Chapter 8 Conclusions
There are two key message that one can take from this book. First, the market for targeting data is a race to the bottom. Second, it is notoriously hard to measure the incremental sales effect of digital advertising.
8.1 Data: A market for lemons
In 1970 the the Nobel prize winner G. Akerlof described a market situation in which buyers can expect only subpar products (so-call lemons) to be offered. The market for data market is a perfect example for a lemon market. The Akerlof’s idea is that if buyers of goods cannot distinguish between high- and low-quality goods/providers, one can expect providers of high-quality goods to leave the market. As a consequence, buyers can expect only low quality (lemon) providers to be active.
This is exactly the case in the data market. As discussed above, advertisers have no way to validate data quality before running a campaign. Even if they run a campaign, measuring the incremental effect of targeting is in most cases very hard and costly. To make matters worse, there is no open rating/feedback system for data providers. If advertisers have had performance issues with data providers, they will keep the information internally in order to not help their competition. Finally, as advertisers run a limited number of campaigns per year the learning about data quality takes at least several years. From the data provider perspective, maintaining data quality is opposing to increasing reach and subsequently profit. When pressured by shareholders and investors, data provider will always lower the classification threshold in order to increase reach. As a decision maker within a data provider, it is very clear that one needs to invest in a sales pitch (including artificial intelligence buzzwords) first before investing in higher data quality. To sum up, all economic incentives drive data quality towards the bottom.
8.2 Ad effectiveness: The industry is not striving for better measurement
In a 2018 article the world leading business consultancy states: “In recent years, the proliferation of technologies that can process massive data sets, combined with the growth of digital advertising channels—which are inherently more measurable—has unlocked a massive opportunity to measure the performance of marketing investments. 6
It is easy to get caught in the number of digital metrics and the promise of better measurement. However, these claims are just wrong. As discussed in the previous chapter, digital targeting and measurement routinely ignores the fact that users have a base probability to buy a product or service. Instead of measuring the incremental number of converted users it takes the total number of converted users as denominator. Hence the digital ROI is often massively inflated.
When confronted with measurement approaches, the industry response is a smoke screen called attribution. This set of methods simply distributes all digitally measured conversions (excluding direct type-ins) between the involved digital channels. For a traditionally channel like TV this would mean that all offline sales of persons who happen to watch a TV commercial in the last 30 days are driven by the commercial.
In contrast to digital, it is the norm for offline channels to measure ad effectiveness as an incremental effect to baseline sales using market mix modelling. This incremental measurement is rarely done in digital channels and most industry participants work hard to distract advertisers to measure the true ROI.
All in all, the value of targeting data for society is very limited. The massive tracking systems are used for a marketing machine which cannot accurately measure the incremental effect of the collected data. Rather, the data collection is used as foundation to argue for the superiority of digital channels over traditional marketing methods.
On the flipside, the public is increasing aware and skeptical of the massive data collection. Public bodies like the EU started to increase regulation. The General Data Protection Regulation (GDPR) is the most important change in data privacy regulation in 20 years. GDPR substantially reduced the extent to which targeting data can be shared, processed and stored. Without explicit consent, data providers are not allowed to collect data on users.
Finally, some technology providers realized that data privacy can be a differentiator. Intelligent tracking prevention (Apple, Firefox, ad-blocking) stops 3rd party pixels from firing in the first place. Hence, by default user groups with these devices and browsers are not tracked and cannot be profiled with the current technology stack.