Data Driven Attribution for Google Shopping / PLAs


This article was published in May 2016 when DDA was still in beta phase. Information can be outdated by now.


Google is finally solving a problem which existed since the beginning of conversion tracking. Soon you will be able to apply different attribution models to your conversion tracking and use the attributed data for bidding! Say hello to the data-driven attribution model.


Why is this important?


Customer journeys can be very long and complex, including many touchpoints with different product-targets/keywords, campaign types, devices and ads. Getting a clear picture of the attribution behavior and attributing a fair credit to each touchpoint, therefore can be very intimidating for many advertisers. But this will get easier now!

Google Shopping and PLA searchers produce many touchpoints during their progression through the purchase funnel. They search for new ideas and browse through many different options, followed by a reduction of these options to a more narrow consideration set and finally hopefully deciding for one of these options converting through a PLA or branded search. Of course, many searchers just enter the purchase funnel at the decision stage, but it would be very short-sighted to ignore the browsing behavior of other searchers in the top funnel.

With this upcoming feature, you will be able to get a more holistic view of your Shopping Campaigns and also the interaction with other campaigns like branded search.


What will change?


Currently, you are able to just analyze the attribution behavior of your campaigns. Furthermore, you are only able to view the attribution data on the campaign, ad group, or keyword/match type-level. This will change: the new feature represents a switch from ‘passive analysis tool’ to ‘campaign steering tool’. You will be able to select an attribution model that will then be applied to your conversion tracker — you are then able to use the attributed data for bid management.

There are six options to choose from: last click (this is/was the default), first click, linear, time decay, positon-based and a new option: data-driven. We will talk about this great new option in a second, but first a quick look at the already existing ones with an overview and data from Google:


2016-05-13 11_03_52-Untitled presentation - Google Slides


Data-Driven Attribution (DDA)


The new data-driven attribution option is the most interesting attribution model. Why? Because it takes out the guesswork of choosing a model. As for all data-driven and machine-learning-based algorithms, you will need enough conversion data for that. Google suggests that you should obtain about 800 conversions per month — the more the merrier.

Data-driven attribution will evaluate all the different paths in your account to determine which touchpoints are the most influential. Several factors like the number of interactions or the order of exposure are incorporated in the calculation. By using a counterfactual approach, the algorithms try to compare what actually happened with what could have happened to determine which ad clicks are the most critical for obtaining a conversion or missing a conversion.


Data-Driven Attribution for PLAs


As Google announced that DDA will be available for the Search Network we think that the technical concept of DDA will be built on the query level and triggering criteria (like keywords and product targets). Based on the currently available information this is not clear yet, but we will keep you updated on that. If this turns out to be the case, then it would be possible to get more information on the assisting behavior of generic shopping terms to closing terms like very specific product terms. (eg. “running shoes” vs. “nike free 5.0 men’s running shoes”). Currently, you would have to create shopping campaigns per query type (let’s say: generic and product-specific) and funnel the traffic with query sculpting (adding negative keywords to control the traffic flow) into the right campaigns and analyze the assisting behavior on the campaign level.

With DDA every search query will get attributed credit and this will shed some light on the real performance of very generic shopping terms and maybe change the way you deal with them.


What does that mean for you?


The new attribution behavior for conversion tracking will be available within the next month. In the meantime, you should dive into your attribution reports and get a feeling of how the specific attribution behavior turns out for your account. If the feature becomes available you are able to apply the attribution behavior which you like best. We encourage you to try out the data-driven-approach, especially if you have way more than 800 conversions per month. One big remark: If you are using Automated Bidding Strategies from Google for your Shopping Campaigns they will continue optimizing based on the last click model. Whoop will continue to optimize based on the resulting conversion data.

We also love to hear feedback from you – so feel free to share your learnings with us and the community!