One of the most powerful aspects of digital marketing is addressability—the ability to target and tailor marketing actions to consumers and directly attribute the resulting business outcomes to those marketing actions. The attribution part is the key to enabling marketers to understand the return on their investment and then optimize that investment accordingly. Unfortunately, while targeting capabilities have improved, the approaches to measurement and attribution widely used today suffer from several fatal flaws, namely last-touch attribution, reliance on the cookie, missing the offline view, and failing to measure incremental impact. Even where these flaws are well known & understood, the typical measures in place to address them fall short, leaving many marketers struggling to realize the promise of digital marketing. In this 4-part series, we’ll dive into each of four common but fatal attribution flaws, including a description of the problems, the complications marketers face in trying to address them with current solutions, and how new technologies and approaches being developed now can successfully resolve them.
Fatal Flaw #1: Last-touch Attribution
The Problem. Perhaps the best-known flaw is that many systems attribute 100 percent of the marketing impact to the last media interaction with the consumer. For example, let’s say a consumer watches an online video ad for a product, prompting her to do a search for that product. She then clicks on the first result (a paid search result) and purchases the product. Under a “last-touch” (aka “last action”) model, the search ad would receive 100 percent of the credit, while the video ad would receive no credit. This approach undervalues upper-funnel marketing touchpoints that generate demand and overvalues lower-funnel touchpoints that merely fulfill demand. The result is an inaccurate view of marketing performance, leading to misaligned budget allocations across media types and channels.
The Complication. Many marketers have developed custom multi-touch attribution (MTA) models—either independently or working with attribution modeling partners—because they understand last-touch attribution is wrong. These MTA models use advanced quantitative techniques to ascribe the correct “partial credit” to various digital touchpoints along the consumer’s path to purchase. However, despite generating better insights, these models are typically not linked back to the real-time buying systems (e.g., DSPs) governing media execution. In the worst cases, the models simply end up languishing as interesting insights in PowerPoint. In slightly better cases, these models are translated into coarse recommendations to manually adjust wholesale “blocks” of media (by channel, publisher, placement, etc.) up or down. Yet that fails to deliver against the true potential of these models in a programmatic context, because it does not de-average buying decisions down to the individual impression level; it effectively treats all impressions within those blocks as equal. The power of MTA models comes from their ability to discover precisely how much impact is driven by different combinations of data and to credit each impression accordingly. Not acting on that granular, impression-level data is simply throwing the information away and leaving money on the table.
The Resolution. Realizing the full benefit of MTA models requires taking the fine-grained output of those models and using them to inform real-time buying decisions in an automated fashion. Not just by channel, publisher, or placement, but by the full combination of all available user- and impression-level data (including audience segments, geography, time of day, viewability and myriad other variables that these models can analyze). This enables programmatic bidding to be adjusted in line with value, which is the key to maximizing performance. We call the integration of MTA model outputs into real-time execution “Closed-loop Attribution.” Closing the loop from attribution back to real-time, impression-level execution in automated fashion not only improves performance, but eliminates the manual effort and workflow associated with the wholesale approaches often used today.
Next time we’ll discuss the second fatal flaw: Reliance on the Cookie.