Digital, Culture, Media and Sport Committee
Oral evidence: Fake News, HC 363
Tuesday 16 January 2018
Ordered by the House of Commons to be published on 16 January 2018
Watch the meeting
Members present: Damian Collins (Chair); Julie Elliott; Paul Farrelly; Simon Hart; Jo Stevens; Ian C. Lucas; Christian Matheson; Rebecca Pow.
Questions 86 - 188
Witnesses
I: Bethan Crockett, Senior Director, Brand Safety and Digital Risk, GroupM EMEA, Eitan Jankelewitz, Partner, Sheridans, and Matt Rogerson, Head of Public Policy, Guardian News and Media.
II: Tim Elkington, Chief Digital Officer, Internet Advertising Bureau, Phil Smith, Managing Director, Incorporated Society of British Advertisers, and Ben Williams, Adblock Plus.
Written evidence from witnesses:
Witnesses: Bethan Crockett, Eitan Jankelewitz and Matt Rogerson.
Q86 Chair: Good morning. Welcome to this further evidence session of the Digital, Culture, Media and Sport Select Committee on our inquiry into fake news. The focus of our evidence sessions today is very much the role that the advertising market plays in supporting the distribution of news on the internet and, in particular, concerns around the placement of digital advertising alongside sources of fake news and other problematic content. We were very interested to see the interest that advertisers have taken in this issue and their concerns around brand safety in advertising appearing on sites that contain problematic content. I wonder if I could start with you, Bethan Crockett, on that point. Procter & Gamble in America, in particular, have raised a number of concerns around brand safety. What does WPP do, in terms of the services it provides to its clients, to ensure brand safety online?
Bethan Crockett: Absolutely, I am happy to share. At WPP—and I sit within GroupM, as well, which looks after our four main big media-buying agencies—we work directly for clients. Brand safety obviously became a huge topic last year, 2017, because of all the reporting. It was always an issue; it has always existed. It is about the unintentional misplacement of an ad next to content that is inappropriate.
We define that contextual issue into three categories. The first category is illegal content. There is definitely illegal content online, for example IP-infringing sites, which are very obvious. Then there is a second level that we determine is illegitimate, which is generally types of content that an advertiser or a brand would not want to appear against. This can be content that is overly sexual or it could be focused on terrorism, extreme violence, extreme political views and so on. Then we have a third level, called, “Inappropriate to the brand.” As you know, certain brands want to avoid certain content for various reasons, and because it does not match up to their brand values. That does not necessarily mean it is illegal or that there is something terrible with the content, at that level, but it does not match their brand values.
We take those three elements and we have a policy that we share with all the clients whereby we recommend they take a multi-layered approach. The first layer is inventory choice: make the right decisions around where your advertising should run in terms of how you buy the placement of ads. The second level is around using baseline brand safety that we refer to often as “whitelist” and “blacklist”, which is intentionally listing out a number of sites or apps where you definitely do not want to appear, or placements, websites and apps where you do want to appear, and balancing both of those.
As a third level we recommend using content verification technology. There are several players in the market. They are independent and they are audited so their technology does what it claims to do. They will look at the content of the webpage or app, the actual text on the website or app, the metadata behind delivery of the ad, and the referral URLs to and from the website. They have evaluated that over time and given it a score for whether it carries this type of illegal or inappropriate material, or just non-matching categories that some advertisers wish to avoid. They can use that score and that history to block an ad appearing. They will know that the website or app falls into a bucket and they will opt not to place an ad.
We recommend that multi-layered approach and then we support them internally with our own strategies around procurement of inventory and who we trade with, and around creating those blacklists ourselves. We create a global blacklist that looks at IP infringement and that blocks off over 250,000 websites and apps by default. We will offer other, optional lists that I am sure we will come on to today, such as controversial news blacklists, where we make an offering and a suggestion around controversial sites that may be intentional disinformation or just extreme in their views. We cannot enforce that—and we will probably talk about that today—because what is the legal definition of a fake news site? That makes it a bit more difficult for us to make a final decision on behalf of our clients.
We will look at new technologies coming into the space, test those and use those, such as fact-checking technologies, which we might touch on today, and quote validation technologies. Again, can they do what they claim? We make sure and then we offer those to clients. Then when it comes down to a manual overview—especially in what we refer to as “programmatic”, the automatic buying space—we will do human validation and checks, using various data points, to determine whether we think a site is legitimate enough to receive monetisation through advertising.
That is the policy that we take to all clients. We help educate them, run some of the baseline by default and recommend third levels of extra technology that should be in place. Sometimes clients may opt not to, depending on their own approach around brand safety and what they are willing to pay for in that space.
Q87 Chair: From what you have said, it sounds like there are two pretty clear categories. There is the dangerous and illegal content, which is probably, using technology, relatively easy to identify and therefore protect clients from. There would then be another area that is harder to define. The difference between a fake news website and a clickbait sensationalist site full of stories that may be true but slightly sensational is difficult to determine. Do you believe there is a technological solution to that, or is it about human checking and identification of sites that are known to be problematic in terms of fake news content?
Bethan Crockett: Technology solutions definitely exist. We are looking at fact-checker technology solutions that will validate the number of facts within a site to say whether this is genuine information or misinformation that is not true or has no factual evidence to support it. We are also looking at technologies that look at quote-checking, the validity of quotes throughout, whether they have been used on various other sources and therefore whether it looks legitimate.
It is difficult, though, because we do not have that legal definition of a fake news site. For example, I have here—we can share it afterwards—one company that has already divided fake news into nine categories. We end up in this weird space where we would like as an industry to have a definition for “fake news” that includes those sites that are deliberately and intentionally sharing misinformation with the aim of influencing in some way or undermining public security or public institutions, and one that would not pull in the clickbait sensationalist headlines that have always existed, fun headlines that are used by the tabloid press to gain readership and attract attention. That is very different from what we need to address in that space. We would like to have clearer guidelines and ways of saying, “Yes, this is what would count as that type of fake news” so that we can build a technology to look for that, to validate that and to create those blacklists and blocklists that we can use across the board.
Q88 Chair: It might be helpful to ask this. What are clients buying when they are buying online advertising? I understand that if I am buying television advertising I am buying an audience. I might be buying some specific spots but I am probably buying a bundle of spots that will reach a certain audience. Is that what I am buying if I am advertising online as well? Am I buying an audience and a bundle and there could be all sorts of sites and people mixed in with that?
Bethan Crockett: Correct. You could either be buying a specific placement on a specific site—you have that option, a direct buy or a direct partner, knowing which category or partner site you wish to buy and run your advertising space on—or you could be buying an audience, which is what a lot of advertisers want to do because, as you just shared, that is a traditional way of buying your advertising across TV, print and radio. They wish to do that on digital as well, more and more.
You have a lot of people buying those audiences that have engaged with their content, with their product. That might be linked to their own users that have registered and used their products and services. They want to find them wherever they go, but they do have a limit and that is where they say, “I do have some limits on brand safety.” That is where they will want to put into place those blacklists, whitelists and a certain level of technology to protect themselves. Yes, they want to reach their audience, but not at all costs.
Q89 Chair: As things stand at the moment, if a client said to you, “My expectation is that my advertising will not appear on any sites that distribute fake news”—as you say, there is no legal definition of what fake news is—could you guarantee that to a client?
Bethan Crockett: No. We can say to them that we can offer our controversial/fake news list. You can see everything we have placed in that list, which we would consider fake news. We continue to grow and update that list monthly based on independent technology feeding in. We can use third-party verification partners who have created categories themselves, scoring sensationalist news or extreme political news, and use those to block. Then we can ask more of our partners to enforce those standards as well. However, there is the potential that someone could set up a website, pull some content that is not factually accurate, place it on their website and opt into a monetisation programme and it might not be picked up. That might not necessarily be the intentional disinformation that I think we need to think about; it might just be someone copying and pasting an article that they agree with but that is not factually correct.
Q90 Chair: The Committee has been given different written evidence and private briefings on this subject. Is it not possible to look at the characteristics of certain sources of information? Could I say, if I am a brand, that I do not want my advertising to be placed on a site that is maybe being operated out of Macedonia, Moldova or Russia, where there is no clear identification of who the author or owner is, and which might be a legitimate site but in terms of brand safety I would rather not be associated with? Is that a line of defence you could build into the system?
Bethan Crockett: Yes, you could build into the system lines of defence looking at where a site has been built, where it is registered, and you could choose to block geographies. That is an option that we could look at.
Q91 Chair: Is that something that is done now, routinely?
Bethan Crockett: It is done in a slightly different way, in the sense that many advertisers will only want to target their audience within their marketplace. They will be positively targeting a UK audience or UK-registered sites as opposed to just running them across the board. It is positive geography tailoring rather than exclusive geography tailoring.
Q92 Chair: OK, but if someone is running an international campaign and has not specified a certain geography because that is where their audience is, that is not offered as a standard matter of course?
Bethan Crockett: No, that would not be a standard procedure right now. I have not heard an advertiser raise that as a question either, to date. I have not had an advertiser come in and ask if we could exclude geographies based on the source of the sites.
Q93 Ian C. Lucas: I have been struck by your very comprehensive replies, but you have not mentioned the platforms themselves, which is quite telling. Do they play any role in this process at all, as far as you are concerned?
Bethan Crockett: Yes, as far as I am concerned, absolutely. They have a big role to play. We work very closely with all the platforms, as you can imagine. Advertisers and brands wish to use the platforms to reach audiences and we work with them to deliver their advertising across those platforms. We meet with the platforms regularly.
Right now we are meeting the YouTube senior team weekly and we are asking them for very specific requirements that we want to see to allow brand safety. We are advising our clients currently, until they see these changes, to be very careful about making the decision to run advertising on a platform that is social, with uncurated content that is uploaded in a lot of volume, that does not have third-party verification and that is not pre-checked before being loaded. We demonstrate the risk there to the client, we explain that there is a limitation on the brand safety controls that we would like to run and that we can run across all other types of websites and apps, which puts them at further risk, and we also recommend they think about other options.
One option we have engaged and got YouTube to agree to is a company called OpenSlate. They have direct API interaction with YouTube and are able to evaluate each of the videos for what the content actually is. They measure a channel—YouTube sells a channel that has multiple videos in it—and give a score to that channel for its relative safety. We can help give some guidance to clients on where there is relative safety compared to other channels.
Q94 Ian C. Lucas: It sounds like you have quite a long list of things that you would like the platforms to be doing.
Bethan Crockett: Yes.
Q95 Ian C. Lucas: You are negotiating with them but they are in a very strong negotiating position, are they not?
Bethan Crockett: They are in a very strong negotiating position but we can ask—
Ian C. Lucas: Some 20% of the advertising goes on—
Bethan Crockett: We are seeing more and more clients hesitating across platforms now. They are starting to realise the inherent risk and they are starting to query whether they should run there. Can they run elsewhere where they can get the same audience reach and delivery of their messaging? Can they move away from using some of these platforms until the controls are improved?
Ian C. Lucas: Thank you.
Q96 Chair: Just on YouTube, going back to my earlier question, can people buy in different ways? Can a client say, “I see a clear association with a particular YouTuber who has a very big following and produces live content consistent with the interests of my brand, and therefore I want to advertise around that channel or channels like it”? Can people buy in that way as well as buying a general bundle of the audience who use YouTube and are generally interested in similar things?
Bethan Crockett: Correct, there are both ways of buying. YouTube and other platforms will offer out their inventory in different bundles and different ways of buying.
Q97 Chair: From what you are saying, it sounds like the area of nervousness for clients would probably be around the second of those, where you are not buying something specific but just buying a general audience.
Bethan Crockett: Correct. That is the biggest concern for probably all the clients. If they are following their audience, can they follow their audience in the safest possible environment rather than risk following them into an area where their audience is seeing something they would not normally want to see their advertising against?
Q98 Chair: I imagine as well that for Facebook and YouTube it is harder for agencies to control where the advertising goes because it is a closed system and you do not have access to the data.
Bethan Crockett: Exactly. That is one of our big asks to the platforms: better viewability and access to the data, as well as third-party validation of the data.
Q99 Chair: How seriously do you think the platforms take the challenge of marketers like Marc Pritchard saying, “If we cannot have guarantees and transparency on this then we will take our business somewhere else”? Do you think that is possible? Has that changed their attitudes?
Bethan Crockett: It has definitely changed their attitude over the last year, with more and more exposés of what can happen on the platforms. As in many cases, the influence will happen when the money is withdrawn. When there is substantial change to revenues coming in there will be more action, but I think now they are looking at trying to resolve many of these issues.
Q100 Christian Matheson: Very quickly, the process that you describe, Ms Crockett, sounds very labour-intensive. It sounds, in what is obviously a very fast-moving sector, as if it would be difficult to keep on top of where you are going to be placing some of these. Is it as labour-intensive as it sounds? Can the process be automated or are you and your team constantly having to check and reassess the sites you look at?
Bethan Crockett: The new technology that I have touched on with the blacklists and the whitelists is very automated, so that is not labour-intensive for the group. The third-party verification technology used is real-time technology running and scoring consistently, making real-time decisions on blocking. That is running at the same time as the advertising. That is not manual; that is the technology doing the work. The manual piece comes when we cannot run that type of technology. Then we have to step in. That is where we have challenges with platforms or networks who will not allow you in with a third party to do the automated work.
Q101 Christian Matheson: It is not like you are fighting a losing battle, then. It is not that there is a tide that is coming in. You can stay on top of it.
Bethan Crockett: You can never guarantee 100% brand safety. You cannot guarantee that in print, on the radio or on television. You could always have a slight mismatch of the brand or the chance of an advertiser showing up next to a news story—a breaking news story is always the example given—that does not play well for the advertiser. You cannot guarantee it 99% no matter what the medium. In digital, there is so much volume and so much room for participants that the challenge is slightly harder, but I believe technology can do a lot of that work. We need strong definitions we can match the technology to, to help us do a lot of that work. Then there is work to be done with other platforms and networks so that we can use the technology in that way on the various platforms.
Q102 Chair: Who needs to generate those strong definitions? The industry can generate those themselves, in a way.
Bethan Crockett: We can generate a definition but we cannot put the legal standing behind it. For IP-infringing sites, I can go and read the legal definition of what an IP-infringing site is and measure the site against it. I can say if it is infringing, black and white, yes or no. For fake news right now I can have a definition that we commonly agree on and I can look for sites with those traits, but I could also easily be pursued by one of those sites saying, “You do not have the right to block. Where is the legal standing on this?”
Q103 Chair: The site would have the grounds to challenge legally your decision not to place your advertising on it?
Bethan Crockett: We could argue the case that we are making a decision about placing advertising, which is our decision to make, but they would say, “You are commercially undermining us and you do not have the right to say we are a fake news site.”
Q104 Jo Stevens: You mentioned networks where you cannot run this third-party checking system. Are any of the major platforms that we think about, like Google or Facebook, such an example?
Bethan Crockett: Yes. We are encouraging them more and more to let us use these third parties. They are starting to let us use some of the third parties, but it is not across the board.
Q105 Jo Stevens: So Google will not? Facebook will not? YouTube?
Bethan Crockett: Twitter, Instagram; yes, a lot of the social platforms. They have their own position and concerns about why they do not want to allow those technologies in. It would definitely be a conversation with them about why.
Jo Stevens: Thank you.
Q106 Chair: Eitan, this is an interesting issue about the lack of a legal definition and whether that makes it harder to guarantee brand safety. Perhaps you could say a little bit about your work at Sheridan’s and the views of your clients on that.
Eitan Jankelewitz: I am a partner at Sheridan’s. We are a technology and media practice and I am advising quite a few adtech companies and advertisers who are advertising online. If a buyer of media decided not to place ads with a company, it is a commercial decision. There is probably a complicated competition law question there as to whether it is possible that buyer was abusing a dominant position that could prejudice other people in the market but it is, generally speaking, a commercial decision to be made by the advertiser.
Q107 Chair: They are not legally compelled to advertise on some of the sites because they are there and they fit the audience definition?
Eitan Jankelewitz: Not as far as I am aware.
Q108 Chair: No. It would seem odd if they were. It would be like saying that if you advertise in every newspaper apart from The Daily Express, The Daily Express could have legal recourse against you for not choosing its title.
Eitan Jankelewitz: Absolutely. From a competition point of view, if there were group decisions being made at an industry level that were beyond general market practice, decisions being made in concert to abuse a position of dominance, then in that scenario it is possible but it would not be a common scenario.
Q109 Chair: In terms of an agreement between an agency and its clients, really that is a commercial agreement between those two parties over how they try to manage brand safety issues online.
Eitan Jankelewitz: Yes, and different advertisers will have different risk appetites. Different agencies and ad exchanges will be able to offer different levels of protection. For example, at the most basic level, if it is an ad for an over-18 film, it should not be shown to kids—these kinds of things. Different ad exchanges, ad networks and ad agencies are able to access different levels of protection.
Q110 Chair: Are you seeing any changes in attitudes of clients in the contracts with their agencies over some of these issues? Do they want to build brand safety into their contractual relationships with their agencies?
Eitan Jankelewitz: From a brand safety point of view there is certainly an inclination towards those protections, just making sure there is not going to be that awkward article or bad exposure as a result of ad misplacement. Overall, some advertisers rely quite heavily on account management, I find. It may be that they are taking guidance from their ad agencies in order to manage their campaigns.
Q111 Chair: Is there a question for clients around whether the platforms should be able to guarantee a level of brand safety? Those are closed networks and you do not have access to the data to micro-target your placement. If brand safety was compromised because of the way in which the advertising model works on Facebook or on YouTube, could there be some protection in law for brands or some recourse they would have against the platforms for allowing that to happen?
Eitan Jankelewitz: Overall, as you have pointed out, the opacity in advertising is now under the spotlight. The idea of performance marketing and advertising online is basically paying for the viewership of your ad. Aspiring to a scenario where you can take your advertising spend and attach it to each unit that you sell is a worthwhile thing to try to pursue. It just turns out that it is not as accurate as people want. Ad fraud exists and there are loopholes that are now being gamed.
Advertisers are definitely looking for more information on what is going on when they place ads, where they are going to and where the traffic is. If an advertiser places an advert with a publisher, they want to understand where that publisher has itself acquired traffic. Instead of viewing it as an individual season advert—it is led towards an advertiser because they click on a link—it is more of a chain of publishers where people are arbitraging their cost of acquiring traffic. In that case the publisher is acting like an advertiser themselves and then, once they have that viewer on their website, also trying to generate more revenue from that viewer and make a profit. People are jumping in and out of this stream all the time. For that to happen, there has to be a problem with the way that advertisers are spending their money.
Q112 Chair: Perhaps, Matt, you could give me the perspective of The Guardian on this as well. It is an interesting point about publishers seeking to raise money from their audience but also effectively cross-selling to them as well, through other people’s advertising. How does a business like The Guardian manage that relationship?
Matt Rogerson: The safest and best way for any brand to buy online advertising is to go direct to the publisher. In direct buys at The Guardian, you see miniscule levels of fraud. There are very few issues with buying direct at The Guardian. The issue, as you are right to point out, is where you aggregate an audience, either via an open ad exchange or via products such as Facebook Audience Network, where the buyer is effectively sold an audience while being completely blind to where that advertising lands.
In terms of the focus on fake news, our viewers focus on the counterpoint to that, which is trusted news. There are various facets of trusted news that are clear in terms of there being an appeal process and a reader’s editor in place or some kind of point of contact where you can get in touch with the organisation to suggest that an article is wrong in one or another respect.
The challenge for us is that in an open exchange auction at the moment, that sort of trusted news is not marked out for buyers of advertising from the rest, from the long tail. That is a real challenge because we are treated the same as a long-tail website that is either a ripped article that is pasted in an attempt to be monetised or a piece of what is just fake news. There is no divide for someone at the other end of that transaction. That is a real issue if you are trying to encourage investment in high-quality content where there is accountability versus the long tail of the web, and that is a debate we have entered into recently as an industry with Google to see what they can do to make it much easier for advertisers to see the difference between trusted content and the long tail.
With Facebook it is much harder for them. One of the reasons that we as a publisher came out of a product called Instant Articles was because it did not yield very much revenue for the business, but also because we had no idea where the advertising was coming from. Advertising would be injected into the Instant Articles unit but there was no accountability or transparency about who those advertisers were. If you are a publisher, especially if you are a premium publisher, that makes you extremely nervous because we are as concerned about brand safety and who is advertising next to our content as advertisers are on the other end.
Q113 Simon Hart: The bit of this I am not entirely clear about is how much work has gone into measuring the impact of all this, for example where a legitimate brand might have mistakenly appeared in an illegitimate location, if I can describe it like that. Does anyone know what effect that is having on the profitability or the success of that particular company? Has anybody done any work on that?
Bethan Crockett: I have not seen any industry-wide studies to demonstrate that. What we can say, and it is very interesting, is in the recent events on YouTube, when the research was done with consumers about how the harm was perceived. This was when there were the terrorist YouTube videos, way back in March. The brand’s reputation was seen to be at fault more than YouTube, but we could not quantify it.
Q114 Simon Hart: I get that when it is blatant, but sometimes when it is a fake news site it is not even clear that it is a fake news site. If a legitimate brand appears on something that does not look particularly legitimate, what is the downstream damage in those circumstances? Are advertisers being a bit cynical and thinking, “We will appear to be horrified if somebody raises it with us, but does it really matter?”?
Bethan Crockett: There are several global brands that are genuinely very concerned. They are concerned that it impacts their brand reputation. They have not quantified that to give us a way of saying what impact that was.
Q115 Simon Hart: There was a bit of a thing the other day over Paperchase. I know it is not an online comparison, but it advertised in the Daily Mail or the Mail on Sunday—something like that—and there was a minor Twitter storm, whereupon they went into a complete panic. In the end, the brand damage was probably more profound in the retreat than it was in the original placement. Should people not be doing it more scientifically than just imagining what the brand damage could be, which seems to be where we are at the moment?
Matt Rogerson: The whole regulatory system for advertising is based on reputation and the damage to PR.
Q116 Simon Hart: Perceived or real?
Matt Rogerson: Perceived, I think. The ASA regime is based on the idea that if they rule against you then there will be negative PR as a consequence. A recent study, published a couple of weeks ago, said that Facebook was the key vector for fake news, so I will just focus on them for a minute. Their client base in terms of advertisers is generally SMEs, not major brands. If you think about how the system of regulation works, if it is about brands being connected with content that fulfils the values that we as a society believe are right, then with a platform like Facebook where it is not major brands that are advertising but SMEs, that dynamic changes a bit. It is not Paperchase that is going to be in the headlines if their advertising appears next to something on Facebook, if, by the way, you ever see it next to the piece of content on Facebook, because there is no transparency and limited accountability around that. It does bring into question how the regulatory regime will work going forward.
What is interesting is that we have come to the end of 2017, when brand safety was a big issue, and there are some in the industry who just say, “Brand safety does not matter, citizens do not care and we do not need to worry about it.” My view is that values in the media ecosystem would collapse if that were true. Advertising has always been connected to high-quality, premium content. That is where advertisers have always wanted to be. If advertisers decide they are just going to follow the audience wherever they go, that has very serious implications for the sort of content that is funded. As a premium publisher, that is a point of major concern.
Q117 Simon Hart: I have just one last question on that. Is it the advertiser who has to take full responsibility for this? Does the advertiser have to be considerably more careful about checking out where their stuff is going than they currently are? Can we pin the responsibility purely on the brands, whether they are SMEs or big brands?
Matt Rogerson: You cannot do that without transparency in the value chain of pragmatic advertising because the opacity in the value chain means that when they are sold a bundle of an audience, it is very hard for them to see where that money is going. Once there is transparency there and they are able to clearly see where their money is going, then absolutely it is up to them to decide. It is up to the digital advertising industry to give them clearer and easier choices about where their advertising is going.
Q118 Ian C. Lucas: Can I just ask a follow-up question on that? If one buys advertising on, say, Facebook, can one secure a report of where that advertising went or which sites it went to?
Matt Rogerson: I do not know about directly in the news feed but with Audience Network, which from memory is turned on by default when you make a buy on Facebook now, at the moment there is no functionality to get a report on where advertising lands through that product. Certainly as a publisher we could not get a report on what advertising had featured within our Instant Articles units.
Q119 Chair: On that, obviously as a publisher you do not have access to the audience data from Facebook for people who move from being on Facebook to reading an article on The Guardian website, people who share the link and click through. You do not know who those users are when they come through; they do not share that data with you. How can you control what advertising that viewer will see on The Guardian website when they land on the website and you do not necessarily know that much about who they are? How is that process managed?
Matt Rogerson: If they come to The Guardian website and it is out of our own inventory, then we would sell that directly via—how would I sell that directly? Can I come back to you on that in more detail?
There are two types of readers from Facebook. One is where you read a piece of content on Facebook and that is via Recent Articles, and that is when they control the advertising experience within the unit. If the reader comes from Facebook over to The Guardian, then it is our responsibility and we are in control of that advertising. That is where a buyer of advertising would work with The Guardian to display advertising on our website when the audience comes into the website.
Q120 Chair: I appreciate this is incredibly technical and, in all aspects of it, slightly opaque, but if, crudely speaking, I was thinking of buying a car, I had been on the Ford website looking at cars and I then decided to go and read an article on The Guardian website, there is a reasonable chance that while I am looking at an article I might see an advert either for a Ford car or a different brand of car. The programmatic model guesses the things I am interested in and the advertising follows me around the web rather than the way it would be in traditional media, where the advertising is pre-placed and I am either interested or not. In general terms that would be a fair definition of the way the programmatic model works, would it not?
Matt Rogerson: That would be tracking and targeting, yes. Bethan might want to comment.
Bethan Crockett: You are correct, we would not have information about what you had been doing or your preferences from Facebook. That would not be available to the website that you visited after being on the Ford website. If you had been to the Ford website that would mean Ford as an advertiser has dropped a cookie on your computer when you visited the website and they have picked up that you have been looking at their information. Then when they have been running their advertising directly on The Guardian, for example, they have shared that information with The Guardian to say, “If you see these cookies, they have come to the Ford website, so please show our advertising.”
Q121 Chair: Yes, but when they are coming from a closed network such as Facebook, you do not have that information.
Bethan Crockett: We would not have any information on what you had been doing on Facebook.
Q122 Chair: No. In that case, is it much harder for you—The Guardian, say—to monetise that audience? Presumably what they saw would fall back to generic advertising that has been placed on The Guardian site, which is a guess as to what they think Guardian readers are interested in.
Bethan Crockett: Yes, or The Guardian will have picked up traits of their own viewership and be offering those out as audiences.
Eitan Jankelewitz: It is also possible to submit information to Facebook and then target those particular individuals on Facebook or target people who are like them on Facebook. When it comes to following people around the internet with an ad, as it tends to be known, it is a scenario where the advertisers, the publishers and the intermediaries are all dropping cookies on people, they are matching those cookies together to try to understand where the journey is going to and from, and then they are meeting people at their destination. Ford would be able to see that the person who looked at this car is now looking at The Guardian and that is where they provide the ad.
Matt Rogerson: The vague irony is that the relationship is asymmetric, because if you have a Facebook button on your website as a publisher, Facebook gathers information from the publisher which it then uses to target advertising on Facebook, but the same is not true the other way around; the publisher would not get information from Facebook about the activities of the user on their site. All the time you see intelligence and information being gathered up, but it is not being shared down to the publisher.
Q123 Chair: For the sake of clarity, does that imbalance in the relationship favour Facebook commercially and how does that work?
Matt Rogerson: They have the most intelligence on the most people on Earth. I saw a Washington Post story suggesting they had 98 points of data on each user on the platform, which means that whether you are a political party or a brand advertiser, you have the ability to target at a very micro level.
Q124 Chair: So Facebook will be gathering information about its users and the other sites they interact with, but it does not share any of that information with any external partners? You might know about those people separately and individually but you do not know about their activity on Facebook in the same way Facebook knows about their activity on your sites.
Matt Rogerson: Unless you are a buyer of advertising via Facebook, in which case you have access to a console of data that enables you to target specific audiences.
Q125 Christian Matheson: Presumably that is data they do not need to share with anyone because they are so big.
Matt Rogerson: Yes, I think if you get to that position you do not need to share.
Q126 Chair: Bethan, if I could just go back to what we were talking about a moment ago, on the issues around brand safety, following up on the points Matt was raising about, is there any way content could be more effectively banded as being risky, very safe or somewhere in between, almost like a traffic lights system for clients where they could say, “I want to pay a premium to have advertising placed on sites where I know I am very safe, even though the audience is slightly more restricted”?
Bethan Crockett: Yes, absolutely, and that is what we do now. There is a lot of banding of quality of sites. There are a few points I have just noted down. For example, there are networks that will not let you know which websites are within the network, only that the audience is there. You referred to them as “blind.” They will not give you that information. We would advise clients not to run there because, again, it is in that higher-risk bucket. Then there are networks that do allow full transparency. There are options. We would advise transparency over just following your audience without any information on the network.
You also touched on brand safety and whether it is within contracts with clients. It is. With many of our clients now we have contractual obligations put in around brand safety. We ourselves have had, for a long time, contractual obligations around brand safety with our publishers. With The Guardian, for example, they will see in our contract that we ask them not to place our ads on certain types of sites, and if they do they guarantee a very short, tight take-down window, which they agree to because they uphold brand safety. It also asks for non-monetisation and refunds. There are all the contractual controls there focused around brand safety. It is key to recognise that it is not the case that everyone just books their runs and is not looking at these issues from advertising all the way through the chain.
This takes me back to my earlier point and the competition law. It does come back to, at the moment, everyone upholding those standards as individual companies throughout the chain, and then obviously access to certain amounts of data we cannot access so we can do the verification as a separate topic.
Also, if we can establish a legal standard for fake news, for example, then we can bring together a blacklist that the industry can use and there would be no excuse for any player to not be using it. It would not be down to an individual advertiser to say they want it, or down to a network to say they are applying it. That is really important. We have seen where that works with PIPCU and the IWL. Everyone can use that list. We have also seen—and my colleagues will talk about this—that when we have gone to the European Union and talked about creating a list on this level for the industry to use, we have had challenges around competition and pushback, saying, “You cannot, as a group, do that.” We are in that tricky position with fake news where we are asking everyone to step up but we cannot say it is required and we cannot enforce it in a contract ourselves to say, “You have to do this because of this standard.”
Q127 Chair: From what Eitan said earlier, is a legal definition of what “fake news” is necessary? This is ultimately just a commercial agreement between an advertiser and their agency.
Bethan Crockett: If you want the whole industry to accept and work with one list where it is established that these are bad players, these are intentional distributors of news that is intended to send misinformation and influence society in a bad way, and we want everyone to sign up and use this list, this one list, then we get into competition difficulties. It is not necessarily the monetisation decision as an individual advertiser saying, “I want to be off that.” If we want to say and I want to say to every partner I work with, “You will apply this blacklist. It is now the standard,” you get into difficulties around competition.
Eitan Jankelewitz: It is easy to say, “This content is appropriate for people over a certain age” or, “This content is pornographic” or, “This content is violent.” Those things are easy to identify. With fake news there are so many different kinds, from clickbait all the way through to full-blown undermining of democracy by a foreign state. There is exaggeration that appears. Because the industry is so connected there needs to be some sort of infrastructure whereby an advertiser and a publisher each know what they are talking about when they go through all the intermediaries and talk about where this advertisement could be placed.
Q128 Chair: Also, there is never going to be a static list because sites that are a source of fake news, as with sites with pirated content, are being set up and taken down all the time. You need a technological solution that identifies likely sources of problematic content and they are steered away from it. I would imagine that is something the industry could create as a code for itself rather than something that has to be defined in law.
Paul Farrelly: I am stuck back in the days of display advertisements in print, posters and television advertising where every agency took a fixed commission. Chair, I do not think you are that far ahead of me actually, and you used to work in advertising. Can I just ask some basic questions? First of all, before I forget it, what is “clickbait”?
Eitan Jankelewitz: Clickbait is a headline that entices you to press the link to see the advertisement. It is called clickbait because you are basically baiting the user into clicking the link because that publisher monetises on the basis of a cost per click; they will charge the advertiser every time someone clicks for that advertisement.
Q129 Paul Farrelly: Right. They charge the advertiser. How do they charge the advertiser, through what mechanisms or conduits?
Eitan Jankelewitz: Through advertisement networks and other intermediaries who deal with the payment of the commissions.
Q130 Paul Farrelly: Back to basics—Matt, please chip in—I have The Guardian here on my iPad through Safari, which uses Google as its search engine. Across the top of The Guardian there is a banner advertisement for a clothing company, which I have never heard of, called Gant. How does that advertisement come to be there?
Matt Rogerson: If you have come through Safari that would be because Gant has recognised the high quality of the audience that is reading The Guardian online and that you, as a high-quality audience member, would potentially want to buy a piece of Gant clothing.
Q131 Paul Farrelly: Right. Does The Guardian have any say in that?
Matt Rogerson: Yes. In terms of who advertises in The Guardian—as you will remember, Mr Farrelly—from the policy we have in place we have a view on who should and should not advertise in The Guardian, both in the new tabloid format and also on The Guardian’s websites and app. We have a blacklist of people we would not want to advertise in The Guardian.
Q132 Paul Farrelly: If my Safari were set to a different search engine, would that advertisement necessarily be there through a different search engine?
Matt Rogerson: Yes, because if that is a Google advertisement that will be served regardless of whether you got there via Bing or via another search engine.
Eitan Jankelewitz: If you went through a different web browser you would have a set of cookies there. If on a Monday you used one browser and on a Tuesday you used another browser, your browsing activity would be different on each day. You will be profiled differently and then you may get served different advertisements. If you clear your cookies, you will get the vanilla advertisement, basically, because at that point you do not have a profile.
Q133 Paul Farrelly: From the advertisers’ point of view, how do they get there and do they know they are there?
Bethan Crockett: In that scenario, yes. That is most likely a direct buy on The Guardian. Gant has said, “I would like The Guardian’s readership and potential audience. I would also like the way advertising was bought in the past.”
Q134 Paul Farrelly: So that works like a traditional display advertisement?
Bethan Crockett: Yes.
Q135 Paul Farrelly: How would a reputable clothing retailer’s advertisement come to appear, not on its own volition, on a scurrilous news site, a jihadist site or a porn site? How would that happen?
Bethan Crockett: Say, for example, Gant or a reputable clothing company had tracked all these visitors to its own website selling the products and those visitors had not purchased anything. It wants to show them advertisements again to entice them back to its product. It has cookies on its own website in the browser so they know that browser has visited its website. It will then share that information with an audience network, someone saying, “I can sell you audiences that match up.” It will share those cookies and that network would be looking for that cookie appearing, no matter what website the person is looking at. It is just seeing the same browser, the same cookie, that could be looking at any website but, “It is that cookie again that Gant told me about so I will show them the advertisement.” The user would have to go to the inappropriate website.
Q136 Paul Farrelly: Right. A different user would be seeing a different advertisement, so it is not static?
Bethan Crockett: Correct.
Q137 Paul Farrelly: Okay. The scurrilous site would get paid through advertising intermediaries for every time the advertisement was clicked?
Bethan Crockett: Not for the clicks, but it could be paid for just the advertisement being displayed to the audience. You do not have to interact with the advertisement.
Q138 Paul Farrelly: There are two ways, per click or just display?
Bethan Crockett: Just display, the traditional way of paying for your impressions.
Q139 Paul Farrelly: We asked for advertising to be included in the terms of reference here because we recognise the new ways of people being rewarded by advertisers for viewing or clicking gave an economic impetus, potentially, to people who are up to no good. How can the economic incentives, the click, be mitigated in trying to come up with a response to a phenomenon such as damaging fake news?
Bethan Crockett: Within display advertising there are very few advertisers paying on clicks; they pay on the impressions, the opportunity for the advertisement to be displayed. The way to disincentivise is to stop the advertising reaching there, which is what we have touched on, and all the measures that can be put in place to stop the advertisement appearing there. The only other element I can see where you could work on disincentivisation is a different topic, not a fake news element. Clickbait, as you have described, is fraud so that is a different topic. There are lots of people in our industry who try deliberately and intentionally to attract a consumer on to a website that they cover in as many advertisements as possible because then they are paid for every single advertisement that is displayed. That is fraud and there is different technology we use to detect fraud. We use that technology to stop advertising going to fraudulent sites but that is a different topic from stopping the advertisements going to fake news.
Stopping the advertisements going to fake news falls into the brand safety contextual controls that we have touched on. I think there are very few intentional fake news sites out there whose goal is to spread the fake news and not to be in an advertisement monetisation programme. However, they can plug into those programmes to do that, so that is where we need to put the brand safety controls in to stop them getting any revenue from that.
Matt Rogerson: I do not see the two as different. I do not see fraud as being different from the underlying reason why fake news is a profitable business for both entrepreneurs and foreign states. AppNexus, a very large advertisement technology company, released a white paper yesterday that said fraud is absolutely the issue that underpins fake news and is the reason why it exists.
If you create a more transparent and open market, that will weed out some of the worst practices. Some of the worst practices, unfortunately, are bundling premium inventory with very long-tail inventory that might look appealing in terms of price but is certainly not appealing in terms of where that money ends up.
Q140 Paul Farrelly: I have one final question, Chair, for Eitan. I am an advertiser and I find that my advertisement has appeared on a website that I would never have chosen it to be on. It has been picked up and people are talking about it, as they were with the jihadi website. What legal recourse would an advertiser have against a platform that facilitated that, or are the platforms in such a position where there will be contracts so they cannot be actioned?
Eitan Jankelewitz: For the savvier advertisers there would be much stricter and more specific contractual obligations on the partners they worked with to place that advertisement and they would go and look to see if those clauses were breached. For the less savvy and for more SMEs, they may find they have agreed that the advertisement can be placed anywhere and there may be wording that says, “If it is that is up to you.” It is a case of looking to the agreement, seeing what the obligations are and seeing if there was any breach of those obligations. Was the service delivered or was it not? If the service did not include checking for this stuff then it would not be a breach.
Q141 Paul Farrelly: Have any of those contracts been tested in law as to whether they are onerous or not?
Eitan Jankelewitz: I do not think so. Bear in mind everyone is having their own experience of the internet. You would not have a scenario where there is a huge volume of these kind of breaches, it may be one or two isolated things. In our experience it tends to be dealt with between the companies. There may be some sort of legal squaring off but it tends to settle quite quickly because it is usually quite a small incident.
Q142 Paul Farrelly: I am asking this in particular to learn whether there is a sturdy framework in place that makes the platforms themselves take more care rather than it being spelt out by the clients.
Eitan Jankelewitz: No, it is just market forces, if I can put it that way.
Q143 Chair: Building on Paul’s online profile as a Gant shopper—just as well it was not Gieves & Hawkes, because otherwise he would probably be in trouble with the Labour party—and going back to the issue about the platforms themselves, if someone has a profile as a Gant shopper and they are getting Gant advertising on The Guardian website, and they then go on to YouTube and Gant then wants to use the programmatic model to allow its advertising to follow the user around, what control does it have over where its advertising then appears on YouTube channels? Does that just follow the user and it could appear anywhere depending on what they happen to go and look at?
Bethan Crockett: Correct. There are some levels of control offered by YouTube. It offers about four layers of control where it describes its content as for mature audience, family, acceptable or any audience. You can choose your audience profile. It has a section where you can opt out of sensitive topics or sensitive categories. It has some elements such as sexual content and so on, there are about five of those so you can choose to opt out there. Then it has a list of topics, how it has categorised all the video content, “This is a topic about sports or football. This is a topic about politics. This is a topic about religion.” You can go down into those multiple layers as well and opt out of specific topics. After that you are at the hands of the content and the advertisement following the user.
Q144 Chair: Depending on what options you have selected, even using this programmatic model of advertising, the advertisement might not appear because the browser is looking at content that the brand considers to be unsafe.
Bethan Crockett: Yes. Let us say you have opted out of mature audience content and the user happens to go to some sort of mature, 18-plus horror movie trailer. That has obviously been blocked because it is considered for mature audiences and you have opted out so the advertisement would not follow the user on that particular video.
Q145 Chair: There is still a risk, depending on how good YouTube has been at classifying these different areas of content, that someone could end up seeing their advertisement placed against problematic content simply because of the way in which the user has moved from one platform to another.
Bethan Crockett: Correct. All those categorisations and opting out are done by YouTube and Google technology. As we have seen, there have been challenges with those opt outs picking up and removing content.
Q146 Chair: I think YouTube has changed its setting so that advertising does not appear against films that have been viewed only a small number of times, believing that those sorts tend to be the more problematic. However, it is a fairly crude tool to use, isn’t it?
Bethan Crockett: Yes. One option it chose was to remove monetisation from a channel that had less than 10,000 views. However, a channel can be made up of multiple videos so it is not even as granular as you might wish. A video that has had 10,000 views or less is considered you or me uploading something to share with our friends and family. That may not be what a target advertiser would want to show its advertisements next to as it is not a large audience. It is not necessarily a great indicator of quality.
Q147 Chair: How effective do you think YouTube is at policing its own platform and making sure videos are categorised correctly?
Bethan Crockett: We can see it is challenged based on all the news we have seen over the last year. It is looking now to try to find a better solution.
Q148 Chair: Is it largely reliant on the user generating the content correctly classifying the content rather than it checking itself?
Bethan Crockett: Yes. It allows the user to categorise and it is not able to check itself every single categorisation.
Q149 Chair: An advertiser who has decided to place certain restrictions on where its advertising goes on YouTube in advance is basically relying on the truthfulness of the person generating the content in the first place and that they have correctly categorised it?
Bethan Crockett: To a certain extent. That is one of the factors feeding into that categorisation model.
Q150 Chair: If you are the sort of person who was creating problematic content to distribute on YouTube, and you wanted it to reach the biggest possible audience, you would incorrectly categorise it to make sure that it did.
Bethan Crockett: If you were deliberately trying to get your fake news spread, why would you flag it to indicate that?
Q151 Chair: Yes. In effect, unless that content is referred to YouTube, it is highly unlikely it would spot that itself?
Bethan Crockett: To date, yes, I think we have seen that it has been difficult for it to spot that.
Matt Rogerson: Publishers are treated by completely different standards. We had a few clients in 2017 when the issue around brand safety hit where they did not want to appear next to any articles that included the words “worst”, “pathetic”, “Westminster”, “mice”, “law on” or “fire”. There was also a list of MPs names where if those articles came up they did not want to appear next to them—none of whom are on this panel.
Chair: I am glad to hear it.
Matt Rogerson: What you are seeing effectively is brands moving away from the idea of funding high-quality journalism, hard journalism and hard news. That is really problematic for us because they have always wanted to appear in the newspaper and in fact some of those same brands will happily appear in a newspaper next to a piece of journalism about an issue like terrorism. The problem is that the tools might spot a word in a Guardian article, which is “terrorism”, but there is a question whether or not that should be treated in the same way as a piece of random UGC on YouTube or Facebook that has the word “terrorism” in it. At the moment these tools risk squeezing out inventory we have next to our serious journalism, cutting the link between advertising and news. That is a serious problem.
Eitan Jankelewitz: One small comment on user-generated content, as platforms they depend on the defence that they are just the hosts of this content and are not responsible for it. They allow an environment where people can share things for themselves. If they exercise editorial control over content they host then they can be made liable. If things are IP infringing, defamatory, obscene or whatever they are the platforms are walking a tightrope. Facebook wants a friendly environment, they all want that. However, they also do not want to be in a position where they have exercised such control over the content that they have basically taken ownership of it.
If you are a business and you have a Facebook page for your business, the way advertising will apply to you depends on how much curation you put over it. In different areas of law this is a theme: are you the printer of the page on which the words are written or are you the newspaper creating the content which you then sell? I imagine that is some part of their thinking because they do not want to take on that huge liability. YouTube cannot possibly check every video.
Q152 Chair: Does the programmatic model work with Facebook in the same way as YouTube or are there differences?
Bethan Crockett: It is a similar way but the advertisement products are slightly different. You cannot programmatically follow someone into Facebook. If you are logged into you Facebook account looking at advertisements within your stream with other posts from your friends and family, that is not going to be programmatically served because it will not allow you to target its audiences within the product that way. It will offer its Facebook audience network, for example, which programmatically follow its audience when they go outside Facebook on to a series of websites or apps.
Q153 Chair: How would advertising follow? When someone arrives on Facebook are they then in a space that effectively is totally curated by Facebook and the advertising experience on Facebook is dictated by a direct commercial relationship people have with Facebook buying against the Facebook audience?
Bethan Crockett: Facebook directly, yes.
Q154 Chair: In terms of brand safety, what controls do you have over where that advertising appears? Does that entirely depend on where the user goes?
Bethan Crockett: No brand safety control is offered within Facebook.
Chair: At all?
Bethan Crockett: Within its audience network it offers some levels of opt out of certain topics and to manage certain categories.
Q155 Chair: As with YouTube, is the way it enforces those opt outs based entirely on the way the users have categorised the content they have created?
Bethan Crockett: More so their evaluation because it is not user-generated content so much. Its audience network is selling Facebook consumer and audiences on Facebook who have gone off onto a website and so on. It is not a video that has been uploaded or user generated, it could be any normal website page.
Q156 Chair: It could be a Facebook page, which is user-generated content, which could include films and texts?
Bethan Crockett: It will not be Facebook-owned content; it will be all different types of websites and apps that have opted in to receive advertising from Facebook selling their site on the bar.
Q157 Chair: Finally from me, Matt, you said there is this imbalance in the relationship between Facebook and other sites. You share with Facebook information about visitors to The Guardian website but it does not share it back. Why do you not refuse to give that data to Facebook? What would be the consequences of that?
Matt Rogerson: It is an interesting question. I honestly think the relationship between publishers and Facebook as a platform is changing. You saw last week that it announced a big algorithm change that will drive more content from friends and family and less ability for publishers to be seen by users. From a Guardian perspective, we are not particularly reliant on Facebook, less so over recent years. Other publishers are very reliant on Facebook so they need to share their data much more and rely on Facebook advertising much more.
I think a number of high-quality publishers are questioning the value of that relationship with the platform. As I mentioned earlier, we exited from an initiative called “Instant Articles” because we did not see it was either paying us decent amounts of revenue or allowing the sort of transparency we would require from a partner. Certainly that is an ongoing debate.
Q158 Chair: The Facebook share price took quite a big hit when it made those announcements of the change to the news feed. Although it is a closed market that probably has more information on its users than any other platform in the world, it sounds like an environment that is increasingly less attractive for advertisers to invest in. However, is it so big that you cannot avoid being there?
Matt Rogerson: There are two things. One is that it seems the advertising base at the moment, as I say, is predominately SMEs. The issues around safety are not as pertinent for a SME as they are for a blue chip company. It has peaked in terms of how much money it can get out of the display advertising market and the advertisement load on the platform seems also to have peaked. The question is where you generate more revenue from. I think it is moving more and more into trying to develop high-quality video that will enable it to get some advertising revenue from people who would otherwise push it towards TV. The nature of the platform is probably changing and the algorithm changes are a reflection of that.
Q159 Chair: That was very interesting. Thank you very much.
Witnesses: Tim Elkington, Phil Smith and Ben Williams.
Q160 Chair: Thank you for joining us for the start of this second panel. Thank you, in particular, to Tim Elkington for stepping into the breach at the last minute for Jon Mew, who had been advertised as being on the panel today but was unable to join us because of a bout of winter flu.
Gentlemen, you will have heard the previous session. To start off with we would be interested in your overview of the advertisement market at the moment and the ability of the internet advertising market to respond to the growing challenges over brand safety that major advertisers, such as Procter & Gamble, have raised in the course of last year that will be of interest to the whole industry. Perhaps, Phil Smith, you could start off from the advertiser’s point of view. How much progress do you think is being made and are they doing enough?
Phil Smith: Thank you. It is a very big question. It is best to start by trying to say there is no one response across the industry. The responses we have seen from the different players have been very different, in terms of both style and content. We have been engaged quite deeply with Google in particular since last year, especially around the stories that came out in The Times in March. We have seen a succession of initiatives, the most recent of which were heralded in a blog that came out from YouTube in December.
Clearly there is an increasing amount of activity that is being put behind the issue of brand safety and also where advertising is appearing as an appropriate context. Google, in particular, has already committed to having 10,000 people on the ground by the end of 2018 addressing the issue of content control and putting more tight rings around the content it deems to be fit for advertising on YouTube. While that is clearly a big step forward, the difficulty from an advertiser’s perspective is knowing when enough is enough because there is no independence or openness around the oversight although we know there is huge activity. In 2016 Google took down 1.6 billion advertisements. We do not really know the basis on which all of those advertisements were taken down, and there is no sense of an independent view that says that cured 95% of the problem—90% or 20%. None the less, there is a lot of work going on there.
With Facebook it is more difficult to understand what is going on behind the scenes. Again, we had an announcement that there would be a significant step up in terms of its efforts in this area and 20,000 people were promised on the ground by the end of 2018. This time last year we were being told that an extra 3,000 on top of the existing 4,000 would answer the question, so the sense is that there is a catch up being played here. I think most advertisers and publishers were taken by surprise by the announcements that came out at the end of last week from Mark Zuckerberg, and indeed a lot of Facebook staff themselves were taken by surprise. We are still waiting to see much more detail around what is there. Again, there is that same issue of lack of independence and oversight is one that is very much on advertisers’ minds.
Q161 Chair: On that point about the people you are referring to, the human checkers who are going to be hired by Facebook to help oversee content and deal more effectively with user referral of problematic content, as politicians we all know it is easy to announce great new initiatives and determination to invest in certain things, but delivering on them is a very different thing. From the ISBA’s point of view, are you reassured that Facebook is actually implementing what it says it is going to do, these things are being set up correctly and are operational?
Phil Smith: I do not think we can feel reassured because there is that lack of oversight. We are very reliant on what we are being told. There is more reporting that we see that comes out of Google, if we are using that as a counterpoint. It does have a transparency report and it has a bad advertisement report. Whilst it is still marking its own homework there is a degree of granularity in what we see. It is harder to get under the skin of what is going on at Facebook. We have not had that direct contact with Twitter so we do not know what the state of play is there. One of the concerns we have is that as this issue becomes larger, and as people put more resources behind fixing it, it is only really the large players who are going to be able to continue to invest in a satisfactory way to be able to attack the issues successfully and the smaller players will get even more left behind and disadvantaged.
If I take the contrast of what we see in advertising regulation in the UK, where the ASA for a number of years—funded by the advertising industry voluntarily, so advertisers fund this as a voluntary levy on the back of their advertising spending—is responsible for all commercial content online and in the offline space where there is independence, governance and an oversight in the reporting. That allows us to feel confident that what it is doing has a degree of impartiality and evidence base, but we do not feel that same degree of confidence in the online world. When we do research with the public as well, it is clear that even with people who are fuzzy about what form regulation takes for TV, press and other media, they know that somebody is in charge but they might say Ofcom or the ASA. People do describe the digital environment as a wild west and do feel that nobody is accountable.
Q162 Chair: I know there are other members who want to come in, but just on that opening question I would welcome other people’s comments. Certainly, Tim, the challenge from Marc Pritchard in America at Procter & Gamble is that technology is always innovating but this is a mature market and the biggest by value advertising market in world. Do you feel the major internet players are able to step up and give brands the assurances they need over brand security advertising online?
Tim Elkington: We are the not-for-profit trade body for digital advertising in the UK. One of the things we launched last October was a new gold standard for digital advertising. We have 23 board members and they are all committed to implementing that. Of interest, of 23 board members six are national news brands such as The Guardian, The Daily Paragraph, News UK and so on. That initiative covers three things: advertisement fraud, brand safety and providing a positive user experience, in terms of advertisements so the advertising is not too distracting and too annoying.
The events of last year have really galvanised the digital advertising industry to get together and do something about it. We announced that last October and people started registering for it just before Christmas. The certification phase, where we will check people are actually doing what they have said they are going to do, will start next week. Big players—such as Facebook, Google and so on—have all committed to be part of that.
Q163 Chair: It is a big commissioning part of it. However, again, do you feel there is an issue that Google and Facebook are very different from the rest of the market in that they do not offer the same level of transparency about the way they target their users?
Tim Elkington: As part of the gold standard initiative both of those players will have to sign up to the DTSG, the Display Trading Standards Group. The brand safety best practice principles there are governed by JICWEBS, the Joint Industry Committee for Web Standards. Maybe that was the case in the past but I think everyone within the industry is realising how serious it is. Phil obviously talks on behalf of the advertisers. If it is a concern for advertisers then, by definition, it is going to be a concern for media owners.
Q164 Chair: We have just heard from the previous panel that Google and Facebook could sign up to that but cannot guarantee it because they are largely reliant on the user-generated content on their platforms correctly categorising what it is. All they can really do is react after a mistake has been made and not guarantee that the space will be safe.
Tim Elkington: It is impossible to ever guarantee 100% brand safety. However, stuff like the DTSG will enable advertisers to make really clear choices and to say, “This company is a member of the DTSG, and therefore we are content they have done the right things. We are going to have a contract with them and there will be a redress if stuff goes wrong.” If a company is not a member of the DTSG then obviously it can be a risk in terms of brand safety.
Our job as an industry is not to make the choices on behalf of advertisers but to give them all the tools and the knowledge so they can make that choice, and if they want to take a risk they can. If they want to be as sure as they possibly can be then they can make sure that the platform they are advertising with is a member of the DTSG, has signed up to the gold standard and does that other industry stuff.
Q165 Chair: The point I make is that YouTube might have signed up to the gold standard but it probably cannot offer the same guarantees another company can offer. That is because people do not have access to the data and its system of even enforcing the preferences the advertisers have selected is largely based on the voluntary description of user-generated content on YouTube and categorising it in the correct way. YouTube at the moment cannot offer a failsafe way of protecting brands from their advertising appearing against content that has been misclassified and could be difficult.
Tim Elkington: I can never remember all the figures but it is something like 400 hours of content uploaded every minute. Obviously in terms of scale it has a very different challenge from a premium publisher like The Guardian that manages and writes all of its own content. You are right; it is a different challenge.
Q166 Chair: Ben, perhaps we can bring you in on this as well before I bring in other members of the panel. You have heard what we have been discussing. Do you think there are more effective tools in place that can help protect brand safety and also make sure users see the content they want to see?
Ben Williams: Thank you. We are a company that makes products that allow people to block advertisements, block tracking and protect themselves while they are online. We therefore come at it with a more user-centric point of view.
It is helpful to frame the larger problem because it is not just Google and it is not just Facebook. It is really that the entire internet infrastructure, as Phil pointed out, is often referred to as the wild west. It is not that fake news or inappropriate content is anything new; it is just that the openness and the opacity of the internet architecture, in particular in advertising, allows for players who would like to spread that information to metastasise it and to grow it in ways that was not seen before. It is helpful to look at industry solutions for doing that.
It is also helpful to look at platform solutions, so what Facebook and Google have done up to this point. Both have made very positive strides forward. The downside to the platform solution, of course, is that it will be limited. Facebook and Google, because of the lion share of advertisement spin that they take in, are often referred to in the advertising world as the duopoly. You could cover the duopoly—probably could not cover it 100%—but you would not be covering the entire internet. Another downside to that is that they are just as open, accessible and opaque as the internet infrastructure itself, as the greater internet advertising world. It would be very difficult for them to police even their own systems, as we have been talking about.
Finally, it would concentrate a lot of power in the platforms themselves. Both publishers and users are feeling a strain on the limitations that Facebook places upon their content conception or, from the publisher’s point of view, the content they are able to share with the public. This concentrates more power in very few people.
There is another solution that is, of course, where we have a bit of expertise and that is to entrust that to third parties. The good thing about a third party who may develop software that would alert users to questionable content, to fake news and even to satire is that it would be independent and the motivation would be strictly to the user. Finally, it would work across the entire web. It works across Facebook, across Google and across any site that a user may stumble upon.
That is how we see the entire problem and, of course, we look at it more from a user point of view.
Q167 Ian C. Lucas: Can I ask a little bit more about the regulatory framework involving the Advertising Standards Authority? We all know and recognise the legal test of honest, truthful and all that. As I see it, it is very much a national organisation. One of the aspects that is novel, in a sense, with the internet is its international scale. As far as internationally based or placed advertising is concerned, how does the ASA exercise authority in those areas? If an advertisement was placed from a foreign jurisdiction, how could it exercise any authority on the issue?
Phil Smith: The truth is that it would find it difficult. Any advertising that is served to somebody in the UK is capable of engendering a complaint, but the ASA’s ability to pursue that if the website hosting that is offshore is going to be problematic.
Q168 Ian C. Lucas: What about if it was on Facebook or on YouTube? Would it then be able to require Facebook or YouTube to do anything?
Phil Smith: It would pursue the complaint in the first instance with the advertiser but it would be able to get it taken down.
Q169 Ian C. Lucas: In a sense the regulatory buck has been passed to the platform in that case. It has signed up so presumably that means it is responsible.
Phil Smith: The broader question of responsibility is difficult because you get into this debate of legal liability between platform and publisher. That is something where I would not feel qualified to make a comment. When it comes to the responsibility, what we are seeing is that there is a de facto recognition that the platforms are taking responsibility for the content they are hosting by the very fact they are putting in place these very large resources on content control and exercising their own community guidelines on their own individual platforms.
Q170 Ian C. Lucas: There would not be any sense in them signing up to the regulatory framework unless they were prepared to act in the way you have described and take an advertisement down.
Phil Smith: I am not sure those are the words I used, but the ASA would be able to pursue with Facebook and Google to have that advertisement taken down, to the best of my knowledge.
Q171 Ian C. Lucas: Facebook and YouTube have agreed to that?
Phil Smith: I would need to give you a response on that separately.
Q172 Paul Farrelly: I wondered whether you could bring me out of my time warp with another basic question. I have just searched for the profile of a prolific fake news generator in the US who, in this profile, said some of his contributors all had Google AdSense accounts. I have looked at Google AdSense and how it works—advertisers bid for space in online auctions. How does that process work?
Phil Smith: Shall I have a first go? Given you have just had Bethan here I think our answers are going to be much less granular, or at least mine will be. When a user visits a page and that page is loading, an auction takes place in real time for that user profile based upon information from his cookie about his previous visiting habits, his purchasing and anything else that might be associated with that cookie. Based on a number of criteria, including the size of the bid, the advertisement will be served to that individual. If you go to a site and then click refresh, you will find each time you will be served with a different set of advertisements because that auction is taking place repeatedly each time that page is being served.
Q173 Paul Farrelly: In a millisecond?
Phil Smith: Yes, it is. Therein lies one of the issues we have been trying to address. The promise of programmatic advertising for big advertisers was that it would allow them to reach very large audiences, or very targeted audiences, very quickly and economically and would do so without wastage. That was particularly attractive for big advertisers who had seen their audience base fragmenting on traditional media. As TV audiences have become more fragmented it has been much more difficult to build schedules very quickly to get to large numbers of people. Programmatic looked as if it was going to be a solution for that.
Unfortunately, within programmatic advertising there are a series of perverse incentives and a very long and rather convoluted value chain that incentivised players along the way to buy and sell cheap, low-quality eyeballs in order to maximise the margin opportunity between buyer and seller. Put simply, if I know you are visiting a very low-quality site and I know you are visiting Guardian.com, I would rather pay to get you in the low-quality environment as it is going to be cheaper because I am buying you as an eyeball. That is the way programmatic currently works today.
We believe that some of those incentives occur on the buy side as well as on the eye side. That is one of the reasons why in the work we have been doing through contractual arrangements with our media services framework with the agency community we have sought to realign agency incentives to more closely align with the business needs of their clients. That has been to look for terms that are more transparent in terms of the ability to see rebates that are taking place within the value chain, to look at the treatment of inventory that is bought and then resold by an agency, for example, and to look at the way in which other forms of money are flowing through agency networks all the way through to the ultimate seller.
Alongside that, also in the contract frameworks that we have now been promulgating in the industry for coming on two years, we have been looking at standards of brand safety. The DTSG framework from JICWEBS, which we talked about, would be built in as one of the framework conditions, along with conditions around advertisement fraud and ownership of data. It is one of the things we have been advocating very strongly to our membership and more broadly to the advertising base to ensure that their terms are well aligned with those of their agencies and there are not built in conflicts of interest or perverse incentives.
Q174 Chair: So that I am clear in my own mind before we move on, would I be right in saying that for an advertiser they probably have an agency who represents them. The agency and the advertiser come to an agreement where they say, “We want to reach this certain audience” and they agree an amount of money that it will cost to reach that audience. The agency then goes into the market, to the auctions, and buys those slots. However, the agency has an incentive to buy the cheapest amount of stuff that will reach the same audience because it makes a bigger margin on it. Is that correct?
Phil Smith: Yes, and in some cases the agency has set itself up as a principal where it has bought inventory that it is then seeking to resell to the client. That, in itself, creates what we can call a misalignment of interest. It is not necessarily illegal but you are buying from a vendor rather than from an agency when you are in that relationship.
Q175 Chair: It could create a perverse incentive for advertising to end up on lower quality sites because there is more margin to be made?
Phil Smith: Correct, and that is one area we have been working on very hard.
Q176 Paul Farrelly: To be clear, the operator of the auction is Google in this instance?
Phil Smith: It could be anybody. There are a number of people through the value chain where there are more or fewer choices. We could certainly supply the Committee with a chart that shows the various steps the money goes through.
Paul Farrelly: That would be really helpful.
Phil Smith: It goes through the agency records, sometimes through a trading desk, through a platform on the demand side, there will be some form of data analysis to help them with the targeting. Then you get on to the supply side where, again, there will be a platform and exchanges involved as well before it gets to the final publisher.
Q177 Christian Matheson: And everyone takes their cut.
Phil Smith: And people take their cut. That is why you hear the stories of people saying that 30 or 40 pence in the pound is what ends up with the publisher.
Paul Farrelly: That will be very useful. If you have not guessed already, we are learning as we are going along.
Phil Smith: I thought that was an area of interest that you identified.
Tim Elkington: Phil is exactly right. The best way to think of it is as a sort of eBay auction that happens in real time. The advertisement is up there. There might be multiple advertisers who are interested. The advertiser who is prepared to bid the most to have their advertisement there is going to be the one that wins the auction.
It is worth putting that in the context of the whole market. We measure the market with PricewaterhouseCoopers and the latest figures we have are for the first half of 2017. The whole digital advertising market in the UK is worth £5.6 billion. Search is roughly half of that and all of the search advertising occurs on that real-time auction basis. Classified is about 12%, which is listings for cars, jobs, holidays and so on. About 36% of the market, £2 billion, is display.
Within that display area it is worth thinking about how it is traded. About 25% will be traded in that very traditional way. I, as an advertiser, might phone you as The Guardian and say we would like to advertise and it is entirely done in the traditional way. Another 50% is done directly. Again, I am phoning you up as the advertiser and we are making an arrangement but I am executing it in an automated programmatic fashion. As an advertiser I know exactly where my advertising is going but the bid will vary depending on the quality of the person.
The remaining 24% goes on that open exchange, which is about £0.5 billion. Less than 10% of the UK digital advertising market is in that area where we are saying the advertiser might not know where their advertisement ends up because they are buying an audience rather than dealing with a particular publication. When we talk about the bidding and so on it is useful putting it in context because for the majority of the market there is still that relationship between either the agency or the advertiser and the media owner in particular.
Really quickly, one other thing I wanted to come back on is when we talk about the motivation of the agency to buy the audience may be in a lower cost environment that might be lower quality, you have to bear in mind that not all impressions are created equally. If I am a very loyal Guardian reader and I am very committed to the editorial, if I see an advertisement there I might respond much more positively to that advertisement than if I see it on a clickbait site, as was talked about earlier. In some research we did with YouGov, we found that 75% of people would not trust advertisements on a fake news site and only 9% would. Phil is right that there are different movers in the market but it is worth bearing in mind that if you are assessing your advertising on what works and what the return on investment is, then quality should see out. It is not just about getting lots and lots of reach.
Phil Smith: I would say that of course is true. Programmatic is the area that has been growing the fastest. It took me five minutes on Saturday to find two of my members’ advertisements served to me on sites where they would not be very happy. Those sites may not have been Times headline-worthy, but they were certainly in the clickbait, fake news arena.
Q178 Paul Farrelly: To be clear, this lightning fast auction is happening while I am loading a page?
Phil Smith: Yes.
Q179 Paul Farrelly: The auction that is going on is for me going to that page and not to fill an advertisement that is static on that page?
Tim Elkington: It is for the opportunity to show you an advertisement. If you had been at the Ford website you might be characterised as someone who is very interested in buying a new car and because of that automotive advertisers are obviously going to put a premium on the opportunity to show you an advertisement.
Q180 Paul Farrelly: The incentive is for the people running those sites to run them with content that attracts people like me. Are there any safeguards against a reputable advertisement following me to a disreputable site, the whitelisting and blacklisting we have talked about in the previous session?
Tim Elkington: If, for example, you were with GroupM and dealing with Bethan’s team then all of the stuff she talked about would be put in place to make sure that that advertisement did not follow a relevant person on to a site where you did not want it displayed.
Q181 Paul Farrelly: That begs the question—the last panel was obviously too short—that because the platforms are there all the time and this is their business, to what extent are they helpful in maintaining and updating banned lists, as it were, that they would advise you not to visit or is the onus pretty much on the advertiser and its agencies?
Phil Smith: Again, you would need to split each of the individual players apart and look at the environments in which the advertisements are being served. The closed system, such as Facebook, is a very different environment from YouTube and from the open advertisement exchanges. There are different degrees of risk associated with each. I would say from what I have seen that Google has stepped up a great deal when it comes to the ability to classify content. Although I heard Bethan say it cannot be 100%, there is more granularity in the control settings now than there was a year ago. Whereas last March we were seeing examples of content being misclassified, we are not seeing that to anything like the same extent as we were then.
Ben Williams: What Facebook is doing is a bit different. At first it started to flag articles as disputed but it found a number of problems with that. The biggest problem was that occasionally that would cause it to become even more viral. It would cause preconceived notions of people who may view the advertisement to kick in and then to think, “Oh, this is just the fake news media that this is telling me this is not real. This is actually real because it is coming from someone I trust, someone within my network.” That is really getting to the heart of the matter of how much exactly Facebook can do. What it does now is it will give you other articles that would dispute that and other fact checkers that might dispute how truthful that article may be.
What we are talking about is, on the one hand, brand safety. We have to understand there is a whole other side of people being misled by content that has gone viral and that happens in a very different way. We have all probably read articles about places in the world where there is a cottage industry that has sprung up to make fake news. In fact, everyone read the article about a couple of villages in Macedonia. That was very instructive because this teenager was able to set up a series of websites that have URLs that are made to look like a legitimate URL but is just slightly different. Because he has access to the levers to buy those eyeballs, to buy those Facebook users, to click on his articles he can make content. If he is really good at what he does then he can make that content go viral and then the damage spreads very quickly. The disputed tag is not really going to help anyone because, first, people may not believe it and, secondly, it is too late.
Phil Smith: Yes, and sensation and outrage travel very fast on the web, as we have seen. In my view there is an inherent issue with user-generated content because it will always push at those margins, if not beyond them, and into activities some of which are borderline legal and some which are plain illegal and you are not far into fraud.
Q182 Rebecca Pow: Thank you. Huge apologies for arriving late. I am much later than I thought I was going to be so you have probably have covered this but just on Facebook, gentlemen, it has said it has already started prioritising friends and family and is going to stop these viral videos. Is that possible?
Phil Smith: I think they said almost the reverse, haven’t they? They will prioritise that which is content—
Rebecca Pow: Fewer viral videos.
Phil Smith: —that has lots of comments on from friends and family. I have seen some commentators say they think that will not do anything to address the issue of fake news because fake news that is spreading virally will continue to spread. What they have said, as I understand it—and we do not know much by way of detail—is that they will put less priority on material that is coming from brands and from publishers. Therefore, much of the commercial content in the news feed will be deprioritised. Again, we do not know much by way of detail, unless any of you have any more on that.
Tim Elkington: No, theoretically Facebook can look at the preference it gives to different content and change that so it displays more from friends and family and less from commercial messages. If you wanted more detail on that, Facebook would be the people to speak to.
Q183 Rebecca Pow: Of course; it was just that it sparked my attention. From the advertising industry’s point of view, do you think you have a responsibility to ensure that advertisements go to the right places? Is that something that is a major focus?
Phil Smith: From an advertiser point of view, we have every interest in making sure people feel confident in the environments where they see advertising and that credibility and trust is built rather than destroyed. We already have oversight over the advertisements themselves. We have talked about the ASA where more and more of its work is being focused on the online environment. What we have seen in the research from the public is that there is a sense that while they feel in the traditional media there is somebody in charge they do not feel that same degree of confidence in the online world. As Tim has said, people feel more confident in brands that are advertising in trustworthy places. We certainly have that motive. We advocate our members should take as much responsibility as they can in the relationships they have with their suppliers, whether that is an advertising agency or directly with any media vendor.
Q184 Rebecca Pow: You talked just now about these auctions. These agencies are buying them on low-quality sites because they can get more for their money. Are the lower quality sites the sites where you are going to get the more dubious news? How are you in control of that?
Tim Elkington: I would bring it back to the fact that as an organisation it is our responsibility to educate the market, to say, “Look, you do have a choice. You can advertise on low-quality sites or you can advertise around premium content.”
Q185 Rebecca Pow: What is a low-quality site? Give me a couple of examples.
Tim Elkington: It is very difficult off the top of my head, but it would be that sort of fake news and clickbait sites we have been talking about. High quality would be professionally produced with quality journalism, like The Guardian, The Daily Telegraph and so on. Our job is to illustrate the benefits of supporting quality journalism and content, not because it is the right thing to do but because it is better for your brand and because your advertising is going to get a better result if you are displayed within that kind of high-quality environment.
Our job is to do research and to put on events to get that message across to advertisers that to some extent you get what you pay for. If you are paying for lots of reach on low-quality sites you might get lots of eyeballs, but it might not be effective, whereas it is more likely to be effective if you are using the relationship that The Guardian, for example, already has with its readers.
Q186 Rebecca Pow: We have had discussions on previous panels that took us into the world of journalism and how you trust something. Do you think the advertising industry is now realising it needs to work with what you call bona fide investigative journalists in order to overcome this fake news issue?
Phil Smith: I think they would not necessarily see it in the context of journalism, but they are looking for high-quality environments for their brands. Different brands have a different appetite for risk, I think it would be fair to say. We have seen more and more of our members take their own measures when it comes to not just embracing the DTSG principles in the contracts they have with their agencies, but extensively using whitelists and blacklists—they have very narrow whitelists if they are from financial services, for example. Some are avoiding the environment altogether because of what they see as a risk to their brands.
There are other brands that are advertising out there—and we talked about those references to SMEs earlier—and there are some businesses that have built their business on the basis of response, so they are paying for a conversion or a click, who frankly show little interest in where the ad appears as long as the economics of their business model are met by the cost per conversion. What you will find if you go to one of these lifestyle sites, “10 wonder foods that will make you lose hundreds of pounds of weight,” is that there would be some brands, some which you would not have thought were that risk averse, who will consistently keep appearing because they do not mind how many times they appear because they are only going to pay on a per-click basis.
Tim Elkington: Yes, you are right that different brands will have different attitudes to risk. Some do not mind where they are showing because that fits in with their brand values, whereas others are going to be much more conscious of that.
Phil Smith: We represent about half of the industry and we have members from across the sector—finance, retail, package goods and automotive—but we represent responsible advertisers. They try very hard to avoid being there. As I say, there is still an issue when it takes five minutes to find a couple of our members in unsavoury places.
Q187 Chair: Paul Farrelly has a question, but just before that I wanted to ask a quick one of my own as well. One of the things that Procter & Gamble called for was greater transparency on margin in the auctions for this advertising space. Are you satisfied that the industry has moved to deliver that, or do you feel that is an area where more work needs to be done?
Phil Smith: We think that there is more work that needs to be done, and we are constructively engaged with the agency community to make that happen, at least on the buy-side of the equation. There are some agencies and suppliers who have built their businesses on the basis of a fully-disclosed model, and there are others who offer a choice. Certainly in the work that we have been doing with our media services framework, which is going to be updated very shortly, we have worked closely not just with those agencies that have transparent models, but with the big holding companies as well who traditionally have made quite a lot of margin in that environment, we believe.
We think that we will see some progress, and we are seeing advertisers drive it too. Advertisers who have gone into habitual renegotiations are saying, “These are the terms on which we expect business to be made,” and in some cases have completely realigned their incentive structures for that to be the case.
Q188 Paul Farrelly: This is a curiosity question for me, but also it is relevant to transparency. Over 25 years ago I advised a company called Media Audits on its sale and management buy-out at the end of the day. Advertising agencies hated that company, because these were the days when they took a fixed commission, so there was no incentive to keep costs down. I do not know in the online world what level of audit there is, or the capability for it is, in judging between the effectiveness of different online campaigns. Obviously if your sales do not increase by one cent if you use Facebook then it is ineffective, but judging what sort of advertising campaign, into which outlets—The Guardian rather than disreputable sites—is something that an advertiser would definitely benefit from. What capability is there out there?
Phil Smith: I am sure Tim will add to this. There are pools, but it is a much more complex environment than it was, and more and more people are focusing much more on outcomes and deliverables from the advertising and its effectiveness rather than how it did with a market place reference?”
Tim Elkington: Yes, absolutely. That is what the thing is: it is a relatively complex world. Imagine that you are one user but you are using multiple devices and you have multiple different browsers. You might search for a holiday on your phone, do a bit more searching on your laptop, book it on your tablet, or you may make a purchase offline. Tying all of those things together to measure effectiveness is really difficult.
Phil is right. Again our job as an industry body is to encourage people to measure on business outcomes rather than just effectiveness of buying, or reach, or whatever. If you are able to do that effectively—and it is relatively complicated—you are able to say, “All of those people in the middle that are taking out 1p or 5p or whatever it is, if I take them away and it works just as well then obviously they were not adding much value.” However, the people who you are spending a bit of extra money on who make sure your ad is appearing in the right place or who make sure the audience is really relevant, it is about understanding the kind of value that they add. The truth is that it is an issue that the industry wrestles with constantly: what works, what does not work, how can we optimise against it, and how can we measure?
Phil Smith: I think if I take that reference book to Media Audits and the TV marketplace, the TV marketplace was relatively straightforward; we knew how much was spent on television, we knew how many ratings were being achieved, and a lot of the rest was around division. One of the challenges that we face in the digital world is that there is no single standard of audience management and very little comparability. Again, it is something where we have been working with other parts of the industry, with the IPA, our industry trade body counterpart, and with the platforms themselves to look to create much more independent and open standards of audience measurement, which would at least give you the part of the equation that allows you to say which audience was really reached.
Tim Elkington: It is worth saying that we do have something called UKOM, which is the UK Online Measurement Company, which is co-owned by the IAB and AOP, and both IPA and ISBA sit on. It does a good job of making quite a complicated picture understandable. It is worth adding that.
Phil Smith: It does not deliver down to commercial audience impacts, for example, so, “How many people saw my ad?”
Tim Elkington: it is more the size of the website and the demographics of the people visiting it.
Chair: Thank you. Are there any further questions? In which case, again that was another really interesting and insightful session. Thank you very much for your evidence.