Month: April 2020

Attribution: the new deal in the post eracookies

With the announced end of cookies, measuring attribution is not getting any easier. On the contrary: the customer journey is becoming increasingly complex to reconstruct. This is an opportunity to revisit methods and question the purpose of attribution.

A jumbled puzzle on the marketers' table. An unfulfilled quest for the Holy Grail for the past 10 years. These are the images that come to mind as soon as one looks at attribution. One of those topics where theory seems far removed from practice.

In theory, attribution promises to evaluate the weight of the various contact points in the customer journey, so that partners can be remunerated at their fair value and the marketing mix optimized. This is obviously a key issue in the age of omnichannel. The subject has therefore seen its share of overbidding, with technological incantations based on artificial intelligence, magic algorithms and models that often look like pretty black boxes.

Attribution in the face of data black holes

In practice, things are... different. First of all, because the vast majority of brands stick to a last click attribution. Secondly, because the picture of contact points remains very incomplete. The situation is not going to improve with the announced end of cookies and the increasing regulatory constraints. If today Safari, which is hunting for cookies third parties, represents a black hole for attribution, tomorrow it will no longer be the only one in this case. Black holes in the customer journey will multiply.

This perspective of a web without cookies third parties questions all the actors involved in attribution. And their methodologies. The probabilistic method, which is based on the analysis of large volumes of data, is unsurprisingly losing interest since, in the technical and legal context, these volumes are shrinking... The deterministic method, which aims to reconcile the contact points for which the data is reliable, is experiencing renewed interest. Admittedly, in practice, it is based on small volumes of data - reaching 6% of reconciled data is already considered satisfactory - but it has the merit of being part of a global effort to make the most of first-party data.

Priority to first party data

This is the good news in this landscape: there are many ways to expand and improve the reliability of first-party data: from logins (on apps and sites) to loyalty cards to be presented in stores. While waiting to see more clearly which solution will take over from cookies, working on the contact points that identify prospects and customers seems to be the most pragmatic way to back up your attribution model with "concrete" data. Brands have understood this and are multiplying their initiatives: systematic packing at the checkout or the multiplication of emails during an order in order to "map" each customer with all his devices.

This approach, which is often centered on "owned" data, makes it possible to combine the last click model with a personalized model based on the mapped touch points. This is another way to go besides the 100% "last click" model or the unique "multi-touch" model. Because, as has been widely observed, there is no ideal attribution model, only personalized models based on the known customer journey and the reliability of the associated data.

Is attribution an end in itself?

Another interest of the deterministic approach, and of the actions that result from it, is that it helps to remind us that attribution is not an end in itself but a means to an end. The goal is to contextualize an increasing part of the activations. In other words, to send the right message at the right time to the right person via... the right contact point.

The measurement of attribution was already a highly constrained exercise before the rethinking of cookies . The latter can be experienced as one more (big) constraint or as a motivation to renew the approach to the subject. For example, by putting less effort into measuring correlations and more into measuring direct impacts according to concrete geographical and temporal variables.

For attribution, the post cookies era could well be the "back to basics" era. With hypercontextualization and first-party data as the main weapons.


Don't settle for just one attribution model!

Find out how MixCommander can help you optimize your investments.

To not miss any of the latest news from Commanders Act, subscribe to our newsletter!  

© Commanders Act. All rights reserved.
Powered by CREAATION.