Fraud, tech taxes, brand risk, and useless metrics
Can behavioral advertising be saved?
The first banner ad saw the light in 1994. On its 25th anniversary, a case could be made to credit online advertising with helping the Internet take off, unlocking the many possibilities we have all enjoyed greatly.
There is probably little appetite for celebration, though, given how it has evolved into an unmanageable gathering of fraudsters, hidden brokers, fortune tellers, and gamblers - which I will duly proceed to introduce before anyone feels terribly insulted.
The fraudsters
According to Juniper Research’s Future Digital Advertising (published on May 21st 2019), Global ad fraud will amount to $42bn in 2019, a 21% increase from 2018, expecting to reach $100bn by 2023.
Whether truly reflective of fraud or possibly mixed with other non-human traffic (legitimate crawlers and bots), this figure will represent 20% of eMarketer’s expected global digital ad spend by 2023, with current fraud levels already surpassing that threshold for programmatic display and video ads (21% in the US according to Pixalate’s Q3 2018 Ad Fraud Report). Click fraud for both video and display ads is also estimated to have already surpassed 25% of the overall spend.
As per eMarketer’s 2019 Ad Fraud Report, three fourths of publishers behind programmatic inventory have adopted the Ads.txt standard in the hope to prevent fraud, but this does not work beyond the open web and can only protect against domain spoofing and arbitrage - while the problem is fast spreading to mobile in-app ads (its toll amounting to 20% of the total, according to Pixalate) and Smart TV/OTT ads.
Being entirely fair, fraud is not unique to behavioral advertising. And we should probably expect it to simply follow the money to whichever alternatives emerge - suffice it to look at Influencer Marketing*.
*Sway Ops (an anti-fraud business) estimated in 2017 that 50% of daily influencer interactions were fake, with 40% of all comments being made by bots. Separately, Point North Group has reported a fraudulent fan base that reached 78% for a particular brand (Ritz-Carlton) in 2018. Overall, the same study reveals an average 20% of fake fans across all micro-influencers globally (50k-100k followers), with certain countries reaching 28%.
The brokers
The US Association of National Advertisers (ANA), along with Ebiquity and others, estimated in 2017 that digital supply chain data and transaction fees take 48 cents out of every dollar of programmatic transactions, reaching up to 62% of the total investment when agency trading desk fees are included. Hoping to help brands avoid unreasonable charges, the Interactive Advertising Bureau (IAB) has even come up with its own Programmatic Fee Transparency Calculator.
Given that such fees are applied against gross budgets (pre-fraud), there is little incentive for any of the parties in the chain to stem ad fraud.
In the meantime, publishers are only able to compete for a fraction of the total media budget (post-brokerage, post-fraud) and, since “quality journalism” competes on equal footing with free, social, ever-multiplying ad inventory in a market that bids for audience profiles rather than context, every placement’s ability to measure and target a given individual matters much more than the quality or stand-alone success of the piece itself.
Eventually, the same marketers that once demanded short-term performance and extreme measurability get to complain about “brand safety”. The problem gets worse in the context of walled gardens (YouTube, Facebook), extremely efficient at such precision game.
Partly as a result, contextual advertising and direct brand-to-media deals seem to be back on the table, with advertisers and publishers realizing that triplicating Click-Through Rates may not be worth all of the side effects of out-of-control intermediation (overall CTR average remains a meager 0,05%, according to the April 2019 Smart Insights benchmark).
As reported by Digiday with regards to The New York Times’ deep-dive into this approach (in Europe alone):
“rather than bombard readers with consent notices or risk a clunky consent user experience, it decided to drop behavioral advertising entirely.”
The newspaper managed to “significantly” increase its ad revenue as a result. Admittedly, this option alone will probably not save most other publishers.
The fortune tellers
Measurability of results has always been online advertising’s best selling point. The case is strongly built on digital “touchpoints” for which marketers can easily define their own systems of record.
Aggregate data collected on websites, mobile apps, or social media can help wonders in the tactical and strategic management of various marketing functions - and I for one have spent a considerable number of years obtaining tangible value from such first-party data and their correlations with other sources.
It is, however, easy to get carried away. An explosive combination of the engineer’s mindset (currently presiding over many a marketing function), financial imperatives, and collective hype, has quickly led to a new religion: Predictable ROI (call it “multi-channel attribution”). As if individual customers could really be "deduplicated" across every digital asset, every device, every decision, by every brand and every broker. As if building a Single Customer View was truly attainable.
As it turned out, people change browsers, delete cookies, switch devices, avoid cross-service credentials… and happen to be human. Right across the aisle, human-driven brands make mistakes, tag their own properties or campaigns with mismatching references, lose their data, define the wrong parameters. But not even in the face of all of the above could we have imagined the many additional impediments thrown upon digital marketers in the past few months:
In case it was not hard enough to ensure that AdTech cookies remained in place, the first half of 2019 has seen more of them wiped away: on May 7th Chrome announced upcoming features to help users easily block third-party cookies and fingerprinting practices; on May 13th Apple released version 2.2 of Safari’s Intelligent Tracking Prevention, which maintains a seven-day life span for first-party cookies set on the browser side, besides blocking third-party cookies as it has done for some time; On June 4th Firefox moved to turn on its so-called Enhanced Privacy Protection, blocking third-party cookies by default (and equally exploring a move to block any first-party cookies set through JavaScript commands);
If we had issues ensuring that people saw an ad (around 35% of those served are not even shown), it is estimated that a third of the Internet population uses Ad blockers on both desktop and mobile devices (with reports oscillating between 11% and 47%, and throwing much higher numbers for younger age groups)
If anyone hoped that the “walled gardens” would allow external tracking or auditing, privacy compliance concerns have naturally forced them to stop sharing attribution IDs or allowing third-party tags
If we thought a cookie notification banner was intrusive enough as per the 2009 EU ePrivacy Directive, the GDPR’s redefinition of consent (when respected) results in a very small sample of users accepting whichever cross-site cookies are still able to make it past the browsers.
As a result, not only do we end up with pretty useless performance metrics. Multi-channel attribution, cross-site measurement, reliable profiling… are simply unviable beyond the context of ID-based walled gardens or the customer's personal sphere.
It certainly makes all the sense that marketers take control of their data, their models, their money. Data-driven capabilities are extremely important, but in a world in which consumers (and the law on their behalf) demand to know the exact purpose of every data point collected prior to it being collected, Big Data is no longer the name of the game. It is not about betting on a blurry future with as much as we can digest of someone else’s information. It is about anticipating needs and clearly showing what we can do with the specific data we are asking others to expose, for the more we collect the higher risk we assume. Call it Smart Data. Call it Distributed Data.
CMOs spend 29% of their budget on marketing technology, according to Gartner’s CMO Spend Survey 2018-2019. As per a recent MarketsAndMarkets report, a total of $1.8bn, out of such budget, went into marketing attribution tools in 2018. Whatever this represents for each individual advertiser, I guess it makes sense to discount it from its yearly demand generation investments, before we move on to an even bigger hole in the brand’s pocket.
The gamblers
As it happens, not only is the advertising circus blessed with all of the above. A dangerous defiance of gravity is not missing either.
With similar regulatory initiatives piling up around the world (and CCPA going into effect on January 1st 2020), EU Supervisory Authorities have started to act against consent gathering and data processing activities that run counter to the GDPR’s defined thresholds for valid user consent.
From a relatively unknown AdTech player (such as Vectaury) to Google (fined 50m EUR for pretty much adhering to industry standards in terms of transparency and choices), no single player in the Internet advertising ecosystem is free from scrutiny.
In the most decisive move yet in this direction, the UK’s Information Commissioner’s Office released a thorough review of AdTech and programmatic advertising practices on June 20th (2019), concluding that current consent gathering practices do not conform to the applicable ePrivacy/GDPR standards, with specific reference to the IAB Europe’s Transparency and Consent Framework.
But something very similar happens with Customer Data Platforms and other pieces of Marketing Technology: building the bridge between pseudonymous audience data and identifiable leads or customers requires a level of consent that very few players have truly put to the test - which, as a natural consequence of true, informed choice, will result in too small a sample for it to bring significance to the overall effort.
Conclusion
Summarizing much of the above, it is very likely that less than a third of a given advertiser’s budget ends up getting past ad fraud and brokerage fees. When it does, it contributes to the creepiest* and worst possible user experience of news, entertainment, or education.
The perks of behavioral advertising include serving useless cookies through annoying consent requests and, thanks to them, chasing consumers to whichever content has most relevance to them, regardless of its compatibility with the brand’s own values. Most consumers will not click on those ads anyhow (despite the many tricks on display to force such miracle).
The effort is also peppered with excitement, as user consent is deemed invalid anyhow and, along with other violations of applicable privacy laws, helps advertisers share eye-watering fines with publishers and "aggregators".
Internal multi-channel attribution and Single Customer View endeavors have not only failed to prevent all of the above, but they also help make a blind case for the perpetuation of lavish ad spend. It is easy to justify outrageous CPCs when data collection opportunities replace tangible business outcomes as a measure of success (“we may not have sold much, and we have paid a fortune, but look at all the data we have gathered!”).
As it happens, data comes with risk, and the more we collect, the less valuable it may be. This is particularly true in a world in which it plugging into Machine Learning as a Service (MLaaS) offerings may become much more efficient than aiming to reinvent the wheel every single time.
In the meantime, consumers are slowly taking control over their own attention and their own data. Empowered by ever-growing transparency and their individual perception of true value, and yet open to the emotional impact of powerful ideas which are artfully communicated. Expecting suppliers and brands to be aligned with their own principles and beliefs.
For all the excitement and sense of purpose that a “digital transformation” mission generates within each particular organization, most of these internal efforts are all but unique. A large majority of brands out there have placed their bets on true one-to-one customer connections as the ultimate means to disintermediate relationships, anticipate demand, save money, or increase Customer Life-Time Value. But, by their own nature, trusted relationships can just not happen through stealth data collection** and "clickwrap" contracts.
Customer Centricity is not about building as many consumer profiles as brands every individual ends up bumping into. It has little to do with deploying a Data Lake, a CRM, or a Customer Data Platform, and a lot with letting individuals obtain a Single Brand View, or facilitating personal data activation at their own request.
Twenty-five years after HotWired published that first AT&T banner, it is about time we all embrace chance and accept that behavioral advertising may have just been one more failed experiment on our journey to a better world :)
*As per Hubspot/AdBlock Plus’ 2016 report, remarketing is considered to reflect badly on the advertiser by 70% of the population.
**79% of 20,000 respondents surveyed by SAP/Hybris in 2017 claimed they would break up with a brand if they found it tracking them.