Be honest—we might talk about marketing attribution, but really attribution hasn’t been possible for about 15 years. Although several models exist in Google Analytics, each is more imperfect than the other. Google honestly tried to offer a dynamic, data-driven version in the last enterprise update of Google Analytics, but it didn’t seem like this was much of a success either, despite their good intentions. The overwhelming majority of marketers use last-click, even though we’ve known for some time that this provides an incomplete view.
For example, take the following simplified customer journey:
Despite the numerous points of contact, Google Analytics will attribute the conversion to brand SEM (the last interaction) despite the fact that this interaction simply could not have happened without the ones prior, which had a much higher causal impact.
But why do marketers continue to use the last-click model when they know it’s totally inadequate? My answer is a very personal one: because they’re human.
In short, I sincerely believe we’ve only been pretending to be interested in attribution and that as digital marketing industry players we simply looked the other way at times when we should have been thinking for ourselves. We have little to lose in reconsidering these practices.
The shift towards digital privacy, mainly involving the decline in third-party cookies and changes to iOS14, continue to impact our sector. Like drivers facing increases in the price of gas, advertisers are seeing the cost of their primary material (audiences) become more burdensome, currently paying more for the same marketing activities, particularly on Facebook. Facebook even stated in internal communications to agencies that its cost per action (CPA) had increased by a factor of 2.5 following these changes.
Third-party cookies also had an effect on a number of other measurement and management tools, particularly Google Analytics and Google Campaign Manager.
Granular tracking of each user seems to be over. Even tools enabling third-party cookies to be turned into first-party cookies, like Google Tag Manager Server-Side or Facebook Conversion API, seem more like band-aids in a context where users are increasingly cautious about their privacy and when tools to protect them are multiplying.
We think that individual tracking approaches are simply not viable in the long term. In this context, the situation needs to be approached from a completely different angle. In addition, I’m of the opinion that our obsession with hyper-granular tracking in digital has blinded us to the overall effects of digital marketing and its effects on an individual company’s revenues.
This granular approach was suitable back when digital had to fight to earn a place in advertisers’ budgets. But now that the battle has been won, it’s time for digital to come of age and consider its contribution to overall performance.
Well before digital arrived to disrupt everything, MMM (media mix modelling) was used to analyze the causal impact of a media tactic on a company’s total revenues. Today, this type of analysis can still be performed, but technology can play a role in improving its quality and increasing its depth.
Marketing data warehouses facilitate unifying the two major types of marketing data—transactional data and behavioural data—and performing a full analysis of them. Transactional data must also include offline data such as from checkout systems, for example. Once these data are unified and ready for analysis, it’s possible to perform causal analyses to refine attribution in a much larger and precise sense than with web tools and their distorted viewpoint.
Even more interesting is that the union of these two types of data also lets you perform predisposition analyses and therefore predict the performance of some segments. These analyses come from a family of statistical inference which proved some time ago that the future can be found in current data.
Here are a few concrete examples:
We’ll be working a lot with these approaches from now on and I don’t think we’ll be going back.
I’ll close with a tip: I increasingly recommend using Google Analytics (4, obviously!) only to capture data and to then immediately export them to a marketing data warehouse like Big Query in real time. This will allow you to free yourself from digital and have a much bigger impact on your organization’s performance.