Marketers in the programmatic digital ad space are no doubt familiar with the term “multi-touch attribution”, or MTA. When we speak of attribution in this context, we are speaking of a specific digital “touch” in the customer journey to which we can attribute more or less weight that leads to a “conversion”. A conversion is something measurable: it can be the click of a mouse, a visit to a website, a sale, or some other behavior desired by the marketer. The availability of data at the granular level (individual, behavior, time) has led to the field of analysis known as MTA modeling. Each touch along the attribution train may carry more or less weight depending on where it falls in the sequence of touches for each individual consumer.
Now enter the more recent notion of “walled gardens” (aka the ‘attribution apocalypse’). Major providers of digital data, e.g., Facebook, Google, Amazon, Apple, smart TV manufacturers such as Samsung, digital ad exchanges, first-party providers, and any other sources that monitor digital journeys (aka ‘digital exhaust’), are starting to say “we don’t want to play with you”. They are beginning to erect significant barriers to digital data access due to two key factors:
Regulatory pressure about privacy (e.g., GDPR, California’s CCPA, and others), and;
Enlightened self-interest: digital data owners (e.g., Facebook) are preventing outside access, and by implication now claim that they are the best choice for performing any analysis of their digital warehouse (a little like grading your own homework). This entirely freezes out independent companies from performing cross-platform/device analysis.
The sheer availability of addressable digital data identifiers (e.g., cookies) is also changing. Google is joining Safari and Firefox in blocking third-party cookies in its Chrome web browser (phased out over the next two years). They are happy to go slow: Google makes its money on search and the ability to target. So Alphabet and Chrome are a bit at odds with each other. While cookies were never intended to share as much information as they currently do, how we will replace them will be fascinating. One solution, written about here before, is blockchain. This is an emerging technology that depends upon decentralized identities (either a public blockchain, or a private/consortium-style blockchain) with data that can be acquired or shared by media measurement companies and attribution consultancies.
Ultimately, this is an economic decision that has to be made by the end user. How much information is a browser user willing to share in exchange for the convenience and power of the tools they now use for free? And how much of that information are browser developers willing to share with the media measurement companies that want their data?
Several important constructs in how advertising actually works are overlooked in the rush to leverage the massive volumes of digital data used in MTA modeling. Data scientists lack industry knowledge about building awareness, memory and message decay, decreasing marginal returns in advertising, and other dynamics that involve brand choice/evoked set, or for that matter, emotion. Three simple examples: context (i.e., environment in which an advertisement is delivered); creative (i.e., the ability of advertising to break through and persuade); and brand (i.e., salience and momentum) have been, more or less, neglected. I have written about the power of great creative in sales forecasting. This concept applies to MTA as well.
The objective of all delivered media/advertising, and especially for MTA, is to “get the right ad to the right consumer at the right time”. The hidden assumption is that the consumer is always in the mood to receive the message, and that the consumer fully understands the message. In a world of screen clutter and six second ads, some companies are beginning to change their media and messaging strategy.
For example, P&G has shifted its focus to brand penetration (i.e., reach). Excess frequency (which MTA delivers well) has been criticized as wasting media dollars (as I noted, data scientists simply aren’t familiar with decreasing marginal returns in advertising spend). With the savings, P&G is (re)investing in reach. P&G’s Chief Brand Officer, Marc Pritchard, stated “The best measurement is people who are searching. So when we see an increase in search, we see an increase in sales.” This is largely consistent with the overall message of the book “How Brands Grow” by Bryron Sharp, which emphasizes this point and provides many data-supported case studies. Arguably, targeting is perhaps less critical if a company’s products have few demographic or media consumption skews. For others, precision targeting is essential.
Up until this point, media and marketing measurement firms have enjoyed a good ride with MTA (such a pun). First-mover companies who are nimble, smart, and have deep pockets can implement big data projects like MTA. And they are achieving significant ROAS – often in the very high double-digit range. And, these companies are able to adjust their models in real time and can continue to reap significant rewards. But, as more competitors build MTA models (or the technology becomes less costly, or the tasks less daunting), a company’s relative advantage will diminish. Think of it as an “MTA trickle-down effect”.
I wonder, in 10 years, whether MTA will be thought of as simply a targeting strategy to deliver excessive frequency for the short-term. Or, alternatively, a tool that really helped to build lasting brands and businesses. No doubt, the models will “learn” and become more precise and more “brand conscious”. One of the thought leaders in this space, Joel Robinson, often talks about “brand” vs. “performance” marketing. This is a very useful and provocative discussion: MTA penetration is now at 45% of US marketers based on the 2019 Mobile Marketing Association marketer study.
Until then, I hope we do not lose sight of what brands are all about, and that some aspects of brands are simply not measurable. That is, after all, the essence of brands – and marketing.