もくじ ▶ ポイント１．日本のアドフラウドは世界最低レベル ▶ ポイント２．日本はブランドリスクでも最も高い増加率を記録 ▶ ポイント３．ビューアビリティも前年に引き続き世界最低レベル ▶ アドベリ対策が遅れる日本は世界から狙われている！ ▶ 《参考》これまでのメディアクオリティ レポート ▶ 《参考》IASの調査レポート
１．アドフラウド：世界的なアドフラウドの低下傾向にも関わらず、日本のアドフラウド率は増加し世界最低レベルに2020年下半期、世界のアドフラウド率は全体的に改善し、アドフラウド対策を実施したインプレッションでは計測対象のすべての端末、環境と広告フォーマットでアドフラウド率が低下しました。この結果、すべてのフォーマットでのアドフラウドの平均率は1.0%以下となり、日本版レポートの公開を開始した2018年以降で最も低い水準でした。 国別のデータでは、日本とオーストラリアだけがアドフラウド率の上昇を示しました。日本では特にデスクトップとモバイル端末のウェブ（ブラウザ）環境で、ディスプレイ広告の対策済みインプレッションのアドフラウドが増加しました。日本のデスクトップ ディスプレイでのアドフラウド率は前年同期の2.6%からさらに悪化し2.9%となり、世界ワースト1を記録しました。同期間にグローバル平均では0.3ポイントの改善が見られたのと対照的です。さらにモバイルウェブ ディスプレイでも2.7%（前年同期1.9%から0.8ポイント上昇）と世界で最も高いアドフラウド率でした。グローバル平均は0.7%（2019年下半期）から0.4%（2020年下半期）と改善しており、アドフラウド対策が世界的に進む中で日本が取り残され、標的とされていることがうかがえます。
２．ブランドセーフティ：世界の中でもブランドリスクの上昇が目立った日本。デスクトップ ディスプレイとモバイルウェブ ディスプレイでは日本が最も高い増加率を記録日本はモバイルウェブ ディスプレイのブランドリスクが2019年下半期の7.6%から2.8ポイントも上昇し、2020年下半期は10.4%と、世界で最も高いブランドリスクの増加を記録しました。2020年下半期のグローバル平均値は5.8%で、前年同期比でわずか0.1ポイント上昇しただけでした。日本はデスクトップ ディスプレイでも2019年下半期の3.2%から2020年下半期は5.6%と大幅な上昇を示し、これはイギリスに次いで2番目に高いリスク値でした。同環境のグローバル平均は4.4%でした。 2020年に発生した予測不可能な世界的パンデミックがデジタルエコシステムに与えた影響の一つが、動画広告におけるブランドリスクの世界的な上昇です。この傾向には、下半期を通して見られた動画インプレッション全体の増加と相関が見られました。つまり、パンデミックによって自宅で過ごす時間が増えたことで動画の視聴時間が伸びたことと、動画広告におけるブランドリスクの増加の間には相関関係が見られたのです。ブランドリスク増加の最大の要因はアダルトコンテンツで、ヘイトスピーチがこれに続きました。
３．ビューアビリティとタイムインビュー：日本は前年同期に引き続き世界最低レベル日本の2020年下半期のビューアビリティは55.7%（デスクトップ ディスプレイにおいて）で、グローバル平均の68.8%を大きく下回りました。2019年下半期に日本数値の公開を開始して以来、初めて5割を超えたものの、2半期連続で下落しています。ビューアビリティが7割を超える国も多い中、日本の55.7%は圧倒的に低い水準にあります。 モバイルウェブ ディスプレイでは日本の数値は43.5%（2019年下半期）から46.4%（2020年下半期）へと改善が見られましたが、グローバル平均の62.9%と比較すると大幅に低い水準にとどまっています。50%を下回ったのは日本のみで、デスクトップ、モバイルともに日本のビューアビリティの低さが際立つ結果となりました。 日本のデスクトップ ディスプレイのタイムインビューは23.53秒で、前年同期の21.95秒から若干伸びました。モバイルウェブ ディスプレイは2019年下半期の16.6秒から14.72秒と、逆に短くなっています。デスクトップでもモバイルでウェブでも、日本のタイムインビューはグローバルの平均値とほぼ同程度です。
アドベリ対策が遅れる日本は世界から狙われている！ここまで見てきたように、日本はどのメディア品質指標も世界と比較して低い水準にあります。各指標とも2018年下半期に一度改善の兆しが見えたものの、その後ふたたび悪化。同期間に世界各国がアドベリ対策を推し進めてメディア品質を改善させたのに対し、日本はアドフラウド、ブランドリスクともに明らかな悪化傾向が続いています。 時系列での推移、グローバルとの比較については別途詳しく解説する予定です。 日本は世界で最も広告投資額の大きい国の一つです。にもかかわらずアドベリフィケーション対策の導入はアメリカやヨーロッパ諸国と比較すると圧倒的に低い水準にとどまっています。これは、特にアドフラウドによって金儲けをしようとする不正業者にとって「狙ってください」と言っているようなものです。 ぜひMQRをダウンロードして、まずは日本の現状をご自身の目でご確認ください。
- 『The Power of Context in APAC：コンテキストが広告受容性に与える影響』（2021年3月4日）
- 『Industry Pulse 2020 日本版』（2021年業界トレンド予測）（2021年1月19日）
- 『The Ripple Effect : 波紋効果～コンテンツの品質が消費者の広告認知に与える影響に関する調査レポート』（2019年11月13日）
UK | The Drum: It’s time to stop incentivising digital ad fraud
Video Ad News: Digital Ad Fraud: is the Industry Stuck in a Never-ending Game of Whack-A-Mole?
The media love a headline, and this summer’s big shocker “60 Percent of Mercedes Ads Served to Bots” which appeared in the sober Financial Times, was swiftly followed by further headlines of wasted advertising revenue. Each week, with ever-larger figures bandied around, the industry seems hell-bent on giving marketers more misgivings about digital media.
The reality is that the online advertising industry — in striving to innovate and embrace technology — has also created opportunities for fraudsters. But is the industry simply playing a game of Whack-A-Mole: where you solve one problem only for another to appear? Before answering that question, we have to look at both the technological and cultural forces that are currently shaping our industry.
Two main areas of concern are CPM and bot fraud. CPM fraud involves publishers knowingly trying to defraud advertisers by serving ads where there is no chance for them to be seen, let alone engage a viewer. Techniques include pixel stuffing and layering impressions on top of each other. Video is another area where fraudsters will again stuff videos into impossible to view 1 x 1 iframes, or loop videos continuously without being shown to users.
Bot fraud, or non-human behaviour is where a computer (or more usually tens of thousands of computers) have been taken over by a bot which instructs the machines to serve ads behind the scenes, and unbeknownst to the user. The scale of this is mind-boggling as millions of ad impressions are generated on a daily basis with absolutely no chance of being seen by a human.
This means that advertisers are paying for impressions that appear to be served. So yes, a worrying amount of advertising revenue being wasted, but it’s an inevitable feature of an increasingly programmatic industry. Put simply, by removing the human element from direct trading, you inevitably give the fraudsters more opportunities to operate.
But will the industry always be this way, or can we beat fraud for good? Because the frausters are constantly evolving, there’s no simple answer to that question, but there are steps we can take to minimise the impact of fraud in the interim. Firstly, we need to see more collective action. If the industry fails to work together to beat fraud, we risk losing the trust of marketers just as the medium is getting exciting with possibilities and being embraced by brands.
Secondly, we need to be realistic about the scale of the problem and minimise the scaremongering as some of the figures about advertising waste and fraud have been greatly exaggerated. We see around 30 billion impressions monthly on a global basis, and this robust data shows us that this is not a 60 percent fraud problem. Our quarterly report released this month shows that 15 percent of video is fraudulent, while 13 percent of digital is fraudulent.
Thirdly, the scale of the industry means that manual processors can’t find and stop fraudsters. Even putting our company’s self-interest aside, the reality is that all we can do is to leverage tools to help brands identify and rid your exchange, network or campaign of fraud.
Finally, with such a lucrative revenue stream, fraudsters are constantly evolving ways to try and stay one step ahead of being caught or shut down. We need to move away from correlation-based last touch models that actively incentivise fraud. The good news is that we are constantly developing ever more sophisticated ways to identify their methods and remove them or block ads being served to them. The technology is there — it’s time to start using it.
Read more here.
UK | ISBA: Don’t Believe the Hype
If you have read the trade press recently, you will no doubt be thinking that robots have taken over digital advertising and that everything sane about using this marketing channel has gone to pot. You will also no doubt be feeling bewildered about what is happening to your marketing budget and the potential waste that is being reported by these fraudulent practices. I don’t blame you, I would too. We spend our money in good faith, and the least that we should expect is a fair transactional environment.
Headlines such as "60% of Mercedes ads served to bots", that ran in the Financial Times this summer, catch the eye! They are supposed to. But can as much as 60% of digital impression really be generated by bots and not humans? And that the cost of this is being passed onto unwitting advertisers?
Let’s be honest with each other. There is a fraud issue within the digital advertising space, just like there was with click-fraud in the Search space around 2002/2003, but the good news is that the problem is nowhere near as big as some people would like to have you believe.
We see over 30bn impressions every month, so we feel very confident that we have robust data. We do not see a 50% or 60% fraud problem. It is closer to around the 13% mark for display and 15% for video according to Integral Ad Science Q4 2014 Media Quality Report. Of course, these are averages, and we see these numbers increase on networks and exchanges, due mainly to the absence of direct trading and a move towards more programmatic channels. This is to be expected, as you remove the human element from the equation, you give the fraudsters more opportunities to operate.
I am not for one minute suggesting that these numbers are acceptable, and we need to work hard to remove this wastage from digital campaigns and mitigate the risk of digital ad fraud.
So why do some people highlight such high fraudulent numbers? If you were being cynical, you would say to sell products or services. Maybe. However, the real answer is more likely to be all to do with the technology being used and the ability to determine if an impression is fraudulent or not.
Not all technology is the same. Some vendors see one impression on a website and brand the whole domain or website as fraudulent. Imagine if that one bad impression was on a site as big as The Daily Mail - you would be dismissing a site with over 70 million impressions every day as a fraudulent site, and removing millions of good impressions with the one bad fraudulent impression.
This response to the fraud problem is no good for genuine publishers or advertisers, as it has the potential to restrict the supply chain and over inflates the real issue.
We do have a problem with fraud, and it is important that we remove this wastage from digital trading, but it’s not the 60% that some people would like you to believe.
If we are going to improve this situation, it is important that marketers and their agencies work with credible vendors who have the technology that understands that this is an impression level issue and who can help you reduce wastage by blocking ads served to the Fraudsters and help you target your ads and advertising revenue to those genuine publisher that are working hard on your behalf to help you reach your campaign goals.
Read more here.
Digital ad fraud: Separating facts from fiction
Combatting ad fraud in some ways is like auto care. If you are like me, your background and education did not include automotive training. When I get my car serviced, I’m at the mercy of the mechanic. The mechanic will explain the problem, I can get a second opinion, but I don’t have the knowledge to evaluate the real needs. At the end of the day, I will go with what seems most logical, which under the circumstances, is my best way of making an informed decision.
The same is true for digital fraud – you hear different things – but who can you trust? Let me share with you some recent real-life examples. At Integral Ad Science, we encounter these often, both from the solution providers as well as those who try to (sometimes openly) outsmart fraud detection and enable fraud practices to continue and thrive.
The first example that comes to mind is creative marketing. Recently, I read about a new fraud operation that was discovered. Usually the story involves new forms of ad fraud with new names to go along with it. To make it an even a juicier story, brands who were “victimized” are named, and to complete the story, ad fraud technology vendors who missed the new form of fraud are thrown under the bus. This is great marketing (and as a marketer, I applaud the news making). However, turns out there is nothing new about this type of fraud, except for the name. More times than not, it is a well known problem, addressed by many. As for the brands who were allegedly victimized and the vendors who didn’t catch it? Well, the media is hardly a court of law and they are not given the opportunity to present their case.
Another classic (and clever) example is a vendor who turns technology limitations into a perceived advantage (again, brilliant marketing). Have you heard that blocking non-human (bot) traffic is bad for the industry because it tips off the fraudsters, thus enabling them to outsmart the technology? The simple truth is that if you are only using one particular detection method – Session Based Signals – it might be possible for botnet operators to learn how to avoid your detection. However, if you layer in additional methods such as anomalous pattern detection based on deep traffic analysis, it becomes a lot more difficult if not impossible for the other side to reverse engineer the solution.
And what about those who are happy to avoid accountability, and will continue to buy and sell very cheap inventory turning a blind eye (or not) to it being fraudulent? We see those who try to sell traffic that “fools” the vendor technology. Well, they will have hard time doing it with Integral. We dynamically detect both bots and inventory sources, so the games with using new sites and traffic sources will quickly fail as our solution is smart enough to identify the machine as bot even on a site never seen before. If it’s a bot, we know it.
What I’ve shared with you is only the tip of the iceberg. There are many more examples and misconceptions about digital fraud, some of which were previously addressed.
The reality is that only a select few who are well versed in both the technology and applications of digital advertising will be able to separate facts from fiction. Digital fraud is here to stay, so we can’t ignore it. The best advice that I can give you is to do your homework, “look under the hood,” ask questions, and find the solution that is right for you.
Read more here.
Can I please see some ID?
Whose IDs To Use?The problem is that every ad tech vendor has its own internal IDs for advertiser, campaign and so on. While every vendor could share a dictionary of its IDs with every other vendor, this is not realistic, as anyone who’s ever looked at a LUMAscape can tell you. There are many, many vendors in the industry, and each one inevitably needs to work with each of the others at some point. What we need is an established standard where we all agree to use one particular set of IDs. Fortunately, we do have a kind of standard today as most in this space use the IDs from the advertiser’s ad server. This makes plenty of sense, given the central role of the ad server to a campaign’s delivery. Ad servers for their part have been supportive in their role as “keeper of the standard IDs.” For every impression, they pass the values of their IDs down the delivery chain, from one ad tech vendor to the next, via the mechanism known as “macros.” The problem lies in knowing the meaning of the IDs. The ad server may tell you the advertiser for this impression is 11981, but who is that? Only the ad server knows. The question then becomes how do the meanings of the ad server IDs get communicated out to the rest of the ad tech ecosystem?
If You Didn’t Get My Mapping File, Check Your Spam FolderThe sad truth is that sharing the meanings of ad server IDs from one vendor to the next is mostly a manual process. The advertiser extracts the IDs from the ad server into a spreadsheet and emails that spreadsheet to the other ad tech vendors supporting its campaign. The vendors then map those IDs to their own IDs, using the spreadsheet as their Rosetta stone. No one is a fan of emailing spreadsheets. Why can’t the other ad tech vendors just query an ad server’s application program interface (API)? For instance, if a company’s pixel serves on an impression with the advertiser ID macro set to 11981, can’t the company just query the ad server’s API? It could send to the ad server the number 11981, which is meaningless to the company, and the ad server would send back the name of an agency. The good news is that most, if not all, ad servers have APIs. The bad news is that access to them is usually limited to their paying customers, meaning only the advertisers. So the other vendors are out of luck.
Sharing The WealthThis is more of a business problem than a technological one. I can imagine a master services agreement between an advertiser and ad server that extends access to the ad server’s API to all other vendors participating in the advertiser’s campaign. I can also imagine each vendor having a single permanent login to an ad server’s API. Someone at the ad server or ad agency would then grant that account access to the relevant campaigns within the ad server, rather than creating a brand new account for the same vendor with every new campaign. If the vendor ever requested information for IDs that are part of a campaign they’re not working on, the API would return the appropriate error message to them. APIs are clearly preferable to spreadsheets that are created, sent and processed manually. The meanings of ad server IDs are not a secret. Every day, these mapping files are passed from hand to hand, and company to company. Ultimately, for an advertiser’s campaign to be successful, it needs all of its supporting vendors speaking the same language when it comes to IDs. A more open API policy for ad servers would be a huge win for all involved. Read more here.
Education Is 2015’s unexpected fraud challenge
Recently, Integral Ad Science surveyed members of the online advertising industry – including agencies, brands, DSPs, networks, publishers, and trading desks – to better understand their concerns for the year ahead. Not surprisingly, ad fraud topped the list.
All agree that it’s a significant issue; 89% of the survey respondents believe it has a direct impact on media quality, and 39% think it trumps other ad quality measurement factors: brand safety, viewability, transparency, and geo-compliance.
But, given the media buyers and suppliers responses about how ad fraud occurs and to combat it, it’s clear that more education on the issue is needed. Less than half of respondents say they understand what’s needed to detect and curtail fraud. That’s understandable, of course, as ad fraud is a complex and evolving issue. Anti-fraud technology changes quickly out of necessity, and can only be driven by teams of data scientists armed with knowledge and skills that the average buyer or supplier doesn’t possess.
While most of us are comfortable leaving the details of algorithms, temporal patterns, and big data parsing to the various data scientists and analysts who spend their days examining these issues, we should at the very least agree on the enemy. But, that’s harder than you may think. To start, the definition of fraud is often misconstrued, and there are precious few hard-and-fast rules for applying a nefarious label to a single impression or site. For instance, bots have a bad name in the industry, yet there are good bots and bad bots – only the latter of which perpetrate fraud. Additionally, while a domain or a Web page may be considered low quality, assessment alone does not make it legitimate ad fraud. Similarly, just because a website previously fell prey to fraudulent traffic doesn’t mean it should be avoided forever; after all, even the most premium publishers can be subject to hit-and-run bot attacks.
So, how should the industry define fraud? The generally accepted definition is any illegal activity that prevents ads from being served to human users. The culprit is nearly always illegal bot traffic, easily scaled and perpetrated by botnet operators and hackers who reap lucrative rewards for their tactics. Less prevalent offenses, such as ad stacking (i.e., placing multiple ads on top of one another in single placement so the only the top is in view), and pixel stuffing (i.e., stuffing an entire ad into a 1×1 pixel) also defraud advertisers of millions of dollars each year.
If we know the types of fraud favored by bad actors, why can’t the industry simply create rules to shut them down once and for all? Most fraud-detection efforts amount to a micro-level game of cat-and-mouse. As soon as one type of fraud is identified and stopped, the fraudsters develop new tactics to get around the current detection systems. That’s why the industry needs to target the fraudsters’ business model – not just their tactics. We need to create an environment where it’s no longer profitable for fraudsters to run manipulative operations. This requires a multi-pronged approach that includes both micro-level methodologies (catching the mice), as well harnessing big data on a macro-level (removing the cheese altogether).
At the 2014 IAB Annual Leadership meeting, the Bureau’s chairman, Vivek Shah, made a compelling case that it’s everyone’s responsibility to fight fraud – from publishers and tech middlemen who are paid whenever they serve impressions, to agencies who buy fraudulent impressions that look real to their clients. While great advice, Shah presents a tough challenge to an industry where everyone is struggling to stay up-to-date on the latest technologies and trends.
Recent coverage of ad fraud has sparked awareness and the need for more education – showing that we’re making progress towards fighting it. The fact that we’re even discussing the complexities of anti-fraud methodologies is a clear indication that the entire online advertising industry is advancing in significant and meaningful ways – and tackling its toughest challenges head on.
Read more here.
Can we resolve online ad fraud?
In my last ISBA blog post, I entitled the piece "Don't believe the hype", as we had seen some big scary fraud numbers in the trade and mainstream press, quoting as much as 60% of impressions being delivered to fraudsters. I wanted to inject some realism and perspective into the debate in the UK backed up by our data here at Integral Ad Science. On average we see 13% ad fraud in display and 18% in video. In that same post however, I argued that even these numbers are unacceptable, and that we need to work hard to remove this wastage from digital campaigns and mitigate the risk of advertising fraud.
So what is the industry doing to tackle online ad fraud? The good news is that progress is underway. JICWEBS and the IAB chaired a meeting in December, bringing together for the first time, its members and fraud technology vendors, in order to identify the extent of online fraud in the UK. This is the first step towards forming a working group tasked with tackling the UK fraud challenge, and is likely to be made up of JICWEBS members (from ISBA and IPA, through to AOP and the IAB, among others). Ad fraud impacts the advertising industry as a whole, from quality publishers, right across the ecosystem to DSPs and exchanges. It is, therefore, important that we engage as many relevant voices as possible and build a broad consensus for dealing with this issue.
What can we expect from the newly formed working group? Well firstly, if we look at the recent brand safety work from the DTSG, I think that we can expect to see recommended best practices for both the buy and sell side for dealing with fraud. As the group is to advise companies from different parts of the digital ecosystem, these guidelines should be realistic and sensitive to the needs of the various businesses.
Secondly, as digital ad fraud is a relatively new topic for us all to grasp, I am sure that education will also need to be a key feature in the working group’s discussions. A recent piece of AdExchanger research aimed at marketers set out to provide practical guidance for dealing with fraud. I think that we need to see more research and guidance of this kind, and the experiences of this working group will hopefully be able to provide that needed actionable advice.
Finally, to win the confidence of the broader advertising industry, I think that vendors in the fraud space will need to be accredited for the claims that they are making on behalf of their technologies’ ability to detect and deal with digital ad fraud. I foresee this accreditation process being managed by ABC, who is already doing a very good job verifying brand safety and viewability vendors.
I think that the UK digital advertising industry has learnt a valuable lesson since viewability came to the forefront of market discussions. Over the last two years I have attended meetings to debate the subject, and I always got the impression that the viewability horse had already bolted. The buy side were very quick to grasp the benefits of viewability, and in many cases left everyone lagging as they rode ahead. Even today, media agencies are defining the viewability market, as we see buyers moving towards customisable viewability metrics that optimise individual campaigns. It does not feel the same way with fraud; it feels much more joined up. Our trade bodies have been quick to react to the fraud problem and have responded by working together to try and address the issue head on.
The second meeting for the newly formed working group is taking place in January; Integral Ad Science will be part of these discussions and contributing to the efforts to tackle ad fraud in the UK and outline some best practices.
We are under no illusions that there is still some way to go, but this is a positive step forward in addressing the issue and educating the advertising industry. Online fraud presents a real challenge to the ecosystem, and is an issue we need to resolve today.
Read more here.