England and Wales Court of Appeal (Civil Division) Decisions

IN THE COURT OF APPEAL (CIVIL DIVISION)
ON APPEAL FROM THE HIGH COURT OF JUSTICE
CHANCERY DIVISION
INTELLECTUAL PROPERTY
COMMUNITY TRADE MARK COURT
The Hon Mr Justice Arnold
[2013] EWHC 1291 (Ch)

		Royal Courts of Justice Strand, London, WC2A 2LL
		05/11/2014

Lord Justice Kitchin:

1. This is the judgment of the court upon an appeal by the defendant ("M & S") against the judgments of Arnold J given on 21 May, 23 May and 12 June 2013 and his consequential orders made on 23 May and 12 June 2013 whereby he entered judgment for the claimants (collectively "Interflora") on their claim for infringement of their national and Community registered trade marks for the word "interflora". M & S also appeals against an earlier judgment of Arnold J given on 15 April 2013 and his consequential order made on that day dismissing the objections made by M & S to the admission of the evidence contained in various academic articles referred to in a Civil Evidence Act notice served by Interflora on 25 March 2013.
2. The activities of M & S which have given rise to these proceedings concern the display on the internet of M & S advertisements in response to the entry into the Google search engine by internet users of search terms consisting of or comprising the word "interflora" or minor variants of it.

Legal framework

3. Interflora alleged that M & S's activities amounted to an infringement of their trade marks under Article 5(1)(a) and 5(2) of First Council Directive 89/104 EEC of 21 December 1988 to approximate the laws of Member States relating to trade marks and Article 9(1)(a) and 9(1)(c) of Council Regulation 40/94 of 20 December 1993 on the Community trade mark. Directive 89/104 and Regulation 40/94 have since been repealed by, respectively, Directive 2008/95/EC of 22 October 2008 which came into force on 28 November 2008, and Regulation 207/2009 of 26 February 2009 which came into force on 13 April 2009. However, the alleged infringing activities began when Directive 89/104 ("the Directive") and Regulation 40/94 ("the Regulation") were in force and since the relevant provisions are in all material respects the same, Arnold J approached the allegations on the basis of the original provisions, and we shall do the same.
4. The tenth recital in the preamble to the Directive reads:

"Whereas the protection afforded by the registered trade mark, the function of which is in particular to guarantee the trade mark as an indication of origin, is absolute in the case of identity between the mark and the sign and goods or services; whereas the protection applies also in case of similarity between the mark and the sign and the goods or services; whereas it is indispensable to give an interpretation of the concept of similarity in relation to the likelihood of confusion; … the likelihood of confusion … constitutes the specific condition for such protection."

5. The seventh recital in the preamble to the Regulation is cast in almost identical terms.
6. Article 5 of the Directive provides:

"1. The registered trade mark shall confer on the proprietor exclusive rights therein. The proprietor shall be entitled to prevent all third parties not having his consent from using in the course of trade:

(a) any sign which is identical with the trade mark in relation to goods or services which are identical with those for which the trade mark is registered;

(b) any sign where, because of its identity with, or similarity to, the trade mark and the identity or similarity of the goods or services covered by the trade mark and the sign, there exists a likelihood of confusion on the part of the public, which includes the likelihood of association between the sign and the trade mark.

2. Any Member State may also provide that the proprietor shall be entitled to prevent all third parties not having his consent from using in the course of trade any sign which is identical with, or similar to, the trade mark in relation to goods or services which are not similar to those for which the trade mark is registered, where the latter has a reputation in the Member State and where use of that sign without due cause takes unfair advantage of, or is detrimental to, the distinctive character or the repute of the trade mark.

3. The following, inter alia, may be prohibited under paragraphs 1 and 2:

(a) affixing the sign to the goods or to the packaging thereof;

(b) offering the goods, or putting them on the market or stocking them for these purposes under that sign, or offering or supplying services thereunder;

(c) importing or exporting the goods under the sign;

(d) using the sign on business papers and in advertising. …"

7. The wording of Article 9(1)(a) and (b) of the Regulation corresponds to that of Article 5(1) of the Directive. The wording of paragraph 2 of Article 9 corresponds to that of paragraph 3 of Article 5. Article 9(1)(c) of the Regulation says:

"A Community trade mark shall confer on the proprietor exclusive rights therein. The proprietor shall be entitled to prevent all third parties not having his consent from using in the course of trade:

…

(c) any sign which is identical with or similar to the Community trade mark in relation to goods or services which are not similar to those for which the Community trade mark is registered, where the latter has a reputation in the Community and where use of that sign without due cause takes unfair advantage of, or is detrimental to, the distinctive character or the repute of the Community trade mark."

The background

8. The factual background is set out by the judge in great detail in his judgment of 21 May 2013 ([2013] EWHC 1291 (Ch)) from [18] to [165]. The following is a summary of those matters necessary to understand the issues arising on this appeal.
9. Interflora operate the Interflora flower delivery network in the United Kingdom. The network comprises over 1,600 members who conduct business through over 1,800 florist shops. Customers may contact one of the network florists and place an order, either for delivery in the local area by that florist or for transmission to another network florist out of the area if the recipient is further afield. A substantial number of orders are also placed via the internet using Interflora's website and by telephone using Interflora's central telephone numbers. Such orders are then fulfilled by the network florist closest to the place where the flowers are to be delivered.
10. Interflora have always had a substantial reputation both in the United Kingdom and in other member states of the European Union for their flower delivery services. The Interflora network florists trade under their own brand names but also under the trade mark Interflora. They therefore have their own reputations under their own names and also benefit from (and contribute to) the reputation attaching to the Interflora brand.
11. In recent years, the second claimant ("IBU") has increasingly developed commercial relationships with third parties, including supermarkets. So, from 1995 to about 2000 IBU had a relationship with Sainsbury's that involved a measure of co-branding in selected Sainsbury's stores. In 2005 IBU entered into a relationship with the Co-Op that continues today and in which the Co-Op offers to its funeral clients the opportunity to order online floral tributes from Interflora, again with a measure of co-branding. In November 2009 IBU entered into a relationship with Tesco which again continues today and in which IBU and Tesco operate together a co-branded website.
12. M & S is one of the best known retailers in the United Kingdom. It trades under the brand names "Marks & Spencer" and "M & S" through over 700 stores and, since 1999, through the M & S website. For many years M & S sold goods under the brand name "St Michael" but in 2000 it dropped that name and since then has used "Marks & Spencer" as its main brand name. For very many years M & S has also sold goods under various subsidiary brand names such as "Autograph" and "Per Una" and, since 2009, it has sold goods of other undertakings under their own brand names, particularly in its food halls.
13. M & S has sold flowers in its stores since 1986. It began operating a telephone flower delivery service in 1992 and has provided an on-line flower delivery service since 2000. It is not and has never been part of the Interflora network and, as at May 2008, it did not have a strong reputation for on-line flower delivery, at least compared to Interflora.
14. M & S advertises its flowers on the internet and one of the ways it does this is by use of the Google AdWords service. In broad outline, this service works in the following way. A user of the Google search engine who has carried out a search is presented with a search engine results page or SERP which usually contains three elements. The first is the search box which contains the search term, a word or phrase typed in by the user. The second contains links to websites which appear to the Google search engine to correspond to the search term. These are known as the "natural" or "organic" results of the search and are usually displayed in order of relevance. The third comprises links, referred to as "sponsored links", to websites which are displayed because the operators of those websites have paid for them to appear. They are, in effect, advertisements and are usually displayed in one or more parts of the SERP, namely in a shaded box in the upper part of the page (above the natural search results), in a panel on the right hand side of the page (to the right of the natural search results), or in a panel at the bottom of the page (under the natural search results). Over the years the labelling of these sponsored links has varied, but they have generally been headed with the words "Sponsored Links" or "Ads" or variations of them.
15. A sponsored link appears when a user enters one or more particular words in to the search engine through the search box. These words, known as "keywords", are secured by the advertiser in return for a fee. This is called "purchasing" or "bidding on" the keyword. The sponsored link contains three elements. The first is an underlined heading which functions as a hyper-link to the advertiser's website. This may or may not contain the keyword. The second is a short commercial message or advertisement which, once again, may or may not contain the keyword. The third is the Uniform Resource Locator or URL of the advertiser's website.
16. The fee paid by the advertiser is calculated on the basis of the number of times users click on the hyper-link to the advertiser's website (a process known as "click through") subject to a maximum daily limit which the advertiser has specified. A number of different advertisers may bid for the same keywords, and, if they do, the order in which their adverts are displayed will depend upon various factors including the maximum daily sum, often referred to as the "maximum price per click", that each is prepared to pay.
17. The search term entered by a user of the Google search engine and the keyword selected by the advertiser do not have to be identical for the sponsored link to appear on the SERP. One reason for this is that advertisers have the ability to choose different match types in relation to each keyword, and these match types govern the circumstances in which the sponsored links associated with those keywords will appear. For example, "exact match" requires the search term to be identical to the keyword; "phrase match" requires the search term to contain the same words as the keyword in the same order, but it may include additional words before or after the keyword; and "broad match" simply requires the search term to correspond to variations of the keyword, such as plurals. By 2008 Google had also introduced a development of broad matching known as "advanced broad matching" which causes the sponsored link to appear if the Google search engine deems the keyword relevant to the search term. So, by way of illustration, a search for the term "flowers" might be deemed to match the keyword "florists".
18. There is another feature of the AdWords service which we should mention at this stage, and which was introduced by Google to allow advertisers to limit the appearance of their adverts in response to search terms which were merely related to their keywords. This feature, called "negative matching", allows advertisers to specify search terms in response to which their advertisements will not appear.
19. Until May 2008 Google operated a policy in relation to AdWords whereby a trade mark owner could notify Google that it had registered a particular word as a trade mark. Once notified, Google would not allow that word to be purchased by another trader as a keyword. However, in May 2008 Google changed its policy for the United Kingdom and from that time permitted other traders to purchase words registered as trade marks as keywords, including for use in relation to goods or services for which those trade marks were registered. This means that rival traders can use keywords such as "interflora" with the result that, when a user enters the word "interflora" into the search engine, the advertisements displayed on the SERP include the advertisements of those rivals for goods and services covered by that trade mark.
20. Two further points should, however, be noted. First, Google does allow trade mark owners to block the use of trade marks in advertisements appearing on the SERP. Second, prior to 2008 it was perfectly permissible for rival traders such as M & S to purchase a generic term such as "flowers" with the result that a user inputting a search term such as "interflora flowers" would see upon the SERP not just natural search results for Interflora but also advertisements for the goods and services of those rival traders.
21. Since 2003 M & S has bid on a large number of generic keywords in its flowers campaigns. But in 2008 it seized the opportunity offered by the change in Google's policy and began bidding on the word "interflora". In addition, it bid on a number of close variants of the word "interflora", a number of combinations of interflora with a descriptive term and Interflora's domain names and URLs and variants of them; and it also bid on these and other interflora related terms using the broad and advanced broad match facilities. All of these keywords have triggered the display of advertisements having the following general format:

"M&S Flowers Online

Beautiful Fresh Flowers & Plants.

Order by 5.00pm for Next Day Delivery.

www.marksandspencer.com/flowers"

As one might expect, the text of the advertisements has, however, been modified from time to time to reflect the seasons, price promotions and special days, such as Mother's Day.

22. There can be no doubt this competitive bidding activity by M & S has been successful and has yielded substantial sales revenues. It is also notable that "interflora" and variants of it have been significantly more successful as keywords for M & S than other brand names such as "DFS" and "Laithwaites".

The proceedings

23. Interflora reacted immediately and, on 7 May 2008, sent a letter alleging trade mark infringement and passing off. On 3 December 2008 they began these proceedings alleging infringement of the following registered trade marks ("the Trade Marks") of which the first claimant is the registered proprietor and IBU is the licensee:

i) "United Kingdom Trade Mark No. 1329840 INTERFLORA registered with effect from 16 December 1987 in respect of various goods and services in classes 16, 31, 35, 38, 39, 41 and 42. These include "natural plants and flowers" in class 31, "advertising services … provided for florists" and "information services relating to the sale of … flowers" in class 35, "transportation of flowers" in class 39.

ii) Community Trade Mark No. 909838 INTERFLORA registered with effect from 19 August 1998 in respect of various goods and services in classes 16, 31, 35, 38, 39, 41 and 42. These include "natural plants and flowers" in class 31, "advertising services … provided for florists" in class 35, "transportation of flowers" in class 39 and "information services relating to the sale of … flowers" in class 42."

24. As we have mentioned, Interflora alleged that M & S's activities amounted to an infringement of the Trade Marks under Article 5(1)(a) and 5(2) of the Directive and Article 9(1)(a) and 9(1)(c) of the Regulation.
25. On 16 July 2009 Arnold J made an order referring ten questions to the Court of Justice under what was then Article 234 EC for reasons which he gave in a judgment dated 22 May 2009 ([2009] EWHC 1095 (Ch), [2009] ETMR 59). Following a request from the Court for clarification, Arnold J gave a further judgment dated 29 April 2010 ([2010] EWHC 925) in which he explained that he would withdraw the fourth to tenth questions in the light of the judgment of the Grand Chamber of the Court of Justice in Joined Cases C-236/08 to C-238/08 Google France SARL v Louis Vuitton Malletier SA [2010] ECR I-0000 ("Google France") which was delivered on 23 March 2010, but that he would maintain the first to fourth questions, subject to a slight clarification of the third question.
26. The Court of Justice gave judgment on 22 December 2011 (Case C-323/09 Interflora Inc v Marks & Spencer plc [2011] ECR I-0000, [2012] FSR 3 ("Interflora (CJEU)") in terms which are entirely consistent with a series of decisions the Court has given in other keyword advertising cases. The Court also gave guidance as to particular matters which the national court might choose to take into account in carrying out its assessment, as we will explain.
27. On 24 February 2012 Arnold J gave directions for trial. So far as material to this appeal, the judge gave the parties permission to apply to adduce expert evidence; directed that neither party had permission to adduce survey evidence without first obtaining the permission of the court; and gave Interflora permission to amend their claim to introduce an allegation of infringement based upon the use by M & S of the advanced broad matching facility without negatively matching the term "interflora".
28. At a hearing before Arnold J on 20 June 2012, Interflora sought permission to introduce evidence in the form of witness statements from an unspecified number of individuals selected from nearly 200 respondents gathered by means of two surveys they had carried out in February and March 2012. Arnold J acceded to that application. M & S appealed to this court and, for the reasons set out in a judgment given on 20 November 2012 ([2012] EWCA Civ 1501, [2013] FSR 21 ("Interflora (CA I)"), that appeal was allowed. The issues raised by the appeal concerned the admissibility of evidence obtained by such a survey where the trade mark proprietor does not intend to rely upon the survey itself. In the course of his judgment, Lewison LJ (with whom Hughes and Etherton LJJ agreed) gave extensive consideration to the average consumer test, and his reasoning has been heavily relied upon by M & S on this appeal.
29. Undaunted, on 22 January 2013, Interflora made a further application for permission to introduce witness statements from 13 individuals identified by means of a survey carried out in September and October 2012. Arnold J was again persuaded that permission should be given for reasons he gave in his judgment given on 21 February 2013 ([2013] EWHC 273 (Ch)). However, once more, an appeal against his order was allowed by this court ([2013] EWCA Civ 319, [2013] FSR 26 ("Interflora (CA II)")). Lewison LJ (with whom Sir Robin Jacob and Longmore LJ agreed) emphasised that evidence of this kind should only be admitted if it was of real value and even then only if the value justified the cost. The evidence did not satisfy this test and the judge had been misled by the impression created by the witness statements when compared to the raw data on which they were based.

The trial

30. The action came on for trial before Arnold J on 15 April 2012. On the first day the judge heard argument upon an objection by M & S to the evidence sought to be introduced by Interflora under the Civil Evidence Act notice which Interflora had served upon M & S on 25 March 2013, shortly before trial. The notice identified 18 articles and reports. Some had been produced by Ofcom, Google and YouGov. Others were academic articles of a broadly scientific nature which had been published in a variety of media, including academic journals. M & S objected to some but not all of the documents in the notice. For example, it did not object to the Ofcom reports. But it did object to most of the academic articles, arguing that they amounted to expert evidence which was only admissible in accordance with CPR Part 35, and Interflora had neither sought nor obtained the permission of the court to rely upon it, as they were bound to do under the provisions of that rule. In the alternative, it contended that the court should exercise its case management powers to exclude them.
31. The judge gave judgment that same day ([2013] EWHC 936 (Ch)) (the "judgment on the admissibility of evidence"). He rejected M & S's objections, holding that these documents did not fall within the scope of Part 35, that they might shed light on the attributes of reasonably well-informed and reasonably observant internet users, and that while he had sympathy with the concerns of M & S that it had only been given notice of the material at a very late stage of the proceedings and although he did not know what a wider search of the literature might have thrown up, this did not make it so unfair that the evidence should be excluded, particularly since M & S had known for some time that the court would be interested in receiving independent evidence from public sources, and it had only itself to blame for not investigating the matter sooner. As we shall explain, M & S says that he fell into error in so doing.
32. On the second day of the trial the judge heard argument upon an objection by M & S to the admission of statements of opinion by Mr Rose, an employee of the online marketing company which managed Interflora's keyword advertising. He had expressed the view in his witness statement that most people did not appreciate the difference between natural search results and sponsored links. The judge upheld the objection, saying that he would attach no weight to Mr Rose's evidence so far as he sought to express opinions about the perceptions of persons other than himself. In consequence he was not cross-examined upon that evidence by counsel for M & S. One of M & S's complaints upon this appeal is that, notwithstanding this ruling, the judge proceeded to rely upon the evidence of Mr Rose in reaching his conclusion on the infringement issue.
33. Again on the second day of the trial, the judge was faced with another dispute, this time concerning the evidence of Mr Barringer, the Marketing Director of IBU and the person who was at all material times in charge of all of Interflora's marketing and advertising in the United Kingdom. Mr Barringer's witness statement dated 25 March 2013 contained perfectly admissible evidence about Interflora's flower delivery business and their marketing efforts. However, Mr Barringer exhibited to his statement a report which Interflora had commissioned for this litigation from an organisation called Experian Hitwise ("Hitwise"), and he expressed opinions in his statement as to the conclusions which he considered could be drawn from it. M & S objected to the admission of these expressions of opinion by Mr Barringer and sought an order that it be permitted to cross-examine the author of the Hitwise report, a Mr Anang Pandya.
34. The judge acceded to this application, determining that Interflora would not be permitted to rely upon Mr Barringer's opinions; nor would they be permitted to rely upon the Hitwise report unless Mr Pandya was made available for cross-examination. The judge also gave Interflora permission to issue a witness summons and to serve it upon Mr Pandya requiring his attendance at trial. Mr Pandya duly did attend on the fourth day of the trial and was cross-examined. M & S says upon this appeal that it became clear from this cross-examination that Mr Pandya considered, for good reason, that the Hitwise report provided no basis for inferring the existence of confusion in the minds of internet users arising from the use by M & S of the word "interflora" as a keyword, yet that is what the judge proceeded to use it for, and that in so doing he again fell into material error. So this is another matter to which we must return.

The main judgment

35. On 21 May 2013 the judge delivered his main judgment ([2013] EWHC 1291 (Ch)) (the "main judgment"). It is a substantial and very carefully considered document, extending to 320 paragraphs. At this stage we do no more that set out an outline of its structure and the judge's main conclusions.
36. The first sections of the judgment are concerned with the judge's assessment of the witnesses that he heard and the factual background. It is primarily in these parts of the judgment that consideration is given to the various evidential matters to which we have referred and about which M & S complains.
37. The judge then embarked upon a detailed consideration of the law and addressed the concept of the average consumer (from [194] to [224]) and the jurisprudence of the Court of Justice on keyword advertising (from [225] to [281]). The large part of the analysis of this jurisprudence (from [225] to [267]) is concerned with the effect of keyword advertising upon the origin function of a trade mark, and in the course of it the judge expounded his view, first expressed in Datacard v Eagle Technologies Ltd [2011] EWHC 244 (Pat), [2011] RPC 17, that the test under Article 5(1)(a) of the Directive (Article 9(1)(a) of the Regulation) is a test of a likelihood of confusion but with a reversed onus, that is to say that the onus, or at least the evidential onus, lies upon the person accused of infringement to show that the use complained of is not objectionable.
38. The judge turned (from [268] to [269]) to the effect of keyword advertising upon the advertising function of a trade mark and explained that, as was common ground, this function could not be affected by use of a sign in keyword advertising of the kind in issue. An analysis of the effect of keyword advertising upon the investment function of a trade mark follows from [270] to [274], and the judge concluded (at [274]) that this function would be adversely affected by an adverse effect upon the reputation of the trade mark.
39. Article 5(2) of the Directive and Article 9(1)(c) of the Regulation are considered from [275] to [281]. The judge explained that, if Interflora failed to establish that the use complained of had an effect upon the origin function then, in his view, the case that it caused detriment to the distinctive character of the mark must also fail. As for unfair advantage, the judge summarised the guidance given by the Court of Justice in Interflora (CJEU) as being that an advertiser who selects a trade mark with a reputation as a keyword, and so obtains custom from consumers instead of the trade mark proprietor, obtains a real advantage from the distinctive character and repute of the trade mark and does so without compensating the proprietor. Further, this must be considered to be an unfair advantage in the absence of due cause. However, if the advertisement offers an alternative to the goods or services of the proprietor and does so without offering a mere imitation of them or causing dilution or tarnishment of the trade mark or adversely affecting the functions of the trade mark, then as a rule the use constitutes fair competition and is not without due cause.
40. The final section of the judgment (from [288] to [323]) contains the judge's assessment. Here the judge dealt first (from [288] to [293]) with what he described as general points. He noted the common ground that keyword advertising was not inherently or inevitably objectionable, and that M & S's advertisements must be considered from the perspective of the reasonably well-informed and reasonably observant internet user, initially as at May 2008, the date of the commencement of the activities complained of. The judge then referred to the attempts by Interflora to adduce direct evidence as to consumer reaction to the advertising in issue. He explained that this evidence took the form of the witness statements from witnesses selected from the respondents to the surveys to which we have referred, but that this court had rejected that evidence as having no real value in its Interflora (CA I) and Interflora (CA II) judgments. Conversely, he noted that M & S had made no attempt to adduce any direct evidence of consumer reaction, and that was so despite the fact that M & S was well placed to do so. He put it this way at [293]:

"… Thus M & S had made no attempt to prove that customers who search for, say, "interflora.co.uk" (and thus are clearly looking for the Interflora website), but click through from one M & S's advertisements and end up ordering flowers from M & S, do so because they have been persuaded by the advertisement to order flowers from what they appreciate is a competitive service. M & S's failure to adduce such evidence does not mean that it should be inferred that the evidence would have assisted Interflora; but it does mean that M & S cannot rely upon such evidence to rebut the conclusions which Interflora say should be drawn from the evidence which is before the court. "

41. This brought the judge to the heart of his reasoning and from [294] to [318] he dealt with the case under Article 5(1)(a) (and Article 9(1)(a)) and the effect of the keyword advertising in issue on the origin function of the Trade Marks. He began (at [295] to [297]) by considering three factors identified by the Court of Justice as relevant to the assessment. The first was whether the reasonably well-informed and reasonably observant internet user was deemed to be aware that M & S's flower delivery service was not part of the Interflora network, but was in competition with it. The judge was not satisfied that this was generally known in May 2008; nor was he satisfied that it was generally known at the date of his judgment.
42. The second was whether M & S's advertisements enabled the reasonably well-informed and reasonably observant internet user to tell that M & S's flower delivery service was not part of the Interflora network. Again, the judge did not accept that they did.
43. The third was the nature of the Interflora network. Here the judge agreed with the Court of Justice that the nature of the Interflora network might make it particularly difficult for the reasonably well-informed and reasonably observant internet user to determine, in the absence of any indication in the advertisement, whether M & S's service was part of that network or not.
44. The judge then turned (from [299] to [310]) to six particular matters relied upon by Interflora. The first was that M & S had achieved much better results by bidding on Interflora-related keywords than it had by bidding on other brands as keywords. Here the judge accepted Interflora's submission that this called for an explanation, and further, that the only credible one lay in the nature of the brands: because Interflora signified a network, whereas the other brands did not, consumers were more likely to believe that there was a connection between M & S and the brand the consumers had searched for.
45. The second matter was the analysis of the Hitwise data conducted by Mr Pandya. As we have mentioned, this is a particular focus of M & S's submissions on this appeal and we deal with it in detail later in this judgment. At this stage we need only say that M & S's complaint has two separate but related limbs; the first is that the judge relied upon the Hitwise data at all, and the second is that the judge injected into his analysis the notion of initial interest confusion.
46. The third, fourth and fifth matters can be taken quite shortly because the judge felt they added little to Interflora's case. They were, respectively, various other Hitwise reports which IBU had obtained in the ordinary course of its business; the fact that M & S had given disclosure from its customer records of six possible instances of confusion; and the fact that the growth in M & S's flower business since 2008 was roughly equal to the decline in Interflora's revenue over the same period.
47. The sixth matter concerned the acceptance by many M & S witnesses that consumers searching for "interflora" were looking for Interflora, that is to say they were not using "interflora" as a generic term. M & S contended at trial that consumers who then clicked on M & S advertisements had changed their minds. The judge reiterated that M & S had not adduced any evidence in support of this contention but that did not mean he could not make his own assessment, which he proceeded to do in these terms (at [310]):

"… I am required to consider the impact of M & S's advertisements viewed in the context of the SERPs on reasonably well-informed and reasonably observant internet users. The evidence in this case enables me to put myself in the position of the reasonably well-informed and reasonably observant internet user. I have no doubt that the majority of consumers who click on M & S's advertisements do so because they have been persuaded to take their custom to M & S and not because they believe that M & S is part of the Interflora network. That does not exclude the possibility that a significant proportion of consumers do believe that there is a connection, however."

48. For its part, M & S relied upon three matters. The first was a submission that Interflora was simply a brand like any other and, perhaps more importantly, that the absence of Interflora branding was sufficient to indicate that transactions undertaken outside the Interflora network by florists who were members of Interflora were unconnected with the Interflora network. The judge rejected the first aspect of this submission, and rightly so, for Interflora signifies a network. As for the second aspect, the judge considered the position was not so clear-cut, and went on to explain that customers might take comfort from the fact that a florist was a member of Interflora even if he was not, on that occasion, making use of the network. Further, a customer might place an order for delivery via the network without seeing any Interflora branding.
49. The second matter concerned the business activities of the second defendant, "Flowers Direct", a company which, at the time proceedings were commenced, sold and delivered flowers in competition with Interflora and also bid on "interflora" and other similar terms as keywords. In 2009 Interflora and Flowers Direct settled the dispute between them on the basis that Flowers Direct would not bid on the term "interflora" or variants of it unless and until there was a decision of a court of competent jurisdiction that such bidding did not constitute use of a sign. Subsequently Flowers Direct was acquired by the Flying Brands group of companies and it seems that it may thereafter have resumed bidding on "interflora" and similar terms, although to an extent which was unclear. Then, in 2009, the business was acquired by IBU. In all these circumstances the judge did not consider the activities of Flowers Direct to be of much assistance.
50. The third and final matter related to the effect of bidding on generic terms. As we have explained, both before and after Google's change of policy in May 2008, advertisers bid on generic keywords such as "flowers" with the result that searches for such terms resulted in advertisements being displayed to users with a variety of alternative brand offerings. So also, searches for a combination of a brand name and a generic term, such as "Asda flowers", produced the same result. The judge felt there were two problems with this submission. The first was that, although the majority of internet users had always been aware of the distinction between natural search results and paid advertisements, a significant proportion of internet users had never appreciated that distinction. The second was that a significant proportion of consumers would not have been aware of Google's change in policy and would not have worked out for themselves that on searching for "interflora" they were being presented with advertisements for competitive brands.
51. The judge then came to his overall conclusion on this aspect of the infringement claim:

"318. Conclusion. Taking into account the factors mentioned by the CJEU, the factors relied upon by Interflora and the factors relied upon by M & S, the conclusion I have reached is that, as at 6 May 2008, the M & S advertisements which are the subject of Interflora's claim did not enable reasonably well-informed and reasonably attentive internet users, or enabled them only with difficulty, to ascertain whether the service referred to in the advertisements originated from the proprietor of the Trade Marks, or an undertaking economically connected with it, or originated from a third party. On the contrary, as at 6 May 2008, a significant proportion of the consumers who searched for "interflora" and the other Signs, and then clicked on M & S's advertisements displayed in response to those searches, were led to believe, incorrectly, that M & S's flower delivery service was part of the Interflora network. Thus M & S's use of the Signs had an adverse effect on the origin function of the Trade Marks. Furthermore, I conclude that this is still the case even now. It follows that M & S has infringed the Trade Marks under Article 5(1)(a) of the Directive and Article 9(1)(a) of the Regulation. "

52. The judge dealt with the remaining heads of the claim in concise terms and on the assumption that M & S's activities had no adverse effect on the origin function of the Trade Marks. On this basis, he was not satisfied there had been any damage to Interflora's image and consequently he was not persuaded that there had been any adverse effect upon the reputation attaching to the Trade Marks or their investment function. For like reasons he held that the claim under Article 5(2) (and Article 9(1)(c)) must also fail. If there was no effect on the origin function, the claim for dilution could not succeed and the activities of M & S did not, without due cause, take unfair advantage of the distinctive character and repute of the Trade Marks.

The additional judgments

53. A further hearing took place before the judge on 23 May 2013, on which occasion he decided ([2013] EWHC 1430) that M & S must pay 95% of Interflora's costs and that it must post an appropriate notice on its website drawing attention to the main judgment. M & S also applied for permission to appeal on a considerable number of grounds, which application the judge acceded to in part, granting permission on some grounds but refusing it on others. On 11 November 2013, M & S's renewed application to this court for permission on all other grounds was adjourned by Kitchin LJ to the hearing of the substantive appeal.
54. Yet another hearing took place before the judge on 12 June 2013 to decide the form and scope of the injunction. Three issues arose, each of which was dealt with by the judge in his second additional judgment ([2013] EWHC 1484 (Ch)) (the "second additional judgment"). The first concerned Interflora's complaint, introduced by amendment, that M & S had bid not just on "interflora" and derivatives of it but also on other flower-related terms without negatively matching "interflora". As the judge explained (at [7]), Google's advanced broad matching algorithm can have the result that, when a consumer searches for "interflora", an M & S advertisement is displayed because M & S has bid on, for example, "flower delivery". In his main judgment the judge explained (at [178]) that there was no dispute that M & S had used the sign "interflora", for that is what the judge understood from the arguments presented at trial. That was not, however, the case, for M & S did not accept that it used the sign "interflora" when it bid on other flower related keywords without negatively matching " interflora". This misunderstanding had become clear by 15 June and so the judge specifically addressed this issue in the second additional judgment. He came to the conclusion that M & S had still used the sign "interflora", although less directly.
55. The second issue concerned the specificity of the injunction. Interflora sought an injunction in conventional form, that is to say, to restrain infringement of the Trade Marks. M & S responded that such an injunction would be disproportionate or otherwise contrary to Article 3 of Directive 2004/48/EC ("the Enforcement Directive"), and particularly so bearing in mind the limited nature of the infringing activity of which it had been found guilty. The judge concluded Interflora was entitled to a conventional injunction, finding that the courts' normal practice was not necessarily disproportionate or contrary to the Enforcement Directive, particularly since the court could, in an appropriate case, grant declaratory relief, as it did in Hotel Cipriani Srl v Fred 250 Ltd (formerly Cipriani (Grosvenor Street) Ltd) [2013] EWHC 70 (Ch). Further, the judge continued, M & S had deliberately engaged in a course of conduct well knowing there was a risk it might be found to have infringed the Trade Marks, and it had used a long list of signs and a variety of matching algorithms.
56. The final issue concerned the geographical scope of the injunction. Interflora contended they were entitled to an EU-wide injunction to restrain infringement of the Community trade mark and relied in that connection upon the ruling of the Court of Justice in Case C-235/09 DHL Express France SAS v Chronopost SA [2011] FSR 38 that a prohibition against further infringement or threatened infringement of a Community trade mark must, as a rule, extend to the entire area of the EU. This was resisted by M & S. Once again the judge accepted Interflora's contention. He held that M & S had not proved that that use of the "interflora" sign in keyword advertising did not affect, and was not liable to affect, the functions of the Community trade mark in other Member States and so the injunction must be EU-wide.

The appeal – the issues

57. The parties have been represented on this appeal as they were before the judge. Mr Geoffrey Hobbs QC and Miss Emma Himsworth QC have appeared on behalf of M & S and Mr Michael Silverleaf QC and Mr Simon Malynicz have appeared on behalf of Interflora. In the course of his submissions Mr Hobbs has mounted a wide ranging and forceful attack upon the approach adopted and conclusions reached by the judge in his judgment on the admissibility of evidence, his main judgment and his second additional judgment. For his part and with equal vigour and determination, Mr Silverleaf has defended all of the judge's conclusions and much of his reasoning. The arguments of both counsel have been of the highest quality and we are most grateful for all of the help they and those assisting them have provided to us.
58. We are of the view that the grounds of appeal and the submissions of Mr Hobbs fall into various groups. First, Mr Hobbs points to the finding of the judge in the main judgment at [310] to which we have already referred but the material part of which bears repetition:

"I have no doubt that the majority of consumers who click on M & S's advertisements do so because they have been persuaded to take their custom to M & S and not because they believe that M & S is part of the Interflora network."

59. Then, at [316], the judge said this about the awareness of internet users of the distinction between paid advertisements and the natural search results on the SERP:

"I do not doubt that, as at 6 May 2008, the majority of internet users were aware of this distinction, or that, since then, the proportion of internet users who are aware of it has steadily increased."

60. A little later (but in the same paragraph) the judge expressed his conclusion about the advertisements in issue in this way:

"No doubt, many consumers worked out for themselves that they were being presented with advertisements by competitors to the brand they had searched for despite not being told this. Furthermore, I am confident that nowadays the majority of consumers appreciate this. But I consider that a significant proportion do not."

61. Mr Hobbs argues that these findings should have led the judge to reject the infringement claim on a correct application of the reasonably well-informed and reasonably observant internet user test. However, he continues, the judge misinterpreted and misapplied this test and wrongly converted what is, on a correct application of the law, a hypothetical person with an objectively defined standard of knowledge and perspicacity into a population of individuals with varying degrees of knowledge and perspicacity. Further, says Mr Hobbs, the judge wrongly set the bar for inclusion of individuals in this population too low in that he only excluded those who were ill-informed or unobservant, and wrongly held that infringement would be established if the advertisements caused a significant number of the class of persons so defined wrongly to believe that the advertised goods or services of M & S were connected with Interflora.
62. Mr Hobbs next contends that the judge wrongly changed the test for liability from a question about whether M & S's advertisements would enable the reasonably well-informed and reasonably observant internet user to grasp without difficulty that Interflora and M & S were independent into a strict obligation of broader scope requiring M & S to ensure that its advertisements made it clear that its advertised goods and services did not originate from Interflora. Further and importantly, continues Mr Hobbs, the judge wrongly held it was for M & S to disprove infringement by showing that the use of the sign "interflora" was sufficiently clear that there was no real risk of confusion as to the origin of its goods and services.
63. The next group of errors contended for by Mr Hobbs are concerned with initial interest confusion. Here, argues Mr Hobbs, the judge fell into error because he wrongly decided that such initial interest confusion was sufficient to establish liability. So, drawing these threads together, Mr Hobbs argues that the judge wrongly decided that for M & S to negate the infringement allegation, it had to prove the absence of any real risk of initial interest confusion as to the origin of its goods and services on the part of any significant proportion of the population of persons regarded by the judge as being proxy for the reasonably well-informed and reasonably observant internet user.
64. All of these errors are, says Mr Hobbs, accompanied and indeed compounded by serious procedural irregularities in that the judge wrongly allowed Interflora to rely upon the academic articles the subject of his judgment on the admissibility of evidence, and wrongly relied upon the evidence of Mr Rose despite having said that he would not do so. Further, continues Mr Hobbs, the judge failed to have proper regard to the absence of any pleaded or sustainable case for treating the performance of "interflora" as a keyword as informative or probative with respect to the relevant test for liability; the evidence of Mr Pandya which positively established that the analysis and observations in his report were neither informative nor probative; the evidence confirming that competitive brand bidding, including on "interflora", was widespread; the evidence that it was impossible for a commercial transaction involving Interflora to take place without Interflora's involvement being signified by some form of Interflora branding; the evidence that Flowers Direct and Interflora were competing operations and presented as alternatives; the evidence that M & S and its brands were at all times widely used and recognised; and the evidence that internet searchers were used to seeing competing offers when they used generic search terms and composite search terms including a generic term and a brand name. In so far as Mr Hobbs needs formal permission to rely upon some of these grounds, we give it.
65. Turning to the second additional judgment, Mr Hobbs contends that even if, which he does not accept, there was any proper basis for granting relief against M & S, the judge granted an injunction which lacked the necessary degree of specificity and precision; which wrongly imposed an obligation upon M & S to negatively match the search term "interflora"; and which wrongly extended to the whole of the EU. Mr Hobbs continues that the injunction granted by the judge and its requirement for negative matching amounted to an entirely disproportionate interference with M & S's right to trade and its right to freedom of commercial expression.
66. We will deal with all of these contentions, and Mr Silverleaf's robust defence of the judge's conclusions, in turn and will do so under the following headings:

i) the average consumer;

ii) the onus of proof;

iii) initial interest confusion;

iv) procedural irregularities and errors on the evidence;

v) injunctive relief and negative matching.

But before doing so we must say some more about the jurisprudence of the Court of Justice on keyword advertising. We will focus, so far as possible, upon the potential effect of such advertising on the origin function, for Interflora have not sought to challenge the judge's conclusion that if the case under Article 5(1)(a) of the Directive (and Article 9(1)(a) of the Regulation) based upon effect on the origin function were to fail then a case based upon an effect upon any other function or under Article 5(2) (and Article 9(1)(c)) must fail too. These other ways of developing the infringement allegation therefore add nothing to the particular claim before us. However, we cannot limit our consideration to the effect of such advertising upon the origin function in light of the wide ranging nature of the judge's analysis and the grounds of appeal to which it gives rise.

Keyword advertising

67. We begin with the common ground that, in order to establish infringement under Article 5(1)(a) of the Directive (or Article 9(1)(a) of the Regulation), the proprietor of a trade mark must prove that the following six conditions are satisfied, namely (i) there must be use of a sign by a third party within the relevant territory; (ii) the use must be in the course of trade; (iii) it must be without the consent of the proprietor; (iv) it must be of a sign which is identical with the trade mark; (v) it must be in relation to goods or services which are identical with those for which the trade mark is registered; and (vi) it must affect or be liable to affect one of the functions of the trade mark. The need for each of these conditions to be satisfied has been made clear in a series of decisions of the Court of Justice over many years, as the judge explained at [177] of the main judgment.
68. A proprietor of a registered trade mark alleging infringement under Article 5(1)(b) of the Directive (Article 9(1)(b) of the Regulation) must also satisfy six conditions, namely (i) there must be use of a sign by a third party within the relevant territory; (ii) the use must be in the course of trade; (iii) it must be without the consent of the proprietor; (iv) it must be of a sign which is identical with or similar to the trade mark; (v) it must be in relation to goods or services which are identical with or similar to those for which the trade mark is registered; and (vi) it must give rise to a likelihood of confusion. In Specsavers International Healthcare Ltd v Asda Stores Ltd [2012] EWHC Civ 24, [2012] FSR 19, this court identified (at [51] to [52]) a number of important decisions of the Court of Justice which explain the correct approach to the assessment of the likelihood of confusion, and we endorsed a series of key principles adopted and used by the Trade Marks Registry as being sufficient for the determination of many of the disputes coming before it.
69. It is convenient also to mention at this stage the essential elements of a claim for infringement of a registered trade mark under Article 5(2) of the Directive (Article 9(1)(c) of the Regulation). They are: (i) the registered trade mark must have a reputation in the relevant territory; (ii) there must be use of a sign by a third party in the relevant territory; (iii) the use must be in the course of trade; (iv) it must be without the consent of the proprietor; (v) it must be of a sign which is identical with or similar to the trade mark; (vi) it must be in relation to goods or services; (vii) it must give rise to a link between the sign and the trade mark in the mind of the average consumer; (viii) it must give rise to one of three types of injury, that is to say, (a) detriment to the distinctive character of the trade mark, (b) detriment to the repute of the trade mark, or (c) unfair advantage being taken of the distinctive character or repute of the trade mark; and (ix) it must be without due cause.
70. Reverting now to the claim in the present case under Article 5(1)(a) (and Article 9(1)(a)), M & S accepts that each of the first five conditions is satisfied in this case in so far as the claim is based upon the use by M & S of "interflora" as a keyword. However and as we have indicated, M & S has never accepted that it has an obligation to negatively match, and in that regard has always disputed that its failure to exclude a sign from broad or advanced broad matching involves any use of that sign in the course of its trade. This is an issue which we address later in this judgment in considering the judge's second additional judgment.
71. We therefore turn to the jurisprudence of the Court of Justice on the trade mark implications of keyword advertising and begin with the decision of the Grand Chamber in Google France to which we have already referred. In two of the three cases before the Court the trade mark owner advanced a claim only against Google France. In the third, the trade mark owner advanced claims against both Google France and the advertiser. The Court first considered the issue of use of the sign in the course of trade. So far as the advertiser was concerned, the position was relatively clear. It chose and selected as a keyword a sign identical to the trade mark. This constituted use of the sign in the course of trade:

"51. With regard, first, to the advertiser purchasing the referencing service and choosing as a keyword a sign identical with another's trade mark, it must be held that that advertiser is using that sign within the meaning of that case law.

52. From the advertiser's point of view, the selection of a keyword identical with a trade mark has the object and effect of displaying an advertising link to the site on which he offers his goods or services for sale. Since the sign selected as a keyword is the means used to trigger that ad display, it cannot be disputed that the advertiser indeed uses it in the context of commercial activity and not as a private matter."

72. By contrast, the reference service provider, here Google France, though carrying on a trade, was not itself using the sign:

"56. In that regard, suffice it to note that the use, by a third party, of a sign identical with, or similar to, the proprietor's trade mark implies, at the very least, that that third party uses the sign in its own commercial communication. A referencing service provider allows its clients to use signs which are identical with, or similar to, trade marks, without itself using those signs."

73. Further, the Court considered, and just as the use by an advertiser, in a comparative advertisement, of a sign identical to the trade mark of a competitor for the purposes of identifying the goods or services of the competitor and comparing its own goods or services with them, constitutes use of the sign in relation to the goods or services for the purposes of Article 5(1)(a), so too does use by an advertiser of such a sign as a keyword for use in an internet referencing service:

"69. In that situation, characterised by the fact that a sign identical with a trade mark is selected as a keyword by a competitor of the proprietor of the mark with the aim of offering internet users an alternative to the goods or services of that proprietor, there is a use of that sign in relation to the goods or services of that competitor."

74. The Court reiterated (at [77] and [79]) that the exclusive right conferred by a registered trade mark must be reserved to cases in which a third party's use of the sign affects or is liable to affect the functions of the trade mark. It then explained (at [81]) that, in the context of the cases before it, the relevant functions to be examined were the origin and advertising functions before continuing, in relation to the origin function, in these terms:

"82. The essential function of a trade mark is to guarantee the identity of the origin of the marked goods or service to the consumer or end user by enabling him to distinguish the goods or service from others which have another origin (see, to that effect, Case C-39/97 Canon Kabushiki Kaisha v Metro-Goldwyn-Mayer Inc [1998] ECR I-5507; [1999] ETMR 1 at [28], and Case C-120/04 Medion AG v Thomson Multimedia Sales Germany & Austria GmbH [2005] ECR I-8551; [2006] ETMR 13 at [23]).

83. The question whether that function of the trade mark is adversely affected when internet users are shown, on the basis of a keyword identical with a mark, a third party's ad, such as that of a competitor of the proprietor of that mark, depends in particular on the manner in which that ad is presented.

84. The function of indicating the origin of the mark is adversely affected if the ad does not enable normally informed and reasonably attentive internet users, or enables them only with difficulty, to ascertain whether the goods or services referred to by the ad originate from the proprietor of the trade mark or an undertaking economically connected to it or, on the contrary, originate from a third party (see, to that effect, Céline [2007] ETMR 80 at [27] and the case law cited).

85. In such a situation, which is, moreover, characterised by the fact that the ad in question appears immediately after entry of the trade mark as a search term by the internet user concerned and is displayed at a point when the trade mark is, in its capacity as a search term, also displayed on the screen, the internet user may err as to the origin of the goods or services in question. In those circumstances, the use by the third party of the sign identical with the mark as a keyword triggering the display of that ad is liable to create the impression that there is a material link in the course of trade between the goods or services in question and the proprietor of the trade mark (see, by way of analogy, Arsenal Football Club [2003] E.T.M.R. 19 at [56], and Case C-245/02 Anheuser-Busch Inc v Budějovický Budvar, Národní Podnik [2004] ECR I-10989; [2005] ETMR 27 at [60]).

86. Still with regard to adverse effect on the function of indicating origin, it is worthwhile noting that the need for transparency in the display of advertisements on the internet is emphasised in the EU legislation on electronic commerce. Having regard to the interests of fair trading and consumer protection, referred to in recital 29 in the preamble to Directive 2000/31, Article 6 of that directive lays down the rule that the natural or legal person on whose behalf a commercial communication which is part of an information society service is made must be clearly identifiable.

87. Although it thus proves to be the case that advertisers on the internet can, as appropriate, be made liable under rules governing other areas of law, such as the rules on unfair competition, the fact nonetheless remains that the allegedly unlawful use on the internet of signs identical with, or similar to, trade marks lends itself to examination from the perspective of trade mark law. Having regard to the essential function of a trade mark, which, in the area of electronic commerce, consists in particular in enabling internet users browsing the ads displayed in response to a search relating to a specific trade mark to distinguish the goods or services of the proprietor of that mark from those which have a different origin, that proprietor must be entitled to prohibit the display of third-party ads which internet users may erroneously perceive as emanating from that proprietor.

88. It is for the national court to assess, on a case-by-case basis, whether the facts of the dispute before it indicate adverse effects, or a risk thereof, on the function of indicating origin as described at [84] of the present judgment.

89. In the case where a third party's ad suggests that there is an economic link between that third party and the proprietor of the trade mark, the conclusion must be that there is an adverse effect on the function of indicating origin.

90. In the case where the ad, while not suggesting the existence of an economic link, is vague to such an extent on the origin of the goods or services at issue that normally informed and reasonably attentive internet users are unable to determine, on the basis of the advertising link and the commercial message attached thereto, whether the advertiser is a third party vis-à-vis the proprietor of the trade mark or, on the contrary, economically linked to that proprietor, the conclusion must also be that there is an adverse effect on that function of the trade mark."

75. The following general points of principle emerge from this passage. First, the critical question to be answered in such a case is whether the advertisement does not enable normally informed and reasonably observant internet users, or enables them only with difficulty, to ascertain whether the goods or services referred to in the advertisement originate from the proprietor of the trade mark or an undertaking economically connected to it or, on the contrary, originate from a third party.
76. Second, the trade mark proprietor is entitled to prevent the display of third party advertisements which such internet users may erroneously perceive as emanating from that proprietor or which suggest that there is a material link in the course of trade between the goods or services in question and the proprietor.
77. Third, if the advertisement, though not suggesting an economic link, is vague as to the origin of the goods or services in question so that such internet users are unable to determine, on the basis of the advertising link and the commercial message attaching to it, whether the advertiser is a third party or, on the contrary, is economically linked to the proprietor, then this will have an adverse effect on the origin function of the trade mark.
78. The Court then went on to consider whether keyword advertising was in this context liable to have an effect upon the advertising function of a trade mark and concluded (at [98]) that it was not.
79. In Case C-278/08 Die BergSpechte Outdoor Reisen und Alpinschule Edi Kobmüller GmbH [2010] ECR I-2517, [2010] ETMR 592, the Court of Justice was asked, in substance, whether Article 5(1) of the Directive must be interpreted as meaning that the proprietor of a trade mark is entitled to prohibit a third party from displaying, or arranging for the display of, on the basis of a keyword identical with, or similar to, that trade mark which that third party has, without the consent of that proprietor, selected or stored in connection with an internet referencing service, an advertisement for goods or services identical with, or similar to, those for which the mark is registered. This question was therefore very similar to that answered by the Court in Google France and to which we have referred save that it now extended to similar signs and marks and similar goods and services.
80. In considering this question the Court confirmed (at [17]-[19]) that the selection by an advertiser of a sign as a keyword in the context of an internet referencing service is the means to trigger the display of its advertisement and so does constitute use of the sign in the course of trade in relation to the goods or services so advertised even if the sign does not appear in the advertisement itself. It then continued (at [20]-[22]) that the proprietor of a trade mark cannot oppose the use in this way of a sign identical or similar to its registered trade unless all of the other conditions set out in Article 5 of the Directive and the case law of the Court are satisfied and, in particular, in the case of a claim under Article 5(1)(a), unless the use of the sign is liable to affect one of the functions of the trade mark; and, in the case of a claim under Article 5(1)(b), unless the use of the sign is liable to cause confusion. Whether Article 5(1)(a) or (b) is applicable depends upon whether the signs in issue are identical with or similar to the trade marks the subject of the claim, and this is a matter for the national court to assess.
81. As for Article 5(1)(a), the Court then restated (at [32]-[37]) its reasoning in Google France. In light of what follows, we should cite [35] to [36]:

"35. In respect of the function of indicating origin, the Court held that the question whether that function is adversely affected when internet users are shown, on the basis of a keyword identical with a mark, a third party's ad depends in particular on the manner in which that ad is presented. The function of indicating the origin of the mark is adversely affected if the ad does not enable normally informed and reasonably attentive internet users, or enables them only with difficulty, to ascertain whether the goods or services referred to by the ad originate from the proprietor of the trade mark or an undertaking economically connected to it or, on the contrary, originate from a third party (Google France [2010] ETMR 30 at [83] and [84]).

36. On that point the Court also stated that, in the case where a third party's ad suggests that there is an economic link between that third party and the proprietor of the trade mark, the conclusion must be that there is an adverse effect on the function of indicating origin. Similarly, in the case where the ad, while not suggesting the existence of an economic link, is vague to such an extent on the origin of the goods or services at issue that normally informed and reasonably attentive internet users are unable to determine, on the basis of the advertising link and the commercial message attached thereto, whether the advertiser is a third party vis-à-vis the proprietor of the trade mark or, on the contrary, economically linked to that proprietor, the conclusion must also be that there is an adverse effect on that function of the trade mark (Google France [2010] ETMR 30 at [89] and [90])."

82. Turning to Article 5(1)(b), the Court explained:

"38. The risk that the public might believe that the goods or services in question come from the same undertaking or, as the case may be, from economically-linked undertakings, constitutes a likelihood of confusion (see, inter alia, Case C-342/97 Lloyd Schuhabrik Meyer & Co. GmbH v Klijsen Handel BV [1999] ECR I-3819; [1999] ETMR 690 at [17], Case C-120/04 Medion AG v Thomson Multimedia Sales Germany & Austria GmbH [2005] ECR I-8551; [2006] ETMR 13 at [26], and Case C-102/07 Adidas AG v Marca Mode CV [2008] ECR I-2439; [2008] ETMR 44 at [28]).

39. It follows that, should the rule set out in Article 5(1)(b) of Directive 89/104 be applicable to the dispute in the main proceedings, it will be for the national court to hold whether there is a likelihood of confusion when internet users are shown, on the basis of a keyword similar to a mark, a third party's ad which does not enable normally informed and reasonably attentive internet users, or enable them only with difficulty, to ascertain whether the goods or services referred to by the ad originate from the proprietor of the trade mark or an undertaking economically connected to it or, on the contrary, originate from a third party.

40. The points made at [36] of this judgment are applicable by analogy."

83. The Court then answered the referred question in these terms:

"41. In the light of the foregoing, the answer to Question 1 is that Article 5(1) of Directive 89/104 must be interpreted as meaning that the proprietor of a trade mark is entitled to prohibit an advertiser from advertising, on the basis of a keyword identical with or similar to that trade mark which that advertiser has, without the consent of the proprietor, selected in connection with an internet referencing service, goods or services identical with those for which that mark is registered, in the case where that ad does not enable an average internet user, or enables that user only with difficulty, to ascertain whether the goods or services referred to therein originate from the proprietor of the trade mark or an undertaking economically connected to it or, on the contrary, originate from a third party."

84. It can be seen that the Court appears here to have drawn no distinction between the general test to be applied under Article 5(1)(a) in considering the impact of the keyword advertising upon the origin function and that to be applied under Article 5(1)(b) in considering the likelihood of confusion. It has also explained that, in assessing the likelihood of confusion, the points made by the Court at [36] are to be applied by analogy. As we shall explain, the judge has, however, expressed the view that he cannot believe that this is what the Court meant to say, and this belief is one of the important matters underpinning his conclusion that the test to be applied under Article 5(1)(a) is the same as that under Article 5(1)(b), but with a reversed onus. At this stage we would simply observe that it is, in our view, striking that the Court has in this answer quite clearly chosen not to draw this distinction between the test to be applied in considering the effect on the origin function of the trade mark under Article 5(1)(a) and that to be applied in assessing the likelihood of confusion under Article 5(1)(b) in the context of keyword advertising of the kind in issue and had it intended there to be such a radical difference as a reversed onus then we believe it rather surprising that it did not say so.
85. The third decision in this series is that of the Court of Justice in Case C-558/08 Portakabin v Primakabin [2010] ECR I-6963. Here the Court was again concerned with the use in the context of an internet referencing service of keywords the same as or similar to a registered trade mark. The Court was asked a number of questions, the first of which was directed to the use in such a service of a sign identical to the trade mark, and its answer (from [26] to [36]) was in all material respects the same as that which it gave in Google France and BergSpechte. The Court went on to explain (from [40] to [44]) that the advertiser used the sign in relation to its goods or services, and that was so whether those goods or services were actually offered for sale in the advertisement itself or only on the advertiser's website to which the internet user was referred on clicking on the link, and further, that the presence or absence in the advertisement of actual offers for sale was not, in general, a decisive factor in the assessment whether the use of the sign was likely to have an adverse effect on the origin function of the trade mark.
86. The fourth question again concerned the use of keywords similar to the registered trade mark, and the Court answered it in very similar terms to that which it had given in BergSpechte:

"51. The risk that the public might believe that the goods or services in question come from the same undertaking or, as the case may be, from economically-linked undertakings, constitutes a likelihood of confusion (see, inter alia, Case C-342/97 Lloyd Schuhfabrik Meyer & Co. GmbH v Klijsen Handel BV [1999] ECR I-3819; [1999] ETMR 690 at [17]; Case C-120/04 Medion AG v Thomson Multimedia Sales Germany & Austria GmbH [2005] ECR I-8551; [2006] ETMR 13 at [26]; and Case C-102/07 Adidas AG v Marca Mode CV [2008] ECR I-2439; [2008] ETMR 44 at [28]).

52. It follows that, should the rule set out in Article 5(1)(b) of Directive 89/104 be applicable, it will be for the national court to determine whether there is a likelihood of confusion when internet users are shown, on the basis of a keyword similar to a trade mark, a third party's ad which does not enable normally informed and reasonably attentive internet users, or enables them only with difficulty, to ascertain whether the goods or services referred to by the ad originate from the proprietor of the trade mark or from an undertaking economically linked to it or, on the contrary, originate from a third party (BergSpechte [2010] E.T.M.R. 33 at [39]).

53. The guidance set out in [35] above is, by analogy, applicable.

54. In the light of all of the foregoing, the answer to the first and fourth questions is that Article 5(1) of Directive 89/104 must be interpreted as meaning that a trade mark proprietor is entitled to prohibit an advertiser from advertising, on the basis of a keyword identical with, or similar to, that mark, which that advertiser has selected for an internet referencing service without the consent of the proprietor, in relation to goods or services identical to those in respect of which the mark is registered, where that advertising does not enable average internet users, or enables them only with difficulty, to ascertain whether the goods or services referred to by the ad originate from the proprietor of the trade mark or from an undertaking economically linked to it or, on the contrary, originate from a third party."

87. A little later, the Court considered a further question concerning a possible defence of the advertiser under Article 6 of the Directive. This provision could only apply in so far as the use by the advertiser was "in accordance with honest practices in industrial or commercial matters" and, in that connection, the Court explained:

"67. That condition is the expression of the duty to act fairly in relation to the legitimate interests of the trade mark proprietor. The issue of whether that condition has been satisfied must be assessed by, inter alia, taking account of the extent to which the use by the third party is understood by the relevant public, or at least by a significant section of that public, as establishing a link between the third party's goods and those of the trade mark proprietor or a person authorised to use the trade mark, and of the extent to which the third party ought to have been aware of that (Case C-245/02 Anheuser-Busch Inc v Budějovický Budvar, Národní Podnik [2004] ECR I-10989; [2005] ETMR 27 at [82] and [83], and Céline [2007] ETMR 80 at [33] and [34]).

68. As has been pointed out in reply to the first and fourth questions, however, use by an advertiser of a sign identical with, or similar to, a trade mark for an internet referencing service comes within Article 5(1) of Directive 89/104 where that use does not enable normally informed and reasonably attentive internet users, or enables them only with difficulty, to ascertain whether the goods or services referred to by the ad originate from the trade mark proprietor or from an undertaking economically linked to it or, on the contrary, originate from a third party.

69. Thus, the circumstances under which a trade mark proprietor is, pursuant to Article 5(1) of Directive 89/104, entitled to prevent an advertiser from using a sign identical with, or similar to, that trade mark as a keyword may, in the light of the case-law referred to in [67] above, easily correspond to a situation in which the advertiser cannot claim that it is acting in accordance with honest practices in industrial or commercial matters, and cannot therefore validly rely on the exception provided for in Article 6(1) of that Directive.

70. In that regard, it must be held, first, that one of the characteristics of the situation referred to in [68] above lies precisely in the fact that the ad is likely to cause at least a significant section of the target public to establish a link between the goods or services to which it refers and the goods or services of the trade mark proprietor or persons authorised to use that trade mark. Secondly, in the event that the national court finds that the ad does not enable average internet users, or enables them only with difficulty, to ascertain whether the goods or services referred to by the ad originate from the trade mark proprietor or from a third party, it is unlikely that the advertiser can genuinely claim not to have been aware of the ambiguity thus caused by its ad. It is the advertiser itself, in the context of its professional strategy and with full knowledge of the economic sector in which it operates, which chose a keyword corresponding to another person's trade mark and which, alone or with the assistance of the referencing service provider, designed the ad and therefore decided how it should be presented.

71. Taking account of those factors, it must be concluded that, in the situation described in [54] and [68] above, the advertiser cannot, in principle, claim to have acted in accordance with honest practices in industrial or commercial matters. It is, however, for the national court to carry out an overall assessment of all the relevant circumstances in order to determine whether there may be evidence to justify a contrary finding (see, to that effect, Case C-100/02 Gerolsteiner & Brunnen GmbH & Co. v Putsch GmbH [2004] ECR I-691; [2004] ETMR 40 at [26], and Anheuser-Busch [2005] ETMR 27 at [84] and the case law cited).

72. In the light of all of the foregoing considerations, the answer to the second question is that art.6 of Directive 89/104 must be interpreted as meaning that, where use by advertisers of signs identical with, or similar to, trade marks as keywords for an internet referencing service is liable to be prohibited pursuant to Article 5 of that directive, those advertisers cannot, in general, rely on the exception provided for in Article 6(1) in order to avoid such a prohibition. It is, however, for the national court to determine, in the light of the particular circumstances of the case, whether or not there was, in fact, a use, within the terms of Article 6(1), which could be regarded as having been made in accordance with honest practices in industrial or commercial matters."

88. In this passage the Court has made clear that where the use by an advertiser of a sign identical to a trade mark as a keyword in an internet referencing service is liable to be prevented under Article 5 of the Directive then the advertiser cannot in general rely upon Article 6 as a defence, and that is so because, in such a case, its advertisement is likely to cause at least a significant section of the relevant public to establish a link between the goods or services to which it refers and the goods or services of the trade mark proprietor or persons authorised to use the trade mark.
89. That brings us to the decision of the Court of Justice on the reference made in this case – Case C-323/09 Interflora (CJEU). In this important decision the Court consolidated and expanded upon its earlier reasoning in response to various questions asked by the judge. By the first group of questions the Court was again asked, in essence, about the proper interpretation of Article 5(1)(a) of the Directive (Article 9(1)(a) of the Regulation) in the context of the use of a sign identical to a trade mark as a keyword for the advertisement using an internet referencing service of goods or services identical to those for which the mark was registered. The Court was also asked whether it was relevant (i) that the advertisement concerned was liable to lead some members of the public to believe, incorrectly, that the advertiser was a member of the trade mark proprietor's commercial network, and (ii) that the provider of the internet referencing service did not permit trade mark proprietors to prevent signs identical to their trade marks being selected as keywords.
90. In answering these questions the Court began by reiterating (at [34]) that, in such a case, the trade mark proprietor was entitled to prevent its use only if it was liable to affect one of the functions of the mark. It then provided the following helpful elaboration of that principle, explaining, as it did so, the rival submissions advanced before it:

"35. Interflora submits that that condition must, in accordance with an already established body of case law, be understood as meaning that paragraph 1(a) of Article 5 of Directive 89/104 and of Article 9 of Regulation 40/94 protects the trade mark proprietor against any adverse effect on any of the functions of the mark. According to M&S, however, such an interpretation does not follow unambiguously from the case law and risks creating an imbalance as between the interest in protecting intellectual property and the interest in free competition. The European Commission, for its part, submits that paragraph 1(a) protects the trade mark proprietor solely against acts affecting the mark's function of providing an indication of origin. In its submission, the trade mark's other functions can, at most, play a role in the interpretation of Article 5(2) of Directive 89/104 and Article 9(1)(c) of Regulation 40/94, which concern the rights conferred by trade marks with a reputation.

36. It follows from the wording of Article 5(1) of Directive 89/104 and from the tenth recital in the preamble thereto that the laws of the Member States have been harmonised inasmuch as the exclusive right conferred by a trade mark affords the proprietor of the mark "absolute" protection against the use by third parties of signs which are identical with that mark in relation to identical goods or services, whilst, where there is not identity on two counts, only the existence of a likelihood of confusion enables the proprietor to rely successfully on its exclusive right. That distinction between the protection conferred by Article 5(1)(a) and that provided for in Article 5(1)(b) was espoused, so far as the Community trade mark is concerned, by the seventh recital to, and Article 9(1) of, Regulation 40/94.

37. Although the EU legislature described as "absolute" the protection against the unauthorised use of signs identical with a trade mark in relation to goods or services identical with those for which the mark is registered, the Court has put that description into perspective by stating that, as extensive as it may be, the protection conferred by Article 5(1)(a) of Directive 89/104 is intended solely to enable the trade mark proprietor to protect its specific interests as proprietor of the mark, that is to say, to ensure that the trade mark can fulfil its functions. The Court has concluded that the exercise of the exclusive right conferred by the trade mark must be reserved to cases in which a third party's use of the sign adversely affects, or is liable adversely to affect, the functions of the trade mark, in particular its essential function of guaranteeing to consumers the origin of the goods (see Case C-206/01 Arsenal Football Club Plc v Reed [2002] ECR I-10273; [2003] R.P.C. 9 at [51]).

38. That interpretation of Article 5(1)(a) of Directive 89/104 has been restated on many occasions and applied in relation to Article 9(1)(a) of Regulation 40/94 (see, inter alia, as regards Directive 89/104, Case C-17/06 Céline Sarl v Céline SA [2007] ECR I-7041; [2007] ETMR 80 at [16], and Case C-533/06 O2 Holdings Ltd v Hutchison 3G UK Ltd [2008] ECR I-4231; [2008] RPC 33 at [57], and, as regards Regulation 40/94, the order in Case C-62/08 UDV North America Inc v Brandtraders NV [2009] ECR I-1279; [2010] ETMR 25 at [42], and Google France [2010] RPC 19 at [75]). There has been further clarification of that interpretation to the effect that those provisions enable the proprietor of the trade mark to rely on its exclusive right where there is, or is liable to be, an adverse effect on one of the functions of the trade mark, irrespective of whether the function concerned is the essential function of indicating the origin of the product or service covered by the trade mark or one of the other functions of the mark, such as that of guaranteeing the quality of that product or service or that of communication, investment or advertising (L'Oréal [2010] RPC 1 at [63] and [65], and Google France [2010] RPC 19 at [77] and [79]).

39. With regard to the functions of the trade mark other than that of indicating origin, it should be noted that both the EU legislature—by using the words "in particular" in the tenth recital to Directive 89/104 and in the seventh recital to Regulation 40/94 —and the Court—by using since its judgment in Arsenal Football Club [2003] R.P.C. 9 the words "functions of the trade mark"—have indicated that a trade mark's function of indicating origin is not the only function of the mark that is worthy of protection against injury by third parties. They have thus taken into account the fact that a trade mark is often, in addition to an indication of the origin of the goods or services, an instrument of commercial strategy used, inter alia, for advertising purposes or to acquire a reputation in order to develop consumer loyalty.

40. Admittedly, a trade mark is always supposed to fulfil its function of indicating origin, whereas it performs its other functions only insofar as its proprietor uses it to that end, in particular for the purposes of advertising or investment. However, that difference between the essential function of the trade mark and its other functions can in no way justify—when a trade mark fulfils one or more of those other functions—excluding from the scope of Article 5(1)(a) of Directive 89/104 and Article 9(1)(a) of Regulation 40/94 acts adversely affecting those functions. Likewise, there are no grounds for holding that only trade marks with a reputation are capable of having functions other than that of indicating origin."

91. In these paragraphs the Court has spelled out a series of important points concerning the scope of Article 5(1)(a) and Article 5(1)(b). First, Article 5(1)(a) affords "absolute" protection against use of signs which are identical to the trade mark in relation to identical goods or services, whereas Article 5(1)(b) extends to the use of similar signs in relation to similar goods or services but only affords protection if there is a likelihood of confusion.
92. Second, although the protection afforded by Article 5(1)(a) is "absolute", the exercise of the right conferred is reserved for cases in which the use of the sign by a third party adversely affects or is liable adversely to affect one of the functions of the trade mark, and that is so whether that function is the essential function indicating the origin of the goods or services covered by the mark, or one of the other functions, such as that of guaranteeing the quality of the goods or services concerned or that of communication, investment or advertising.
93. Third, a trade mark is always supposed to fulfil the origin function but will only fulfil one of the other functions in so far as it has been used to that end. It is for the national court to consider whether the condition of an adverse effect on one of the functions of the trade mark is met.
94. The Court then proceeded to give what it described as interpretative guidance in relation to a trade mark's origin, advertising and investment functions in the context of an internet referencing service, beginning with the origin function. At the outset it reiterated (at [44] to [46]) the guidance it had given in Google France and Portakabin and which we have summarised at [75] to [77] above. It continued (at [47]) that the fact that the referencing service provider had not permitted trade mark proprietors to prevent the use of a sign identical with the trade mark as a keyword was irrelevant. By contrast, the fact that some internet users might be misled might be relevant but was not necessarily sufficient to establish liability:

"48. By contrast, a situation such as that described in question 4(a) [the fact that the sign was liable to mislead some members of the public] may be relevant for the purpose of applying the rule set out in Article 5(1)(a) of Directive 89/104 and Article 9(1)(a) of Regulation 40/94 .

49. Indeed, if the referring court's assessments of the facts were to show that M&S's advertising, displayed in response to searches performed by internet users using the word "interflora", may lead those users to believe, incorrectly, that the flower-delivery service offered by M&S is part of Interflora's commercial network, it would have to be concluded that that advertising does not allow it to be determined whether M&S is a third party in relation to the proprietor of the trade mark or whether, on the contrary, it is economically linked to that proprietor. In those circumstances, the function of the INTERFLORA trade mark of indicating origin would be adversely affected.

50. In that context, as has been observed at [44] of this judgment, the relevant public comprises reasonably well-informed and reasonably observant internet users. Therefore, the fact that some internet users may have had difficulty grasping that the service provided by M&S is independent from that of Interflora is not a sufficient basis for a finding that the function of indicating origin has been adversely affected."

95. The Court next identified (at [51] to [52]) three interconnected factors the national court might choose to assess, namely first, whether the reasonably well-informed and reasonably observant internet user was deemed to be aware, on the basis of general knowledge of the market, that M & S's flower-delivery service was not part of the Interflora network but was in competition with it; secondly, should it become apparent that that was not generally known, whether M & S's advertisement enabled that internet user to tell that the M & S service did not belong to the Interflora network; and third, the fact that the Interflora commercial network was composed of a large number of retailers which varied greatly in terms of size and commercial profile and that, in these circumstances, it might be particularly difficult for the reasonably well-informed and reasonably observant internet user to determine, in the absence of any indication from M & S, whether or not M & S was part of that network. These are matters to which the judge did indeed refer, as we have explained at [41] to [43] above.
96. It was then for the national court to make its assessment, and to do so from the perspective of the reasonably well-informed and reasonably observant internet user:

"53. Having regard to that situation and to the other matters that it may consider relevant, the referring court will, in the absence of any general knowledge such as that referred to at [51] of this judgment, have to determine whether or not the use of words such as "M&S Flowers" in an advertisement such as the one set out at [19] of this judgment is sufficient to enable a reasonably well-informed and reasonably observant internet user who has entered search terms including the word "interflora" to tell that the flower-delivery service offered does not originate from Interflora."

97. Guidance on the effect on advertising function is set forth in the decision from [54] to [59]. The Court restated that the use of a sign identical to another person's trade mark in the Google referencing service did not adversely affect the advertising function of that mark, even though it might have repercussions on the use of the mark in advertising. In this connection the Court recognised that the use of such a sign as a keyword meant that the proprietor of the trade mark might have to pay a higher price per click than the competitor if it wished to ensure that its advertisement appeared before that of the competitor. But this did not necessarily mean that the trade mark's advertising function was adversely affected:

"57. However, the mere fact that the use, by a third party, of a sign identical with a trade mark in relation to goods or services identical with those for which that mark is registered obliges the proprietor of that mark to intensify its advertising in order to maintain or enhance its profile with consumers is not a sufficient basis, in every case, for concluding that the trade mark's advertising function is adversely affected. In that regard, although the trade mark is an essential element in the system of undistorted competition which European law seeks to establish (see, in particular, Case C-59/08 Copad SA v Christian Dior Couture SA [2009] E.C.R. I-3421; [2009] FSR 22 at [22]), its purpose is not, however, to protect its proprietor against practices inherent in competition.

58. Internet advertising on the basis of keywords corresponding to trade marks constitutes such a practice in that its aim, as a general rule, is merely to offer internet users alternatives to the goods or services of the proprietors of those trade marks (see, to that effect, Google France [2010] RPC 19 at [69]).

59. The selection of a sign identical with another person's trade mark, in a referencing service with the characteristics of "AdWords", does not, moreover, have the effect of denying the proprietor of that trade mark the opportunity of using its mark effectively to inform and win over consumers (see, in that regard, Google France [2010] RPC 19 at [96] and [97])."

98. Again this is a significant passage for it shows an appreciation by the Court that internet advertising using keywords which are identical to trade marks is not an inherently objectionable practice. On the contrary, its aim is, in general, to offer to internet users alternatives to the goods or services of trade mark proprietors and it is not the purpose of trade marks to protect their proprietors from fair competition.
99. Similar guidance was given by the Court in relation to the investment function. First, it explained the nature of this function:

"62. When the use by a third party, such as a competitor of the trade mark proprietor, of a sign identical with the trade mark in relation to goods or services identical with those for which the mark is registered substantially interferes with the proprietor's use of its trade mark to acquire or preserve a reputation capable of attracting consumers and retaining their loyalty, the third party's use must be regarded as adversely affecting the trade mark's investment function. The proprietor is, as a consequence, entitled to prevent such use under Article 5(1)(a) of Directive 89/104 or, in the case of a Community trade mark, under Article 9(1)(a) of Regulation 40/94.

63. In a situation in which the trade mark already enjoys such a reputation, the investment function is adversely affected where use by a third party of a sign identical with that mark in relation to identical goods or services affects that reputation and thereby jeopardises its maintenance. As the Court has already held, the proprietor of a trade mark must be able, by virtue of the exclusive right conferred upon it by the mark, to prevent such use (Case C-324/09 L'Oréal SA v eBay International AG [2011] RPC 27 at [83])."

100. But again, the investment function does not provide a means to protect trade mark proprietors against the effects of fair competition, even if such competition means that these proprietors have to adapt their efforts to acquire or preserve a reputation capable of attracting and retaining customers:

"64. However, it cannot be accepted that the proprietor of a trade mark may—in conditions of fair competition that respect the trade mark's function as an indication of origin—prevent a competitor from using a sign identical with that trade mark in relation to goods or services identical with those for which the mark is registered, if the only consequence of that use is to oblige the proprietor of that trade mark to adapt its efforts to acquire or preserve a reputation capable of attracting consumers and retaining their loyalty. Likewise, the fact that that use may prompt some consumers to switch from goods or services bearing that trade mark cannot be successfully relied on by the proprietor of the mark."

101. The Court therefore answered the questions concerning the interpretation of Article 5(1)(a) of the Directive (and Article 9(1)(a) of the Regulation) in these terms:

1. Article 5(1)(a) of First Council Directive 89/104/EEC of 21 December 1988 to approximate the laws of the Member States relating to trade marks and Article 9(1)(a) of Council Regulation 40/94 of 20 December 1993 on the Community trade mark must be interpreted as meaning that the proprietor of a trade mark is entitled to prevent a competitor from advertising—on the basis of a keyword which is identical with the trade mark and which has been selected in an internet referencing service by the competitor without the proprietor's consent—goods or services identical with those for which that mark is registered, where that use is liable to have an adverse effect on one of the functions of the trade mark. Such use:

• adversely affects the trade mark's function of indicating origin where the advertising displayed on the basis of that keyword does not enable reasonably well-informed and reasonably observant internet users, or enables them only with difficulty, to ascertain whether the goods or services concerned by the advertisement originate from the proprietor of the trade mark or an undertaking economically linked to that proprietor or, on the contrary, originate from a third party;

• does not adversely affect, in the context of an internet referencing service having the characteristics of the service at issue in the main proceedings, the trade mark's advertising function; and

• adversely affects the trade mark's investment function if it substantially interferes with the proprietor's use of its trade mark to acquire or preserve a reputation capable of attracting consumers and retaining their loyalty.

102. The Court then considered Article 5(2) of the Directive (Article 9(1)(c) of the Regulation) and explained (at [67]) that it had been asked, in essence, whether these provisions must be interpreted as meaning that the proprietor of a trade mark with a reputation was entitled to prevent a competitor from basing its advertising on a keyword corresponding to that trade mark which the competitor had, without the proprietor's consent, selected in an internet referencing service. A little later (at [75]) it continued that the referring court was not seeking guidance on the concept of detriment to the repute of the mark (tarnishment), but rather wished to ascertain the conditions under which such a competitor who caused to be displayed an advertising link towards its website must be regarded as causing detriment to the distinctive character of the mark with the reputation (dilution) or as taking unfair advantage of the distinctive character or repute of that mark (free-riding).
103. The Court dealt first with the concept of dilution and accepted (at [79]) that the use by a third party in the course of trade of a sign identical with or similar to a trade mark with a reputation reduces the latter's distinctiveness when it contributes to turning the trade mark into a generic term. But, it continued, this is not necessarily the case when such a sign is used as a keyword:

"80. However, contrary to Interflora's contention, the selection of a sign which is identical with or similar to a trade mark with a reputation as a keyword within an internet referencing service does not necessarily contribute to such a development.

81. Thus, when the use, as a keyword, of a sign corresponding to a trade mark with a reputation triggers the display of an advertisement which enables the reasonably well-informed and reasonably observant internet user to tell that the goods or services offered originate not from the proprietor of the trade mark but, on the contrary, from a competitor of that proprietor, the conclusion will have to be that the trade mark's distinctiveness has not been reduced by that use, the latter having merely served to draw the internet user's attention to the existence of an alternative product or service to that of the proprietor of the trade mark."

104. Here the Court has recognised, once again, that the use of a sign as a keyword in the context of an internet referencing service is not an inherently objectionable practice. Fair use of the sign which enables the reasonably well-informed and reasonably observant internet user to tell that the goods or services offered originate not from the proprietor of the trade mark but from a competitor does not dilute the reputation attaching to the mark.
105. Much the same guidance is given in relation to free-riding. The Court accepted that where a competitor of the proprietor of a trade mark with a reputation selects a sign identical to the trade mark as a keyword in an internet referencing service, the purpose of that use is to take advantage of the character and repute of the trade mark, and further, that the competitor will derive a real benefit for which he has not paid. Such use will, if it is "without due cause", amount to an infringement, and that is likely to be the case if the goods are mere imitations of those of the trade mark proprietor. But it will not be the case if the goods are genuine alternatives and the use does not cause dilution or tarnishment or adversely affect one of the functions of the trade mark:

"91. By contrast, where the advertisement displayed on the internet on the basis of a keyword corresponding to a trade mark with a reputation puts forward—without offering a mere imitation of the goods or services of the proprietor of that trade mark, without causing dilution or tarnishment and without, moreover, adversely affecting the functions of the trade mark concerned—an alternative to the goods or services of the proprietor of the trade mark with a reputation, it must be concluded that such use falls, as a rule, within the ambit of fair competition in the sector for the goods or services concerned and is thus not without "due cause" for the purposes of Article 5(2) of Directive 89/104 and Article 9(1)(c) of Regulation 40/94. "

106. All of this guidance is reflected in these answers given by the Court to the second set of questions:

"2. Article 5(2) of Directive 89/104 and Article 9(1)(c) of Regulation 40/94 must be interpreted as meaning that the proprietor of a trade mark with a reputation is entitled to prevent a competitor from advertising on the basis of a keyword corresponding to that trade mark, which the competitor has, without the proprietor's consent, selected in an internet referencing service, where the competitor thereby takes unfair advantage of the distinctive character or repute of the trade mark (free-riding) or where the advertising is detrimental to that distinctive character (dilution) or to that repute (tarnishment).

Advertising on the basis of such a keyword is detrimental to the distinctive character of a trade mark with a reputation (dilution) if, for example, it contributes to turning that trade mark into a generic term. By contrast, the proprietor of a trade mark with a reputation is not entitled to prevent, inter alia, advertisements displayed by competitors on the basis of keywords corresponding to that trade mark, which put forward—without offering a mere imitation of the goods or services of the proprietor of that trade mark, without causing dilution or tarnishment and without, moreover, adversely affecting the functions of the trade mark with a reputation—an alternative to the goods or services of the proprietor of that mark."

The average consumer

107. We have summarised M & S's complaint about the approach adopted by the judge to the reasonably well-informed and reasonably observant internet user test at [58] to [61] above. Mr Hobbs has developed his submissions on this issue in the following way. He contends that the reasoning of the Court of Justice in Interflora (CJEU) unequivocally establishes that, for the purpose of establishing liability under Article 5(1)(a) of the Directive (Article 9(1)(a) of the Regulation), it is not only essential to assess the advertising from the perspective of the reasonably well-informed and reasonably circumspect internet user but also wrong to assess it from the perspective of "at least some members of the public". In this regard Mr Hobbs focuses on [48] to [53] of the decision which (or the substance of which) we have set out at [94] to [96] above. Here, he contends, the Court has expressly rejected the notion that it is enough that "some internet users may have had difficulty in grasping that the service provided by M & S is independent from that of Interflora" and made it perfectly clear that the referring court was required to determine whether the advertising did or did not enable the reasonably well-informed and reasonably circumspect internet user to tell that the flower-delivery service offered did not originate from Interflora.
108. Mr Hobbs continues that this ruling is in accordance with established principles and that, properly understood, the court is faced with a binary question, not a statistical test. It must consider the issue before it from the perspective of the average consumer of the goods or services in question and, from that perspective, answer the question posed by that issue one way or the other. Put another way, in the context of a likelihood of confusion, the average consumer, a hypothetical person, is either liable to be misled or he is not.
109. Further, says Mr Hobbs, this position was recognised by the Court of Justice in considering the likelihood of confusion in Case C-251/95 SABEL BV v Puma AG [1997] ECR I-6191 at [23] and again in Case C-342/97 Lloyd Schuhfabrik Meyer & Co. GmbH v Klijsen Handel BV [1999] ECR I-3819 at [25]. What is more, Mr Hobbs continues, the Court of Justice has emphasised that general abstract data such as predetermined percentages should not be regarded as determinative of either the existence or any relevant incidence of distinctiveness among consumers for any given class or category of goods or services, and in that connection he referred us to Joined Cases C-108/97 and C-109/97 Windsurfing Chiemsee Produktions-und Vertriebs GmbH (WSC) v Boots-und Segelzubehör Walter Huber [1999] ECR I-2779 at [52] and Lloyd Schuhfabrik at [24].
110. Mr Hobbs also argues that this position is reflected in recital 18 of European Parliament and Council Directive 2005/29/EC of 11 May 2005 concerning unfair business-to-consumer commercial practices in the internal market ("the Unfair Commercial Practices Directive"). This reads, with emphasis added:

"It is appropriate to protect all consumers from unfair commercial practices; however the Court of Justice has found it necessary in adjudicating on advertising cases since the enactment of Directive 84/450/EEC to examine the effect on a notional, typical consumer. In line with the principle of proportionality, and to permit the effective application of the protections contained in it, this Directive takes as a benchmark the average consumer, who is reasonably well-informed and reasonably observant and circumspect, taking into account social, cultural and linguistic factors, as interpreted by the Court of Justice, but also contains provisions aimed at preventing the exploitation of consumers whose characteristics make them particularly vulnerable to unfair commercial practices. Where a commercial practice is specifically aimed at a particular group of consumers, such as children, it is desirable that the impact of the commercial practice be assessed from the perspective of the average member of that group. It is therefore appropriate to include in the list of practices which are in all circumstances unfair a provision which, without imposing an outright ban on advertising directed at children, protects them from direct exhortations to purchase. The average consumer test is not a statistical test. National courts and authorities will have to exercise their own faculty of judgement, having regard to the case-law of the Court of Justice, to determine the typical reaction of the average consumer in a given case."

111. Finally, says Mr Hobbs, his argument was accepted by this court in Interflora (CA I) and, he reminds us, this decision is binding upon us. So, he continues, we must consider the advertising from the perspective of the average member of the group of consumers to whom the advertising is directed, and the judge fell into error in failing so to do.
112. There was a certain amount of common ground before the judge, and Mr Silverleaf has not sought to resile from it before us. For our part, we endorse it. First, in the context of internet advertising, the average consumer (who is reasonably well-informed and reasonably observant and circumspect) and the reasonably well-informed and reasonably circumspect internet user are one and the same.
113. Second, the average consumer in any context is a hypothetical person or "legal construct" as Lewison LJ described him in Interflora (CA I) at [44] and [73]. We would add that he is a person who has been created to strike the right balance between various competing interests including, on the one hand, the need to protect consumers and, on the other hand, the promotion of free trade in an openly competitive market, and also to provide a standard, defined in EU law, which national courts may then apply.
114. Third, the average consumer test is not a statistical test. The national court must exercise its own judgment, in accordance with the principle of proportionality and the principles explained by the Court of Justice, to determine the perceptions of the average consumer in any given case in light of all the relevant circumstances.
115. Fourth, and again as Lewison LJ explained in Interflora (CA I) at [45] to [50], in a case concerning ordinary goods or services, the court may be able to put itself in the position of the average consumer without requiring evidence from consumers, still less expert evidence or a consumer survey. In such a case, the judge can make up his or her own mind about the particular issue he or she has to decide in the absence of evidence and using his or her own common sense and experience of the world.
116. So we come to the crucial question, namely whether the average consumer, as a hypothetical person, necessarily has a single reaction so that the average consumer test has, in the shorthand of Mr Hobbs, a binary nature and so precludes a consideration of the perceptions of a proportion of the relevant public. The judge cited a good deal of case law in the course of his analysis. However, we believe the citations which follow and to which we were referred are sufficient to answer the particular question now before us.
117. We begin with Case C-210/96 Gut Springenheide GmbH v Oberkreisdirektor des Kreises Steinfurt – Amt für Lebensmittelüberwachung [1998] ECR I-4657. In this case the Court of Justice was asked to define the nature of the consumer to be used as a standard for determining whether a statement designed to promote the sale of eggs was likely to mislead the purchaser in breach of Regulation (EEC) No 1907/90. In providing its answer, the Court enunciated the average consumer test and how it should be applied in these terms:

"30. There have been several cases in which the Court of Justice has had to consider whether a description, trade mark or promotional text is misleading under the provisions of the Treaty or of secondary legislation. Whenever the evidence and information before it seemed sufficient and the solution clear, it has settled the issue itself rather than leaving the final decision for the national court (see, in particular, Case C-362/88 GB-INNO-BM [1990] ECR I-667; Case C-238/89 Pall [1990] ECR I-4827; Case C-126/91 Yves Rocher [1993] ECR I-2361; Case C-315/92 Verband Sozialer Wettbewerb [1994] ECR I-317; Case C-456/93 Langguth [1995] ECR I-1737; and Case C-470/93 Mars [1995] ECR I-1923).

31. In those cases, in order to determine whether the description, trade mark or promotional description or statement in question was liable to mislead the purchaser, the Court took into account the presumed expectations of an average consumer who is reasonably well-informed and reasonably observant and circumspect, without ordering an expert's report or commissioning a consumer research poll.

32. So, national courts ought, in general, to be able to assess, on the same conditions, any misleading effect of a description or statement designed to promote sales.

.…

34. In Case C-373/90 X [1992] ECR I-131, paragraphs 15 and 16, in which Directive 84/450 was in point, the Court held, inter alia, that it was for the national court to ascertain in the circumstances of the particular case and bearing in mind the consumers to which the advertising was addressed, whether advertising describing cars as new despite the fact that they had been registered for the purposes of importation, without ever having been driven on a road, could be misleading insofar as, on the one hand, it sought to conceal the fact that the cars advertised as new were registered before importation and, on the other hand, that fact would have deterred a significant number of consumers from making a purchase. The Court also held that advertising regarding the lower prices of the cars could be held to be misleading only if it were established that the decision to buy on the part of a significant number of consumers to whom the advertising in question was addressed was made in ignorance of the fact that the lower price of the vehicles was matched by a smaller number of accessories on the cars sold by the parallel importer.

35. The Court has not therefore ruled out the possibility that, in certain circumstances at least, a national court might decide, in accordance with its own national law, to order an expert's opinion or commission a consumer research poll for the purpose of clarifying whether a promotional description or statement is misleading or not.

36. In the absence of any Community provision on this point, it is for the national court, which may find it necessary to order such a survey, to determine, in accordance with its own national law, the percentage of consumers misled by a promotional description or statement that, in its view, would be sufficiently significant in order to justify, where appropriate, banning its use.

37. The answer to be given to the questions referred must therefore be that, in order to determine whether a statement or description designed to promote sales of eggs is liable to mislead the purchaser, in breach of Article 10(2)(e) of Regulation No 1907/90, the national court must take into account the presumed expectations which it evokes in an average consumer who is reasonably well-informed and reasonably observant and circumspect. However, Community law does not preclude the possibility that, where the national court has particular difficulty in assessing the misleading nature of the statement or description in question, it may have recourse, under the conditions laid down by its own national law, to a consumer research poll or an expert's report as guidance for its judgment."

118. We think a number of points emerge from this passage, some of which are confirmatory of the common ground we have summarised. First, the average consumer test provides the court with a perspective from which to assess the particular question it has to decide, for example whether a statement is liable to mislead purchasers. Second, a national court may be able to assess this question without the benefit of a survey or expert evidence. Third, a national court may nevertheless decide, in accordance with its own national law, that it is necessary to have recourse to an expert's opinion or a survey for the purpose of assisting it to decide whether the statement is misleading or not. Fourth, absent any provision of EU law dealing with the issue, it is then for the national court to determine, in accordance with its own national law, the percentage of consumers misled by the statement that, in its view, is sufficiently significant in order to justify banning its use.
119. In Joined Cases C-108/97 and C-109/97 Windsurfing Chiemsee the Court followed broadly the same approach in considering whether a mark had acquired a distinctive character through use. It explained (at [51] to [53]) that, in assessing the distinctive character of a mark, the competent authority may take into account the proportion of the relevant class of persons who, because of the mark, identify goods as originating from a particular undertaking; that if the class of persons, or at least a significant proportion of them, identify the goods as coming from a particular undertaking because of the mark, then the authority must hold that the requirement of distinctive character laid down in Article 3(3) of the Directive is satisfied; but this proportion is not to be assessed by reference to general abstract data, such as predetermined percentages.
120. In Lloyd Schuhfabrik (at [24]) the Court reiterated that is not possible to state in general terms, for example by referring to given percentages relating to the degree of recognition attained by a mark within the relevant section of the public, when a mark has a strong distinctive character. It then explained (at [25]), and as it had in SABEL (at [23]), that the wording of Article 5(1)(b) of the Directive "… there exists a likelihood of confusion on the part of the public" shows that the perception of marks in the mind of the average consumer of the category of goods or services in question plays a decisive role in the global appreciation of the likelihood of confusion. Mr Hobbs submits, and we accept, that the same must apply to the reasonably well-informed and reasonably observant internet user. But in our judgment neither of these cases supports the contention that the national court may not properly have regard to the perceptions of what it considers to be a significant proportion of the relevant public in assessing either distinctiveness or a likelihood of confusion under the provisions of the Directive and the Regulation.
121. We also find instructive this reasoning of the Court of Justice in Case C-299/99 Koninklijke Philips Electronics NV v Remington Consumer Products Ltd [2002] ECR I-5475:

"60. As is clear from paragraph 51 of the judgment in Windsurfing Chiemsee, in assessing the distinctive character of a mark in respect of which registration has been applied for, the following may inter alia also be taken into account: … the proportion of the relevant class of persons who, because of the mark, identify goods as originating from a particular undertaking; …

61. The Court has also held that if, on the basis of those factors, the competent authority finds that the relevant class of persons, or at least a significant proportion thereof, identify goods as originating from a particular undertaking because of the trade mark, it must in any event hold that the requirement for registering the mark laid down in Article 3(3) of the Directive is satisfied (Windsurfing Chiemsee, paragraph 52).

62. However, it must first be pointed out that the Court has made clear that the circumstances in which the requirement under Article 3(3) of the Directive may be regarded as satisfied cannot be shown to exist solely by reference to general, abstract data, such as predetermined percentages (Windsurfing Chiemsee, paragraph 52).

63. Second, the distinctive character of a sign consisting in the shape of a product, even that acquired by the use made of it, must be assessed in the light of the presumed expectations of an average consumer of the category of goods or services in question, who is reasonably well-informed and reasonably observant and circumspect (see, to that effect, the judgment in Case C-210/96 Gut Springenheide and Tusky [1998] ECR I-4657, paragraph 31).

64. Finally, the identification, by the relevant class of persons, of the product as originating from a given undertaking must be as a result of the use of the mark as a trade mark and thus as a result of the nature and effect of it, which make it capable of distinguishing the product concerned from those of other undertakings.

65. In the light of those considerations, the answer to the third question must be that, where a trader has been the only supplier of particular goods to the market, extensive use of a sign which consists of the shape of those goods may be sufficient to give the sign a distinctive character for the purposes of Article 3(3) of the Directive in circumstances where, as a result of that use, a substantial proportion of the relevant class of persons associates that shape with that trader and no other undertaking or believes that goods of that shape come from that trader. However, it is for the national court to verify that the circumstances in which the requirement under that provision is satisfied are shown to exist on the basis of specific and reliable data, that the presumed expectations of an average consumer of the category of goods or services in question, who is reasonably well-informed and reasonably observant and circumspect, are taken into account and that the identification, by the relevant class of persons, of the product as originating from a given undertaking is as a result of the use of the mark as a trade mark."

122. The Court has here made clear once again that the average consumer provides a benchmark, for it is the presumed expectations of this person that are to be taken into account by the national court in assessing distinctiveness and in determining whether a significant proportion of the relevant class of consumers identify the goods or services in issue as coming from a particular undertaking. This is an approach to which the Court has continued to adhere as may be seen from, by way of example, Case C-478/07 Budějovický Budvar, národní podnik v Rudolf Ammersin GmbH [2009] ECR I-7721 at [89].
123. It is also important to have in mind that the issue of a trade mark's distinctiveness is intimately tied to the scope of protection to which it is entitled. For example, it is well established that, in assessing an allegation of infringement under Article 5(1)(b) of the Directive (or Article 9(1)(b) of the Regulation) arising from the use of a similar sign, the court must take into account the distinctive character of the trade mark, and there will be a greater likelihood of confusion where the trade mark has a highly distinctive character either per se or as a result of the use which has been made of it. It necessarily follows that the court must therefore have regard to the impact of the accused sign on the proportion of consumers to whom the trade mark is particularly distinctive.
124. In our view the terms of recital 18 of the Unfair Commercial Practices Directive are entirely consistent with this jurisprudence. The EU legislature has explained that the practice in issue must be assessed from the perspective of an average consumer. However, in referring to "the average" member of the group at which a practice is specifically aimed, we do not believe the legislature had in mind anything so formalistic as a mathematical average and consider it was instead referring to a hypothetical person within that group who represents an appropriate normative standard. Further, in assessing whether a practice is deceptive from the perspective of such a person, we see no reason why, as a matter of principle, a national court may not properly have regard to the effect it has upon a significant number of the persons to whom it is addressed. Indeed, this was the approach adopted by the Court in considering the interpretation of Council Directive 84/450/EEC (later amended by the Unfair Commercial Practices Directive) in Case C-356/04 Lidl Belgium GmbH & Co KG v Etablissementen Franz Colruyt NV [2006] ECR I-8501 at [78] to [84], and more recently in Case C-159/09 Lidl SNC v Vierzon Distribution SA [2010] ECR I-11761 at [46]-[50]. In the Lidl Belgium case the Court said this:

"82.       The Court has thus already been led to state that advertising relating to the lower prices of cars that are parallel imports can be considered to be misleading only if it is established that the decision to buy on the part of a significant number of consumers to whom the advertising in question is addressed was made in ignorance of the fact that the lower price of the vehicles was matched by a smaller number of accessories on the cars sold by the parallel importer (X, paragraph 16).

83.       Analogously, comparative advertising relating to the general level of the prices charged by competing chains of stores in respect of their comparable ranges of products and to the amount that can be saved by consumers purchasing their basic consumables from one of those chains rather than the other should, for example, be considered to be misleading if it is established, in the light of all the relevant circumstances of the particular case, that the decision to buy on the part of a significant number of consumers to whom that advertising is addressed is made in the mistaken belief that all the advertiser's products have been taken into account in calculating the general price level, and the amount of savings, that are claimed by the advertising. The same must be true if it is established that such a decision is made in the mistaken belief that that amount will be saved by consumers irrespective of the nature and quantity of the products which they acquire from the advertiser or, for example, in the mistaken belief that all the advertiser's products without exception are cheaper than those of his competitors."

125. We see no change in the approach adopted by the Court in giving its guidance in this case (Interflora (CJEU)). In the passages of its judgment to which we have referred and upon which Mr Hobbs relies ([48] to [53]), the Court has explained that if the M & S advertisements in issue caused at least some internet users to believe, incorrectly, that M & S was a member of the Interflora commercial network then this might be a relevant consideration but would not, of itself, be a sufficient basis for a finding of liability. At the end of the day, the crucial question was whether the advertisements enabled the average consumer to tell that the flower delivery service so offered did not originate from Interflora. The judge suggested (for example at [265] of the main judgment) that confusion on the part of internet users who are ill-informed or unobservant must be discounted. Of course it must. But this formulation runs the risk of setting the bar too low and we prefer to put it differently. It is only the effect of the advertisements on internet users who are reasonably well-informed and reasonably observant that must be taken into account.
126. Considered in this way, we think it makes no difference whether the question is asked and answered from the perspective of the single hypothetical well-informed and reasonably observant internet user or whether that hypothetical person provides the benchmark or threshold for the purposes of identifying the population of internet users whose views are material. The Court has itself used the two interchangeably, as shown by the passages of its decisions in the keyword advertising cases to which we have referred. Thus the Court referred to a population of reasonably well-informed and reasonably observant internet users in Google France at [84], [87] and [90]; in BergSpechte at [39]; and in Portakabin at [52] and [68]. Indeed the decision of the Court in Portakabin is particularly illuminating because, in considering the application of Article 6 of the Directive, it explained the limited scope for the application of this defence in circumstances sufficient to satisfy Article 5(1), namely that the advertisement is likely to cause at least a significant section of the target public to establish a link between the goods or services to which it refers and the trade mark owner, and does not enable average internet users to ascertain whether the goods or services originate from the trade mark proprietor or an unconnected third party.
127. Mr Hobbs also relies heavily upon the decision of this court in Interflora (CA I) and he focuses on this passage in the judgment of Lewison LJ:

"33. I should also refer to Reed Executive Plc v Reed Business Information Ltd [2004] EWCA Civ 159; [2004] ETMR 56; [2004] RPC 40. In the course of his judgment in that case Jacob LJ (with whom Auld and Rix LJJ agreed) said (at [82]):

'Next the ordinary consumer test. The ECJ actually uses the phrase "average consumer" (e.g. Lloyd paras [25] and [26]). The notion here is conceptually different from the "substantial proportion of the public" test applied in passing off (see e.g. Neutrogena Corp v Golden Ltd [1996] RPC 473). The '"average consumer" is a notional individual whereas the substantial proportion test involves a statistical assessment, necessarily crude. But in the end I think they come to the same thing. For if a "substantial proportion" of the relevant consumers are likely be confused, so will the notional average consumer and vice versa. Whichever approach one uses, one is essentially doing the same thing—forming an overall ("global") assessment as to whether there is likely to be significant consumer confusion. It is essentially a value judgment to be drawn from all the circumstances. Further conceptional over-elaboration is apt to obscure this and is accordingly unhelpful. It may be observed that both approaches guard against too "nanny" a view of protection—to confuse only the careless or stupid is not enough.'

34. I agree entirely that the average consumer (in trade mark infringement) is conceptually different from the substantial proportion of the public test (in passing off). What I find difficult to accept is that they come to the same thing. If most consumers are not confused, how can it be said that the average consumer is? I do not think that this particular paragraph of Jacob LJ's judgment is part of the ratio of the case and, with the greatest of respect, despite Jacob LJ's vast experience of such cases I question it. In some cases the result will no doubt be the same however, the question is approached; but I do not think that it is inevitable.

35. There is, of course, no doubt that a valid survey can be an accurate diagnostic or predictive tool. They are used daily to sample public opinion on a variety of different topics. For example, they are used to gauge support for rival candidates in an election and to predict the eventual result. Suppose that a valid survey shows that in an election 49 per cent of the electorate support candidate A and 51 per cent support candidate B. It would be possible to say on the strength of such a survey that B will win the election. It would also be possible to say that a substantial proportion of the electorate will vote for candidate A. But what a survey does not, I think, tell you is: for whom will the average voter vote? In cases where acquired distinctiveness of a mark is in issue a survey may accurately identify that proportion of the relevant public which recognises the mark as a badge of trade origin. It will then be for the fact finding tribunal, with the aid of such a survey, to decide whether a significant proportion of the relevant public identify goods as originating from a particular undertaking because of the mark: see Windsurfing Chiemsee Produktions-und-Vertriebs GmbH v Boots-und-Segelzubehor (C-108/97) [1999] ECR I-2779; [1999] ETMR 585 at [52], [53].

36. In our case the question is whether M & S's advertisement would enable a reasonably well-informed and reasonably observant internet user to grasp without undue difficulty that Interflora and M & S were independent. This, as the Court of Justice has emphasised is not a question of counting heads, but is a qualitative assessment. The fact that some internet users might have had difficulty in grasping that Interflora and M & S were independent is not sufficient for a finding of infringement. If, by analogy with Neutrogena and Chocosuisse, the court were to conclude that most internet users would have grasped that, but that some would not, I cannot see that the court would be any closer to answering the legal question."

128. In these paragraphs Lewison LJ has emphasised two important points: first, that the average consumer is a hypothetical person and so conceptually different from a "substantial proportion" of the actual public; and second, that the average consumer test is not a statistical test. Neither of these points is remotely controversial. However, we do not understand Lewison LJ to have been suggesting that the average consumer is some form of mathematical average. To the contrary that consumer is, as we have said, a notional person whose presumed expectations are to be taken into account by the national court in assessing the particular question it is called upon to decide. Nor do we understand Lewison LJ to have decided that, in carrying out its assessment, the court may not have regard to the effect of the accused conduct upon a significant proportion of consumers.
129. As we have seen, the average consumer does not stand alone for it is from the perspective of this person that the court must consider the particular issue it is called upon to determine. In deciding a question of infringement of a trade mark, and determining whether a sign has affected or is liable to affect one of the functions of the mark in a claim under Article 5(1)(a) of the Directive (or Article 9(1)(a) of the Regulation), whether there is a likelihood of confusion or association under Article 5(1)(b) (or Article 9(1)(b)), or whether there is a link between the mark and the sign under Article 5(2) (or Article 9(1)(c)), the national court is required to make a qualitative assessment. It follows that it must make that assessment from the perspective of the average consumer and in accordance with the guidance given by the Court of Justice. Of course the court must ultimately give a binary answer to the question before it, that is to say, in the case of Article 5(1)(b) of the Directive, whether or not, as a result of the accused use, there exists a likelihood of confusion on the part of the public. But in light of the foregoing discussion we do not accept that a finding of infringement is precluded by a finding that many consumers, of whom the average consumer is representative, would not be confused. To the contrary, if, having regard to the perceptions and expectations of the average consumer, the court concludes that a significant proportion of the relevant public is likely to be confused such as to warrant the intervention of the court then we believe it may properly find infringement.
130. In the circumstances of this case we are, of course, concerned with a claim under Article 5(1)(a) (and Article 9(1)(a)) in the context of internet advertising and the question to be answered was whether the advertisements in issue did not enable reasonably well-informed and observant internet users, or enabled them only with difficulty, to ascertain whether the goods and services so advertised originated from Interflora or an undertaking economically linked to Interflora or, on the contrary, originated from M & S, a third party. In answering this question we consider the judge was entitled to have regard to the effect of the advertisements upon a significant section of the relevant class of consumers, and he was not barred from finding infringement by a determination that the majority of consumers were not confused.

Onus of proof

131. The judge approached the issue of the burden of proof in the main judgment in the following way. After citing paragraphs [82] to [90] of the decision of the Court of Justice in Google France, he then referred to his own decision in Datacard v Eagle Technologies Ltd [2011] EWHC 244 (Pat), [2011] RPC 17 where he said this:

"261. At first sight the Court of Justice's reasoning appears broadly consistent with the Court's treatment of this question in Arsenal, Anheuser-Busch, Céline, Case C-48/05 Adam Opel AG v Autec AG [2007] ECR I-1017 and L'Oréal v Bellure. On closer examination, however, the Court's approach appears to have shifted in two respects.

262. First, unlike in Arsenal (see the passage at [54]-[60] …) and L'Oréal v Bellure (see the passage at [61]-[62] …), the Court does not draw a distinction between use which is liable to affect the origin function on the one hand and 'purely descriptive' use on the other hand. Nor, unlike in Anheuser-Busch (see the passage at [59]-[64] …), does it draw a contrast between use which is liable to affect the origin function on the one hand and use 'for purposes other than to distinguish the goods concerned' on the other hand. Nor, unlike in Céline (at [27]), does the Court say that use of a sign is liable adversely to affect the origin function if 'consumers are liable to interpret it as designating the origin of the goods or services in question'. Thus the Court appears no longer to be saying that it is enough that the sign is used as a trade mark in relation to the goods or services in question.

263. Secondly, and perhaps more importantly, the Court introduces a new test in [84], which despite the reference to Céline is not to be found in that case. The new test is that the origin function of the trade mark is adversely affected if the use of the sign considered in context does not enable average consumers, or enables them only with difficulty, to ascertain whether the goods or services referred to under the sign originate from the proprietor of the trade mark or an undertaking economically connected to it, or from a third party. It appears from [84]-[85] and [89]-[90] that this is a test of likelihood of confusion, but with a reversed onus i.e. the onus lies upon the third party to show that the use of the sign in context is sufficiently clear that there is no possibility of confusion on the part of the average consumer as to the origin of the advertised goods or services.

264. It will be appreciated that, in stating the test in this way, I have generalised it from the specific form stated by the Court, which is only applicable to keyword advertising on the internet. It is difficult to see, however, why the test for adverse effect on the origin function should be different in that situation to the situations in issue in cases like Arsenal and L'Oréal v Bellure. Nor does the Court say that the test is different in that situation."

132. We would say straight away that we accept that the Court of Justice has, in its decision in Google France (at [82] to [90]), enunciated a new test to be applied by the national court in assessing whether the accused use has adversely affected, or is liable adversely to affect, the origin function of a trade mark at least in the context of keyword advertising cases. We have set out the key elements of that test at [75] to [77] above and it appears to be founded upon certain generic characteristics of advertisements of the kind in issue and a recognition by the Court of the need for transparency about the origin of the goods and services so advertised. However, and contrary to the view expressed by the judge, we do not recognise in this passage from the decision of the Court in Google France any conventional formulation of a likelihood of confusion test; nor do we detect any suggestion, still less a finding, that it imposes a burden of proof upon the third party. This is a matter which we develop below.
133. It was argued before the judge at trial that he was wrong in Datacard to interpret the Court as imposing a reversed onus. But the judge adhered to his view for seven reasons he proceeded to set out:

"234. … First, in DataCard I was not addressing the legal burden of proving infringement under Article 5(1)(a) of the Directive and Article 9(1)(a) of the Regulation, which I agree lies on the trade mark proprietor. Rather, I was considering who bore the evidential onus in relation to this particular issue.

235. Secondly, while I accept that, in general, neither the Directive nor the Regulation harmonise national procedural rules, as counsel for Interflora pointed out, we are concerned here with infringement claims under Article 5(1)(a) of the Directive and Article 9(1)(a) of the Regulation, and thus strictly speaking the passage from the tenth recital does not apply. Furthermore, the CJEU has allocated the burden of proof with regard to particular issues in other Article 5(1)(a)/Article 9(1)(a) cases. Thus in Joined Cases C-414/99 to C-416/99 Zino Davidoff SA v A & G Imports Ltd [2001] ECR I-8691, a case about parallel imports, the Court of Justice held at [54] that "it is for the trader alleging consent to prove it and not for the trade mark proprietor to demonstrate its absence". In Case C-244/00 Van Doren + Q GmbH v Lifestyle sports + sportswear Handelgesellschaft mbH [2003] ECR I-3051, another case involving parallel imports, the Court of Justice "split" the burden, holding that where the defendant established a risk of market partition, the onus of proving that the goods were first placed on the market outside the EEA fell on the trade mark proprietor: see [37]-[42].

236. Thirdly, I find it difficult to see how else to interpret what the Court of Justice said in Google France at [84]. The Court's phraseology involves a negative condition: "The function of indicating the origin of the trade mark is adversely affected if the ad does not enable … users, or enables them only with difficulty to ascertain…". That is to say, the origin function is adversely affected unless the advertisement does enable users without difficulty to ascertain this. This implies that it is for the advertiser to show that its advertisement does enable users to ascertain this without difficulty.

237. Fourthly, I consider that this reading is supported by what the Court of Justice says in the following paragraphs. At [85] the Court says that "the use by the third party of the sign identical with the mark as a keyword triggering the display of that ad is liable to create the impression that there is a material link in the course of trade …". Thus the Court is saying that the fact that the advertisement is triggered by the entry by the users of a keyword which is identical to the trade mark is likely to make the user think that the advertisement is for the trade mark proprietor's goods or services. It follows that it is for the advertiser to ensure that that impression is dispelled by the advertisement. At [86]-[87] the Court refers to the requirement of transparency in commercial communications forming part of or constituting an information society service laid down by Article 6 of European Parliament and Council Directive 2000/31/EC of 8 June 2000 on certain legal aspects of information society services, in particular electronic commerce, in the Internal Market ("the E-Commerce Directive"). Again, this indicates that it is for the advertiser to ensure that the advertisement is transparent as to the source of the goods or services advertised. At [90] the Court says that if the ad is "vague to such an extent on the origin of the goods or services … that … users are unable to determine … whether the advertiser is a third party … or … economically linked to the proprietor, the conclusion must also be that there is an adverse effect". Again, this indicates that it is for the advertiser to ensure that the advertisement is clear as to the origin of the goods or services advertised.

238. Fifthly, if the onus is not on the advertiser, the consequence would appear to be to impose on the trade mark proprietor a requirement to prove something akin to a likelihood of confusion. That cannot be correct for at least two reasons. First, it is plain on the face of Article 5(1)(a) and Article 9(1)(a) that there is no requirement for a likelihood of confusion, in contradistinction to Article 5(1)(b) and Article 9(1)(b). Secondly, the Court of Justice expressly acknowledges at [77] that "the protection afforded by Article 5(1)(a) … and Article 9(1)(a) … is … more extensive than that provided for in the respective paragraphs (1(b) of those articles, the application of which requires that there be a likelihood of confusion".

239. Sixthly, my interpretation of what the Court of Justice has said is consistent with Article 16(1) of the TRIPS, which requires there to be at least a rebuttable presumption of a likelihood of confusion in double identity cases, whereas an interpretation which required the trade mark proprietor to prove a likelihood of confusion, or something akin to it, would not be.

239. Seventhly, I consider that my interpretation is supported by the subsequent case law, as discussed below."

134. The judge did, however, qualify what he had said in Datacard in one minor respect:

"… I accept that, when I referred to "no possibility of confusion", I put the matter too strongly. It would, I think, be more accurate to say that the onus lies on the third party to show that the use of the sign in context is sufficiently clear that there is no real risk of confusion on the part of the average consumer as to the origin of the advertised goods or services. "

135. We have carefully considered all of the reasons given by the judge in support of his conclusion but we do not find them persuasive, and that is so whether they are considered individually or collectively. In our view the starting point is the express recognition by the Court that a trade mark proprietor cannot oppose the use of a sign unless all of the conditions set out in Article 5 of the Directive (or Article 9 of the Regulation) are satisfied. Specifically, and notwithstanding Article 16(1) of TRIPS, the Court has now held in a long line of decisions that the right under Article 5(1)(a) (and Article 9(1)(a)), though "absolute", is conferred to enable the trade mark proprietor to protect his interests as proprietor, that is to ensure the trade mark can fulfil its functions. The exercise of the right must therefore be reserved to cases in which a third party's use of the sign adversely affects, or is liable adversely to affect, one of those functions.
136. Second, it seems to us that, as matter of principle, the burden of establishing the necessary elements of a claim for trade mark infringement must normally lie upon the party making the allegation, and that is so as a matter of EU or English law. We are confirmed in this view by the decision of the Court of Justice in Case C-405/03 Class International BV v Colgate-Palmolive Company [2005] ECR I-8735 which concerned a claim for infringement of a registered trade mark arising from the introduction into the EU, under the external transit or customs warehousing procedures, of original goods bearing the mark. The Court was asked a series of questions concerning the proper interpretation of Article 5 of the Directive (and Article 9 of the Regulation), including a question concerning the burden of proof and, in particular, which party had the burden, in a situation such as that in issue, of proving the facts which would give rise to a claim for exercising the prohibition provided for in Article 5(3)(b) and (c) (Article 9(2)(b) and (c)). The Court answered that question in these terms:

"70 In that regard, the issue of proof is raised when a dispute arises, that is to say, when the trade mark proprietor pleads interference with the exclusive rights conferred on it by Article 5(1) of the Directive and Article 9(1) of the Regulation.

71 The interference which may be pleaded consists either in the release for free circulation of the goods or an offering or sale of those goods which necessarily entails putting them on the market in the Community.

72 Interference is the condition for the exercise of the right of prohibition provided for in Article 5(3)(b) and (c) of the Directive and Article 9(2)(b) and (c) of the Regulation.

73 In respect of the issue of the onus of proving that interference, it must be pointed out, first, that if it were a matter for the national laws of the Member States, the consequence for trade mark proprietors could be that protection would vary according to the legal system concerned. The objective of 'the same protection under the legal systems of all the Member States' set out in the ninth recital in the preamble to the Directive, where it is described as fundamental, would not be attained (see, on the subject of the Directive, Zino Davidoff and Levi Strauss, cited above, paragraphs 41 and 42).

74 It must then be stated that, in a situation such as the one in the main proceedings, the onus of proving interference must lie with the trade mark proprietor who alleges it. If that is proven, it is then for the trader sued to prove the existence of the consent of the proprietor to the marketing of the goods in the Community (see, on the subject of the Directive, Zino Davidoff and LeviStrauss, cited above, paragraph 54)."

137. Accordingly, the general position under EU law is that the burden of proving an allegation of infringement lies on the person making the allegation. The position under English law is, of course, exactly the same.
138. Third, it is not the purpose of trade mark law to protect the proprietor of a trade mark against fair competition. Further, as we have seen, the aim of internet advertising using keywords corresponding to trade marks is, in general, to offer to internet users alternatives to the goods or services of those trade mark proprietors, and so is not inherently objectionable.
139. Fourth, it follows from the foregoing that the burden of proving all of the necessary six conditions of a successful claim for infringement under Article 5(1)(a) of the Directive (or Article 9(1)(a) of the Regulation) would ordinarily be expected to lie upon the trade mark proprietor. None of those conditions is an exception to protection such as to justify the reversal of the burden of proof seen in the cases to which the judge referred at [235] of the main judgment.
140. Fifth, the judge appeared to recognise the force of these points for, at [234], he sought to clarify his reasoning in Datacard by explaining he was not there addressing the legal burden, which he agreed lay upon the trade mark proprietor. He maintained he was instead considering the evidential burden. We are not impressed by this purported distinction for, as Patten LJ observed in argument, the effect of the judge's decision, effective as it is from the outset and creating as it does a presumption, imposes on the third party advertiser what is tantamount to the legal burden.
141. We must therefore return to the reasoning of the Court in Google France to ascertain whether, notwithstanding the foregoing, the Court has held that the burden, whether characterised as legal or evidential, lies upon the advertiser to prove that the accused use does not infringe, for it is this reasoning which lies at the heart of the judge's third and fourth reasons, which are, in substance, a development of his earlier reasoning in Datacard.
142. The key question in any such case is whether the advertisement does not enable normally informed and reasonably attentive internet users, or enables them only with difficulty, to ascertain whether the goods or services referred to in the advertisement originate from the trade mark proprietor. We believe this question is cast as it is because the essential function of a trade mark is to guarantee the origin of the marked goods or services to the consumer by enabling him to distinguish those goods or services from others which have another origin. If a third party uses a sign identical to the trade mark in relation to goods or services identical to those for which the mark is registered and in such a way that the average consumer is not so enabled then the origin function of the mark is liable to be affected adversely.
143. As we have said, the Court has recognised that keyword advertising is not inherently objectionable. But it has also acknowledged that there are features of such advertising which may lead the consumer to mistake the origin of the goods or services so advertised. It has noted too the importance of transparency in the display of advertisements on the internet. In this context the Court has therefore explained that the trade mark proprietor must be entitled to prohibit third party advertisements displayed as a result of the use of the mark as a keyword if average internet users may wrongly perceive that the goods or services so advertised originate from the trade mark proprietor, or which are so vague that such users cannot determine whether they do or not. This of course means that the third party advertiser using a trade mark as a keyword must take care to ensure that his advertisements do enable average internet users to ascertain whether the goods or services originate from the trade mark proprietor or an unconnected third party. If he fails to do so, he may be found to have infringed the trade mark.
144. In our judgment, however, none of the foregoing means that, in proceedings for infringement under Article 5(1)(a) of the Directive (or Article 9(1)(a) of the Regulation), the third party advertiser bears the burden of proving non infringement. We believe the error in the judge's reasoning is that he wrongly elided the burden of proof in legal proceedings with the need in practice for third party advertisers to craft their advertisements with appropriate clarity. The one does not follow from the other. There is no inconsistency between, on the one hand, requiring advertisers to ensure their advertisements enable the average consumer to ascertain that the advertised goods or services originate not from the trade mark proprietor but from an unconnected third party and, on the other hand, saying that the burden of proving that they have failed to do so lies on the trade mark proprietor.
145. The judge also found support for his interpretation in the subsequent decisions of the Court. First, he referred to the decision in BergSpechte, the material parts of which we have set out above (at [81] to [83]). He then cited again from his own decision in Datacard :

"266. At first blush, this passage appears to indicate that precisely the same test should be applied to determine both whether the origin function of the trade mark is liable to be affected under Article 5(1)(a) and whether there is a likelihood of confusion under Article 5(1)(b). I cannot believe that that is what the Court of Justice meant to say, however. Article 5(1)(a) contains no requirement for a likelihood of confusion, whereas Article 5(1)(b) does. The reason why Article 5(1)(a) contains no requirement for a likelihood of confusion is that a likelihood of confusion is presumed where a sign identical to the trade mark is used in relation to goods or services identical to those for which the trade mark is registered, as is required by Article 16(1) of the Agreement on Trade-related Aspects of Intellectual Property Rights. That is what recital (11) of the Directive means when it says that 'the protection afforded by the registered trade mark … should be absolute in the case of identity between the mark and the sign and the goods or services'. (In another sense, the protection is not absolute, of course, because it is subject to the defences under Articles 6 and 7 of the Directive.) Furthermore, in the case of Article 5(1)(b) it is clear from the Court's earlier jurisprudence that the onus lies upon the trade mark proprietor to establish the presence of a likelihood of confusion, not upon the third party to establish the absence of a likelihood of confusion.

267. Accordingly, it seems to me that the correct reading of [39] must be that the words 'it will be for the national court to hold whether there is a likelihood of confusion when' mean that the trade mark proprietor must establish a positive likelihood of confusion in order to succeed under Article 5(1)(b), and not merely the presence of circumstances which would suffice for the purposes of establishing an effect on the origin function under Article 5(1)(a).

268. Even if that is right, BergSpechte appears to confirm that the test which must be applied in order to determine whether the use of the sign complained of is liable to affect the origin function of the trade mark is a reverse likelihood of confusion test as discussed in paragraph 263 above. Thus the combined effect of the judgments in Google France and BergSpechte is to narrow the distinction between Article 5(1)(a) and Article 5(1)(b) even if they do not eliminate it."

146. This passage was also the subject of submissions before him in this case, which he assessed in this way:

"247. Counsel for M & S described this aspect of the Court of Justice's case law in his closing submissions successively as follows: "a conundrum", "unfathomable" and "it makes no sense". He submitted, however, that it was nevertheless this court's duty to apply the "plain unvarnished words" of the Court's judgments without struggling to understand the jurisprudential basis for them. I do not accept that submission. In my view one cannot hope to be able properly to apply the Court's judgments without trying to understand what they mean. I adhere to the opinion which I expressed in DataCard that the Court cannot have meant to say that the same test should be applied under both Article 5(1)(a)/Article 9(1)(a) and Article 5(1)(b)/Article 9(1)(b). I also adhere to the opinion that the interpretation which I offered in DataCard (as qualified above) is the best way in which to make sense of the Court's case law. "

147. It can be seen that the judge was deeply troubled by the notion that, at least in this context, the same test should be applied under Article 5(1)(a) (and Article 9(1)(a)) and under Article 5(1)(b) (Article 9(1)(b)) and considered that, as he put it, the best way to make sense of the case law was to interpret the decision in Google France as laying down a likelihood of confusion test but with a reversed onus.
148. Again we do not find this reasoning persuasive. In our judgment it is important always to have in mind that, despite the "absolute" nature of the protection afforded by Article 5(1)(a) (and Article 9(1)(a)) against the use of signs identical to the trade mark in relation to identical goods or services, the Court has consistently held that it is limited to those cases in which the use of the sign by a third party adversely affects or is liable adversely to affect one of the functions of the trade mark. Those functions include but are not limited to the essential function of indicating origin. By contrast, Article 5(1)(b) (and Article 9(1)(b)) has a wider ambit in that it extends to the use of similar signs in relation to similar goods or services. But it only affords protection if there is a likelihood of confusion.
149. In the present context and perhaps reflecting the inherently unobjectionable nature of fair competition, the Court has enunciated the particular test which must be applied in assessing whether the accused use affects or is liable to affect the origin function in a claim under Article 5(1)(a) (and Article 9(1)(a)). It has also explained the test which must be applied in assessing the likelihood of confusion in a claim under Article 5(1)(b) (and Article 9(1)(b)). It appears from its answer to the referred question in BergSpechte that, in this context, the essential test is the same. But in light of the foregoing we do not find this particularly surprising. What is striking, however, is that the Court has chosen not to say that the test under Article 5(1)(a) (and Article 9(1)(a)) is the same as that under Article 5(1)(b) (and Article 9(1)(b)) but with a reversed onus, and we consider that if this is what it intended then it would indeed have said so.
150. We do not believe the other decisions of the Court on keyword advertising materially affect this conclusion. In Portakabin (at [70]) and in Interflora (CJEU) (at [51] to [53]) the Court emphasised that it is the advertiser who chooses the keyword and decides how it will be presented. This is certainly the case but it does not mean to say that, in infringement proceedings, it also has the burden of proving it has ensured that its advertisements make it clear that the advertised goods and services do not originate from the trade mark proprietor or adversely affect any other function of the trade mark. Further, it was not suggested to us that the judge's approach to the issue of burden of proof derives any support from any of the decisions of the courts of other Member States of the EU concerning keyword advertising. In that regard we have given particular consideration to the decision of the German Federal Supreme Court in case I ZR 53/12 Fleurop. Here the Supreme Court explained that it was satisfied that the Court of Appeal had made no error in finding that infringement had been established on the particular facts of that case.
151. In our judgment this criticism of the approach adopted by the judge is therefore a sound one. The judge wrongly held (at [241]) that the onus lies on the third party advertiser to show that the use of the sign in context is sufficiently clear that there is no real risk of confusion on the part of the average consumer as to the origin of the advertised goods or services. In our judgment the onus lies on the trade mark proprietor to establish that the advertisement complained of does not enable normally informed and reasonably attentive internet users, or enables them only with difficulty, to ascertain whether the goods or services referred to by the advertisement originate from the trade mark proprietor or an undertaking economically connected to it or, on the contrary, originate from a third party.

Initial interest confusion

152. The judge introduced the concept of initial interest confusion into his analysis at [306] of the main judgment in considering the evidence of Mr Pandya and the Hitwise report. There the judge said this:

"… I understood [Mr Silverleaf] to be relying on my decision in Och-Ziff Management Europe Ltd v OCH Capital Ltd [2010] EWHC 2599 (Ch), [2011] ETMR 1, although he did not cite it. In that case I held at [79]-[101] that initial interest confusion could be relied upon to establish a likelihood of confusion under Article 5(1)(b)/Article 9(1)(b). The implicit submission was that initial interest confusion is equally relevant to show an adverse effect on the origin function of the trade mark for the purposes of Article 5(1)(a)/Article 9(1)(a). I accept both of these submissions."

153. We return to the issue of the Hitwise report and the evidence of Mr Pandya later in this judgment, but for the moment focus on the doctrine of initial interest confusion and the role it has to play in any analysis under Article 5(1) of the Directive (or Article 9(1)(a) of the Regulation). Mr Hobbs contends that here the judge again fell into error because initial interest confusion has no role to play at any stage of any such inquiry. Mr Silverleaf agrees with him, at least for the purposes of this appeal. He has also made it clear that he does not rely upon this aspect of the judge's reasoning. But, he says, he does not need it and it made no difference to the outcome of the trial and it should similarly make no difference to the outcome of this appeal.
154. As the passage of the main judgment we have cited immediately above makes clear, the judge considered the doctrine of initial interest confusion in some detail in his judgment in Och-Ziff Management Europe Ltd v OCH Capital Ltd [2010] EWHC 2599 (Ch), [2011] ETMR 1. As he there explained, initial interest confusion is an expression which derives from US trade mark law and has been used to encompass a range of situations, and the doctrine it identifies is highly controversial. For the purposes of that decision, he defined it (at [87]) as confusion on the part of the public as to the trade origin of the goods or services in relation to which the impugned sign has been used arising from use of the sign prior to purchase of those goods or services, and in particular confusion arising from use of the sign in promotional or advertising materials. He went on to hold (at [101]) that, as so defined, it was actionable under Article 9(1)(b) of the Regulation.
155. In our judgment it is not helpful to seek to import the doctrine of initial interest confusion into EU trade mark law, at least so far as it applies to the use of a sign the same as or similar to a trade mark as a keyword in an internet referencing service, and it has the potential positively to mislead. We say that for the following reasons. The Court of Justice has already enunciated in clear and unambiguous terms the test that must be applied under the Directive and the Regulation in determining whether the accused sign has an adverse effect on the origin function of the trade mark. As we have seen, it did so first in Google France in addressing Article 5(1)(a) of the Directive (and Article 9(1)(a) of the Regulation) and then again in BergSpechte and Portakabin in addressing both Article 5(1)(a) and (b) (Article 9(1)(a) and (b)). In either case it must be shown that the advertisement does not enable an average internet user, or enables that user only with difficulty, to ascertain whether the goods or services referred to therein originate from the trade mark proprietor or an undertaking economically connected to it or, on the contrary, originate from a third party. Of course, Article 5(1)(a) and Article 9(1)(a) also afford protection against use in this way of a sign identical to the trade mark if that use is liable adversely to affect one of the other functions of a trade mark, as the Court elaborated in its decision in Interflora (CJEU).
156. These tests have been formulated by the Court with great care and reflect the importance of trade marks in developing a system of undistorted competition whilst recognising that their purpose is not to protect their proprietors against fair competition. Moreover, the Court has acknowledged that internet advertising on the basis of keywords corresponding to trade marks is not inherently objectionable because its purpose is, in general, to offer to internet users alternatives to the goods or services of the trade mark proprietors. The tests enunciated by the Court therefore incorporate appropriate checks and balances. In particular, the national court is required to consider the matter from the perspective of the average consumer, a concept we have discussed, and to decide whether the advertiser has enabled that average consumer to ascertain the origin of the advertised goods or services and so make an informed decision. We would emphasise it is not the duty of such advertisers to avoid confusion.
157. Returning now to the doctrine of initial interest confusion, it is, as the judge recognised, highly controversial and, as he also recognised, it has been applied to a range of situations in which a sign which is the same as or similar to a registered trade mark is used by a third party in advertisements for goods or services for which it is registered. They extend from, at one end of the spectrum, initial attraction of consumers based upon some kind of association of advertised goods or services with those of the trade mark proprietor or even mere diversion, to, at the other end of the spectrum, initial confusion which gives rise to a real risk that consumers will actually buy the advertised goods or services even though they know they are not the goods or services of the trade mark proprietor. Second, and significantly, it does not incorporate the checks and balances to which we have referred.
158. In our view the doctrine of initial interest confusion is therefore an unnecessary and potentially misleading gloss on the tests the Court has articulated and we think it should perform no part of the analysis of our national courts in claims of the kind before us. We consider the judge was therefore wrong to approach the matter as he did.

Procedural irregularities and errors on the evidence

159. We begin with the judgment on the admissibility of the evidence contained in the academic papers ([2013] EWHC 936 (Ch)), and the reliance the judge placed upon those papers and the evidence of Mr Rose. The purpose for which Interflora sought to rely upon this evidence was, in summary, to support their argument that reasonably well-informed and observant internet users were not all sophisticated users of search engines who were alive to precisely what Google were doing with regard to paid advertising. They covered a range of subject matters, but those considered to be of particular relevance were based upon studies of consumer behaviour.
160. We have summarised the nature of M & S's objections to the academic papers and the reasons for the judge's rejection of those objections earlier in this judgment (at [30] to [31]). Mr Hobbs originally sought to argue before us both that the judge was wrong to rule that they did not fall within the scope of CPR Part 35, and further and in any event, that the judge was wrong to allow Interflora to rely upon them in the exercise of his discretion, and that he should have excluded them.
161. In the event Mr Hobbs has not pursued the first of these arguments in light of the recent decision of this court in Hoyle v Rogers & Anor [2014] EWCA Civ 257. Christopher Clarke LJ (with whom Arden and Treacy LJJ agreed) explained from [61] to [67] that CPR Part 35 is not a comprehensive and exclusive code regulating the admission of expert evidence but rather is only concerned with a person who has been instructed to give or prepare expert evidence for the purposes of the proceedings, and that its purpose is to regulate any such person's expert evidence. Accordingly Mr Hobbs now accepts that the academic papers do not fall in the scope of CPR Part 35 and that the judge was right so to hold.
162. Mr Hobbs has pursued his second argument, however. He maintains that the judge failed adequately to deal with his complaint that the Civil Evidence Act notice had been served so late that his clients had insufficient time to deal with the evidence and that the judge ought therefore to have excluded it. He also contends that he was placed at a serious disadvantage because he had no proper notice of precisely how the papers were going to be deployed or what points were going to be made in relation to them; because there was no independent witness who could assist the court as to their value or significance; and because he had no witness he could cross-examine about them. Much the same concerns were expressed to the judge, and he dealt with them in this way:

"29. I am bound to say that I have sympathy with Marks and Spencer's complaint that it has been given notice of this material at a very late stage of these proceedings. As I observed at the beginning of this judgment, however, Marks and Spencer does not contend that the Civil Evidence Act Notice was served out of time. The lateness of service of the Civil Evidence Act Notice, as I understand it, is due to the lateness with which the parties ultimately agreed to exchange expert evidence. That being so, the mere fact that the material comes into the case at a rather late stage does not seem to me, in and of itself, to be a sufficient reason to regard Interflora's reliance upon it as unfair or prejudicial to Marks and Spencer. That is particularly so when it is borne in mind that I made the original suggestion that this kind of material be considered as long ago as 24 February 2012 and that Interflora made it clear that they were going down this line in their list of documents served on 16 August 2012.

30. The point which most concerns me is whether it is fair for Interflora to be permitted to rely upon a selection of the academic literature in circumstances where I do not know what a wider search of the academic literature might throw up and whether or not there is academic literature, or indeed other official reports like the Ofcom reports, which could be relied upon by Marks and Spencer to rebut the proposition advanced by Interflora.

31. At the end of the day, it seems to me that that is not a matter which amounts to sufficient unfairness that I should exclude this evidence. Marks and Spencer has been on notice that this court would be interested in receiving independent and impartial evidence available from public sources simply at the cost of copying it for well over a year. If Marks and Spencer has not undertaken the kind of search of official reports and the academic literature that Interflora have undertaken, then it seems to me it has only itself to blame.

32. Finally, I think I should deal with one further point made by counsel for Marks and Spencer. He submitted that a further consideration to be taken into account in these circumstances was that the court should be astute to an attempt by parties like Interflora to turn the court itself into its own expert. I think that is a very real concern. I confess to discomfort at the proposition that scientific literature can be put before courts without the benefit of an expert's report to put the literature into context and without the opportunity for an expert to be cross-examined upon the contents of such literature.

33. Upon analysis, however, I think my discomfort relates to the ability of the court to understand and evaluate the material in question. Whatever might be the position in other cases, I do not consider that that is a difficulty which applies in the circumstances of the present case. I have looked at much of the material that is relied upon by Interflora during the course of counsel for Interflora's opening submissions in this case, when he took me to many of the documents in question. I have seen nothing which leads me to believe that I have any difficulty in understanding and evaluating the material in question. Accordingly, whatever might be the position in other cases, this factor does not cause me to exclude the evidence in this case."

163. We see considerable force in the points which Mr Hobbs has advanced, and believe that a judge should exercise considerable caution before seeking to interpret and evaluate academic papers of the kind in issue here, some of which are of a generally technical or scientific nature, without the benefit of expert evidence. But that, it seems to us, is how the judge approached the issue before him in this case. He was conscious of the risk of unfairness arising from the service of the notice only shortly before trial (although not, we should say, out of time) and he was also alert to the danger of, as he put it, the court turning itself into its own expert. He considered the nature and contents of the papers and took the view that he could understand them, that they were relevant to the issues he had to decide and that, in the interests of justice, they should be admitted. That was, in our view, a reasonable exercise of the discretion conferred upon him and it is not one with which we can properly interfere.
164. In his main judgment, the judge focused on two of these papers. The first was a paper by Benjamin Edelman and Duncan Gilchrist, both of the Harvard Business School, entitled "Advertising Disclosure: Measuring Labeling Alternatives in Internet Search Engines", Information Economics and Policy 24, 75-89 (2012). This was based on consumer research carried out in the USA using a demographically varied sample of 723 participants recruited using an online survey service. As the judge explained at [149] to [151], this was one of the best papers from Interflora's perspective, and the authors concluded from their study that the label 'Paid Advertisement' was more effective than the label 'Sponsored link' at conveying to users that a given link was an advertisement. We accept that may have been the case but we question how much assistance it gave the judge in ascertaining the perceptions of the average consumer in the UK when faced with one of the SERPs the subject of these proceedings, particularly bearing in mind there was no expert witness before the court who could speak to it.
165. The second was a report of an eye tracking study commissioned by Google and published in May 2011. As the judge described at [152], the study was carried out in Germany, and its purpose was to compare the effects of keyword advertising on "mobile websites", such as websites displayed on smartphones, with the effects of such advertising on "regular websites", such as websites displayed on desktop computers. The results included "heat maps" showing how strongly focused participants' eyes were on different parts of the SERP. The study showed that users focused strongly on the top of the SERP which was where the advertisements were displayed. In the judge's view, expressed at [156], it also showed a striking lack of awareness of the advertisements. This opinion was, it seems, formed on the basis of the response by users to the question "which brands and products have you seen advertising for?". As summarised by the judge, just 3% of desktop users and 12% of smartphone users were able to identify them. Even when users were asked "have you seen advertising for the brands listed below", only 24% of desktop users and 44% of smartphone users were able to identify them.
166. We recognise the potential relevance of this analysis but think it must again be approached with care for it seems to us that an inability to remember the names of competitive brand offerings may be very different from an inability to distinguish natural search results from sponsored advertisements when looking at a SERP. Moreover, there was considerable evidence before the court that competitive brand bidding was very common. It was, perhaps, with some of these considerations in mind that Mr Silverleaf, in closing, indicated to the judge (transcript, day 5, page 860 to 861) that Interflora were not putting this material at the forefront of their case.
167. That brings us to the evidence of Mr Rose which, together with that of Mr Bond, may, in light of the foregoing, have carried some considerable weight in the judge's mind. We have explained (at [32] above) that Mr Rose was an employee of the online digital marketing company which managed the online paid search advertising strategies and campaigns for Interflora in the UK. He was not an expert. In his third witness statement he said:

"PPC [pay per click] ads appear for different reasons to the natural search results. I will explain this in more detail below. However the links often appear on the screen in very similar ways. As a result, in my experience, not only do most people not know or appreciate the difference between the search results which have been generated by PPC and which have been generated by SEO [search engine optimisation] but many people do not know or understand that there is a difference."

168. As we have also explained, on the second day of the trial, the judge ruled that he would attach no weight to Mr Rose's extrapolation and, as a result, he was not cross-examined. Yet, at [159] to [160] of the main judgment, the judge said this:

"[159] … Mr Rose expressed the view in his evidence that "in my experience, not only do most people not know or appreciate the difference between the search results which have been generated by PPC and which have generated by SEO, but many people do not know or understand that there is a difference". Again, Mr Rose cannot know this, but it represents an informed assessment from someone whose job it is to have an informed assessment on this subject.

[160] Still further, Mr Bond's and Mr Rose's assessments are consistent with (a) the Ofcom reports concerning internet literacy, (b) the academic literature relied upon by Interflora (in particular, the Edelman and Gilchrist article) and (c) the Google eye tracking study."

169. Mr Hobbs submits, and we accept, that this was not permissible in light of the judge's previous ruling. The judge should not have attached any weight to the opinion and assessment of Mr Rose. In this respect too, the judge fell into error.
170. We now come to the Hitwise report and the evidence of Mr Pandya. This report was produced and the evidence came to be given in the circumstances we have outlined at [33] to [34] above. The judge dealt with the report, the evidence of Mr Pandya and the parties' submissions in his main judgment in these terms (incorporating the reference to Och-Ziff we have already discussed):

"301. Secondly, Interflora rely upon the analysis of Hitwise's data conducted by Mr Pandya which I have mentioned above. Mr Pandya identified a subgroup of consumers within its sample (referred to as "Segment A") who searched for Interflora-related terms on Google UK, clicked on a paid advertisement and then visited the M & S website provided that they did so within the same session (the end of a session being marked by 30 minutes of inactivity). The analysis was carried out on a quarterly basis, and the size of Segment A varied between about 4,400 and about 8,800 consumers.

302. Mr Pandya then compared the proportion of those in Segment A who visited Interflora after visiting M & S with the proportion of all visitors to the M & S website in its sample who did so. He was able to report statistically significant results for six quarters, namely Q1 2009, Q2 2009, Q3 2009, Q4 2009, Q1 2010 and Q1 2011. He found that people in Segment A were between 44 (Q1 2011) and 106 (Q2 2009) times more likely to visit Interflora after M & S than the average visitor to M & S. Thus in Q2 2009, 16% of Segment A did so, whereas on 0.15% of all visitors did so. Mr Pandya also compared the proportion of Segment A who visited Interflora after visiting M & S with the proportion of all visitors to the flowers section of M & S website in Hitwise's sample who did so. In this case he was able to report statistically significant results for four quarters, namely Q3 2009, Q4 2009, Q1 2010 and Q1 2011. He found that people in Segment A were between 7.3 (Q1 2011) and 10.9 (Q3 2009) times more likely to visit Interflora after M & S than the average visitor to the M & S flowers section. In my view it is the latter figures which are more relevant for present purposes. In addition, Mr Pandya compared the proportion of Segment A who visited Interflora after visiting M & S with the proportion of all visitors to 20 other flowers websites in the sample who did so. In this case he was able to report statistically significant results for three quarters, namely Q4 2009, Q1 2010 and Q1 2011. He found that people in Segment A were between 1.4 (Q1 2011) and 2.8 (Q1 2010) times more likely to visit Interflora after M & S than the average visitor to the other flowers websites.

303. A number of criticisms were put to Mr Pandya in cross-examination, but in my judgment none of them undermined his analysis. The main point which counsel for M & S relied on in his closing submissions was that Mr Pandya accepted that he was unable to determine what consumers' thought processes were, all he could report was their observed behaviour. As counsel for Interflora submitted, however, the fact that Mr Pandya was properly cautious about interpreting the data does not prevent the court from doing so.

304. Mr Pandya's analysis shows that consumers who searched for Interflora, clicked on an advertisement and then visited the M & S website were an order of magnitude more likely to visit Interflora's website after visiting the M & S website than the average visitor to the flowers section of the M & S website. Interflora's interpretation of this consists of two propositions. The first is that a significant number of consumers in Segment A decided after they had clicked through to the M & S website that it was not where they wanted to be and went to the Interflora website instead. The second is that the reason for this change of mind was that those consumers had clicked through from the M & S advertisement because they assumed from the appearance of the advertisement in response to their search that M & S was part of the Interflora network, but they realised that that was not the case when they clicked through to the M & S website and saw no reference to Interflora.

305. I do not think that M & S seriously challenges Interflora's first proposition, but M & S vigorously disputes that the second proposition follows from the first. As counsel for Interflora pointed out, however, M & S has not risen to the challenge of providing an alternative explanation of the data. In my judgment Interflora's interpretation of the data is persuasive and I accept it.

306. The second proposition involves an acceptance by Interflora that consumers who were initially confused by the M & S advertisement ceased to be confused after they had clicked through to the M & S website. Counsel for Interflora submitted that this did not detract from Interflora's case for two reasons. First, if there was initial confusion, it was probable that some consumers would remain confused after clicking through. Secondly, and in any event, such initial interest confusion was sufficient to establish Interflora's case under Article 5(1)(a)/Article 9(1)(a). So far as the latter point is concerned, I understood him to be relying on my decision in Och-Ziff Management Europe Ltd v OCH Capital Ltd [2010] EWHC 2599 (Ch), [2011] ETMR 1, although he did not cite it. In that case I held at [79]-[101] that initial interest confusion could be relied upon to establish a likelihood of confusion under Article 5(1)(b)/Article 9(1)(b). The implicit submission was that initial interest confusion is equally relevant to show an adverse effect on the origin function of the trade mark for the purposes of Article 5(1)(a)/Article 9(1)(a). I accept both of these submissions."

171. The judge therefore found persuasive the fact that consumers who searched for Interflora, clicked on an advertisement and then visited the M & S website were an order of magnitude more likely to visit Interflora's website after visiting the M & S website than the average visitor to the flowers section of the M & S website. He concluded that this was evidence of confusion, or at least initial interest confusion.
172. Mr Pandya explained in the course of his cross-examination that the data did not reveal how long the persons in segment A had spent looking at the SERP; whether those persons then visited the Interflora website before visiting the M & S website; or how long those persons spent on the M & S website before going to the Interflora website. Indeed, he considered the limitations of the data meant that it was simply not possible to draw any conclusions about the state of mind of the persons in segment A, as he elaborated in these terms (on day 4 from page 681 to 682):

"Q. Really, am I not right in saying to you that looking at a sequence of events, without any timing data, but looking at a sequence of events, either a pathway or a sub-pathway is basically the same thing as looking at browsing on the net; all you are doing is looking at people browsing on the net?

A. Yes, so our data is based on people's browsing behaviour on the net and then segmenting that based on the criteria that we are after.

Q. Knowing the sequence in which they browse is not of itself telling you anything except that they browsed in the way and in the sequence they did, is it?

A. Yes, so this is the actual behaviour that we saw from the data but it is difficult to [then] comment on the intent or any of that sort of thing, yes.

Q. You say it is difficult and thank you for that, but is it in fact impossible to come to any conclusions on intent from this data that you have collected?

A. Yes, so we cannot give the intent behind a particular behaviour. All we can report is a particular behaviour, but we are not in a position to give intent behind that behaviour.

Q. You certainly cannot look into a consumer's mind to understand what their thought processes from this type of data?

A. Precisely, that is correct."

173. A little later (at page 683), he emphasised once again "we cannot answer the confusion question." This seems to us to be entirely understandable. The data illustrate that persons who have first searched for "interflora" are much more likely to visit the Interflora website after the M & S flowers website than those who arrived at the M & S flowers website without having first searched for "interflora". That is, one might suppose, entirely consistent with the proposition that those persons who arrived at the M & S flowers website after searching for "interflora" were much more interested in Interflora than those persons who arrived at the M & S flowers website without first carrying out such a search. We put it to Mr Silverleaf in the course of oral submissions that this was a perfectly rational explanation for the data and he was unable to provide any persuasive reason why it was not. In the circumstances we do not understand how the judge drew the conclusion that he did and, in our judgment, having rejected the evidence of Mr Pandya, it was incumbent upon him to explain his reasoning for doing so. That he did not do. We are therefore satisfied the judge erred in giving this evidence the weight that he did.
174. We have carefully considered the other complaints that Mr Hobbs makes about the judge's approach to the evidence but these were not developed in oral argument and we do not believe they have any independent force or add to the points we have addressed, although, as we reiterate below, we recognise that the judge's approach to the burden of proof has likely influenced his assessment of all of the evidence and consequently affected all of his findings.

Consequence of the errors of the judge

175. Mr Hobbs contends that the errors made by the judge are very serious and betray a fundamental lack of appreciation of the correct approach in law to the issues he had to decide. He continues that, having regard, inter alia, to the accepted fact that keyword advertising is not inherently objectionable from a trade mark perspective; to the ubiquity of competitive brand bidding as a commercial practice and so also, inevitably, the widespread recognition by consumers of the presentation on a SERP of both natural search results and competitive offerings; to the extensive reputations and consumer recognition of M & S and Interflora as independent businesses; and to the absence of any evidence of actual confusion arising from the activities of M & S the subject of these proceedings, we should allow the appeal and dismiss the action.
176. Mr Silverleaf submits that any errors of the judge are minor and have not affected the ultimate conclusion to which he came. He contends that, having regard to, amongst other things, the exceptional performance of "interflora" as a keyword; the findings of the judge in relation to each of the three factors to which the Court of Justice has said the national court must pay particular regard; and the judge's approach to the overall assessment, we can be confident that his ultimate conclusion is correct and has not been materially affected by any errors he has made, and that is so whether those errors are considered individually or collectively. Mr Silverleaf also emphasised to us that we have not had the benefit of seeing or hearing any of the witnesses give their evidence and we have not had an opportunity to consider all of the documents, and so we should be very slow to interfere in any way with the evaluation the judge has made.
177. In assessing these rival submissions, we accept that appellate courts should not interfere with findings of fact made by trial judges unless compelled to do so. This is a matter we have had well in mind throughout. We also accept that, in a case such as this, a trial judge is called upon to make an evaluative or qualitative assessment from which an appeal court should be very cautious in differing. We acknowledge too the guidance given by the Court of Justice in its decision in this case, Interflora (CJEU). Nevertheless, we are far from confident that the judge would have come to the same conclusion had he not made the errors we have identified. We say that for the following reasons.
178. First, it is, we think, important to have well in mind that keyword advertising is not inherently or inevitably objectionable from a trade mark perspective, a matter the judge himself acknowledged in carrying out his assessment in the main judgment at [288].
179. Second, in considering the perceptions of the average internet user, the judge found that internet literacy had increased over the last 5 years. He put it this way at [290]:

"290. More specifically, counsel for Interflora submitted that the evidence showed that many internet users in the UK do not appreciate the distinction between the natural search results and the paid advertisements on Google's SERPs. In my view, this is something that is likely to have changed over time. As the Ofcom reports show, as a general proposition, internet literacy has steadily increased over the last five years. Counsel for M & S submitted, and I agree, that many internet users learn by doing. In general, I consider that internet users are more likely to be aware both that there is a distinction and the broad nature of the distinction now than they were in May 2008. Considering all the evidence, however, I conclude that, even now, a significant proportion of internet users in the UK do not appreciate that, unlike the natural search results, the advertisements appear on the SERP because the advertisers have paid for the advertisements to be triggered by a keyword consisting of or related to the search term entered by the user."

180. We would note that his finding must be based, at least in part, upon the evidence of Mr Rose which the judge took into account in the manner we have described. Further, the finding is a qualified one, namely that a significant proportion of internet users do not appreciate that advertisers have paid for their advertisements to be triggered by a keyword, not that such users cannot distinguish natural search results from advertisements, however they may have got there. This latter point was something to which the judge returned at [316] where he accepted that the majority of internet users were aware of this distinction.
181. Third, another general point to which the judge attached importance is made by him at [293] in the passage we have cited above (at [40]). Here he appeared to consider it significant that M & S had made no attempt to prove that customers who search using the term "interflora", but click through to the M & S website and end up ordering flowers from M & S, do so because they have been persuaded by the advertisement to order flowers from what they appreciate is a competitive service. It is true that the judge went on to qualify this observation but it indicates to our minds that the judge was very conscious of his finding that the burden (if only the evidential burden) of proving non infringement lay on M & S. Indeed we believe that the attention the judge devoted to the burden of proof indicates that this is an issue which underpinned much of his reasoning.
182. Fourth, and turning now to the particular matters relied upon by Interflora, there were two which the judge considered particularly significant, and the second of these was the Hitwise report which, together with the evidence of Mr Pandya, the judge addressed from [301] to [306] in the manner we have described, and, as he did so, he introduced initial interest confusion. We are satisfied that this was highly relevant to his assessment of the likelihood of confusion.
183. Furthermore, despite all of the foregoing and having considered all of the matters to which the Court of Justice had indicated he should have regard, together with those matters relied upon by Interflora, the judge's overall assessment (at [310]) was that the majority of consumers who click on M & S's advertisements do so because they have been persuaded to take their custom to M & S and not because they believe that M & S is part of the Interflora network. Nevertheless, as he went on to explain, that did not exclude the possibility that a significant proportion did believe that there was a connection between M & S and Interflora. Then, in his overall conclusion (at [318]), he held that a significant proportion of the consumers who searched for "interflora" (and like signs) and then clicked on M & S's advertisements displayed in response to those searches, were led to believe, incorrectly, that M & S's flower delivery service was part of the Interflora network. This was therefore a finely judged decision and we are satisfied that it depended in significant part upon those particular findings in relation to which the judge had earlier fallen into error.
184. We have also given anxious consideration to the possibility of simply determining the issue of infringement for ourselves, a course which would have obvious benefits in terms of time and cost. It is with regret that we have decided that is not something we can do. In that connection we consider there is force in Mr Silverleaf's submission that we have not been taken to all of the documents, and we have not had the benefit of hearing the oral evidence of the witnesses. This is a case of considerable importance to the parties, and it is one in which they have invested substantial time and resources. We therefore conclude that we have no alternative but to allow the appeal and remit the case for retrial. We find limited consolation in the thought that such a retrial should now be relatively straightforward.

Injunctive relief and negative matching

185. We now turn to the three issues considered by the judge in his second additional judgment and begin with the use by M & S of the various Google algorithms, including advanced broad matching, without negatively matching the word "interflora". We have outlined the nature of those algorithms and the negative matching feature earlier in this judgment at [17] to [18].
186. To recap, Interflora complain not just about the bidding by M & S on the word "interflora", phrases including the word "interflora" and variants of the word "interflora"; they also complain that M & S has bid on other generic flower related terms without negatively matching the word "interflora". The effect of such bidding may be that, when a consumer searches for "interflora", an M & S advertisement is displayed.
187. Accordingly, the question considered by the judge in his second additional judgment was whether M & S had used the word "interflora" by so conducting itself that an M & S advertisement was displayed when an internet user searched for "interflora", whatever the precise keyword M & S had chosen. This was a matter about which the parties vigorously disagreed, as had become clear by the time of that judgment.
188. The judge decided this issue in favour of Interflora. He considered that by bidding on flower related generic terms and by failing to negatively match, M & S had still used the word "interflora", albeit less directly, and this was so because what M & S had done had the object and effect that a search by a consumer for "interflora" resulted in the display of an M & S advertisement. He put it this way:

"11. I accept that the position is different, and less straightforward, in the scenario presently under consideration, because in this scenario M & S has not selected a keyword which is identical with, or even similar to, the Trade Marks. Nevertheless, I consider that M & S does "use" the sign "interflora", albeit in a less direct way. What M & S is doing has the object and effect that a search by a consumer for "interflora" results in the display of M & S's advertisement. Thus the SERP will contain both the sign "interflora" and M & S's advertisement. In my judgment it is immaterial that this is achieved by a technical means which does not involve M & S selecting "interflora" itself as a keyword. I would add that I do not consider that this reasoning is inconsistent with what the CJEU said in Interflora (CJEU) at [47]."

189. Mr Hobbs argues that in so finding the judge fell into error. He submits that the imposition of an obligation on M & S to negatively match the term "interflora" in relation to advertisements for its flowers and flower delivery services pre-supposes the existence of some sort of legal duty or obligation which requires an advertiser using keyword advertising to opt out of or into Google's broad matching or negative matching facilities. He continues that no such duty or obligation is to be found in EU or national trade mark law. Moreover, this impugned activity does not involve any selection by M & S of "interflora" as a keyword, and such selection is an essential requirement of liability. What is more, this newly devised legal duty or obligation purports to turn back the tide with regard to keyword advertising as practised for many years, both before and after Google's change of policy in May 2008, for it contradicts the lawfulness of the long standing practice in which the selection by traders of a generic term as a keyword has resulted in the display of their advertisements in response to searches by consumers using a brand name together with that generic term. Accordingly, says Mr Hobbs, the judge ought to have found that this activity cannot amount to trade mark infringement.
190. We find ourselves unable to accept these submissions. The issue of what constitutes use of a sign in the course of trade in this context was considered in some detail by the Court of Justice in Joined Cases C-236/08 to C-238/08, Google France. We have set out the critical parts of the reasoning of the Court at [71] to [73] above. As the Court explained at [51] to [52], [56] and [69], the choice and selection by a third party of the sign as a keyword has the object and effect of displaying an advertisement, and this constitutes use of the sign in the course of trade. By contrast, the referencing service provider, Google France, was not using the sign, and that was so because use by a third party of a sign implies, at the very least, the use by that third party of the sign in its own commercial communication. That was not something that Google France did. This essential reasoning has been followed by the Court in all of its later decisions on keyword advertising to which we have referred.
191. In our judgment it follows from the foregoing that the choice and selection by a trader of a generic term as a keyword cannot be considered in isolation and as an activity separate and distinct from the Google algorithms and match types used in relation to them. Further, account must also be taken of the opportunity available to the trader to negatively match. If, having regard to all these matters, a trial judge comes to the conclusion that a trader has chosen a generic term with the object and effect of triggering the display of his advertisement in response to a search by a consumer of a term including or consisting of a sign which is the same as or similar to the trade mark of a rival then we consider it is permissible for the judge to find that the trader has used that sign in the course of trade, although indirectly and by different technical means. It is our view that in these circumstances the judge may properly find that the trader has used the sign in the course of its own commercial communications.
192. In the present context, and as the judge observed, it is not said that the generic terms selected by M & S as keywords are the same as or similar to any of the Trade Marks. Instead, complaint is made of its choice and selection of such generic terms as keywords without negatively matching the word "interflora". We have come to the conclusion that bidding by M & S on generic flower related terms without negatively matching the word "interflora" may be characterised as use by M & S of the sign "interflora" within the meaning of Article 5(1) and (2) of the Directive (and Article 9(1) of the Regulation). Whether it does or not must depend upon an assessment by the judge to whom this case is remitted of all the circumstances, including the length of time for which the impugned activities have been carried on. If, in the light of that assessment, it is concluded that the object and effect of M & S's activities has been to trigger the appearance of its advertisements in response to searches for "interflora" then we believe that this would justify a finding that M & S has used the sign "interflora" as a matter of trade mark law. We would emphasise, however, that this would not in and of itself be sufficient to establish infringement because, as we have sought to explain, it must also be shown that the advertisements did not enable the average consumer to ascertain whether the goods or services referred to originated from Interflora or from M & S. Here again the judge must consider all relevant circumstances including the length of time for which the impugned or similar activities have been carried on.
193. The other two matters raised by Mr Hobbs concern the scope of the injunction granted by the judge. In light of our findings it is not strictly necessary to decide these on this appeal. Nevertheless, we have heard argument upon them and so we think we should express our views, albeit briefly.
194. Mr Hobbs first complains about the geographical scope of the injunction granted by the judge in relation to the Community trade mark. As the judge observed, the correct approach to this issue was elaborated by the Court of Justice in Case C-235/09 DHL Express France SAS v Chronopost SA [2011] FSR 38. The Court explained (at [39] to [45] and [50]) that the scope of the prohibition against further infringement or threatened infringement of a Community trade mark extends, as a rule, to the whole of the EU. But, it may, in certain circumstances, be restricted:

"46. However, the territorial scope of the prohibition may, in certain circumstances, be restricted. The exclusive right of a Community trade mark proprietor, as provided for under Article 9(1) of Regulation No 40/94, is conferred in order to enable that proprietor to protect his specific interests as such, that is, to ensure that the trade mark is able to fulfil its functions. The exercise of that right must therefore be reserved to cases in which a third party's use of the sign affects or is liable to affect the functions of the trade mark (see, to that effect, Joined Cases C-236/08 to C-238/08 Google France and Google [2010] ECR I-0000, paragraph 75 and the case-law cited).

47.       It follows, as the European Commission has pointed out, that the exclusive right of a Community trade mark proprietor and, hence, the territorial scope of that right, may not extend beyond what that right allows its proprietor to do in order to protect his trade mark, that is, to prohibit only uses which are liable to affect the functions of the trade mark. The acts or future acts of a defendant (namely the person whose use of the Community trade mark is complained of) which do not affect the functions of the Community trade mark, cannot therefore be prohibited.

48.       Accordingly, if a Community trade mark court hearing a case in circumstances such as those of the main proceedings finds that the acts of infringement or threatened infringement of a Community trade mark are limited to a single Member State or to part of the territory of the European Union, in particular because the applicant for a prohibition order has restricted the territorial scope of its action in exercising its freedom to determine the extent of that action or because the defendant proves that the use of the sign at issue does not affect or is not liable to affect the functions of the trade mark, for example on linguistic grounds, that court must limit the territorial scope of the prohibition which it issues."

195. The following points emerge from this passage. First, it must always be borne in mind that the injunction may only prohibit uses which are liable to affect the functions of the trade mark. Second, and as a consequence, if the court finds that the acts of infringement or threatened infringement are limited to a part of the EU, perhaps because the claimant only complains about the activities of the defendant in that part of the EU, or if the defendant proves that the use of the sign complained of is not liable to affect the functions of the trade mark in all territories, perhaps for reasons of language, then the injunction must be limited. It seems to us that the judge had these points well in mind in deciding the territorial scope of the injunction in this case.
196. Mr Hobbs also contends the judge fell into error in his approach to the form of the injunction and, in particular, in granting it in conventional general form. He argues that such an injunction has the effect of restricting (if not completely preventing) the right of M & S to engage in keyword advertising and so to offer to consumers an alternative source of flowers and flower delivery services. Mr Silverleaf responds that the judge considered all of the competing interests carefully and properly.
197. We consider that, at this stage, it is not appropriate to do more than make the following general comments. We accept, as does Mr Silverleaf, that Article 3(2) of the Enforcement Directive requires the court to ensure that any injunction is proportionate. We also accept, as again does Mr Silverleaf, that the right to advertise (including by keywords) may engage the right to freedom of commercial expression. No doubt the court considering the scope of any relief in light of such infringements as may be found and carrying out the proportionality assessment will take into account, on the one hand, the need to ensure the protection of Interflora's rights and provide an appropriate deterrent, having regard to the principles of equivalence and effectiveness, and, on the other hand, the right of M & S to engage in legitimate competitive keyword advertising and the potential chilling effect of an injunction upon that right.

Overall conclusion

198. We have therefore come to the conclusion that the appeal must be allowed and the case remitted to the High Court for retrial of the claims for infringement under Article 5(1)(a) of the Directive (as implemented in section 10(1) of the Trade Marks Act 1994) and Article 9(1)(a) of the Regulation.