Last Click Attribution: A Simple Way to Misallocate Your Budget

With many new options in the marketplace for purchasing display media, advertisers are taking a hard look at the value of each “link” in the traditional advertising value chain, from agencies to networks to data providers.

The links in the purchase funnel (display ads, search, etc.), however, still manage to escape similar scrutiny. Imagine if you rewarded your ad server with the bulk of your media budget, simply because they are closest to the ad’s final placement. Last-click attribution essentially takes the same approach. While reluctantly accepted for its simplicity, the model’s strengths end there. Currently, a substantial number of our advertisers are applying a last-click attribution model for their campaigns. Advertisers are well aware of its shortcomings, which have fueled the debate about display attribution models for years.

Recognizing these challenges, DataXu’s Advanced Analytics Group frequently works with customers to analyze their unique attribution data and develop customized strategies for their display campaigns. Taking a look at this data for six campaigns, the team identified some interesting trends.

Our analysis above shows the cumulative volume of conversions from the first exposure to the ad. In each campaign, consumers received one or more additional impressions as the campaign progressed.

Data Highlights

• For all campaigns, last click attribution ignored 97% of spend driving conversions — which often results in over-spending in search and re-targeting, and under-spending in display that drives demand creation.
• The recommended attribution period for short consideration products, such as CPG, is two weeks. In the campaign shown above, this window includes 90% of impressions that converted.
• The recommended attribution period for long consideration products, such as Insurance and Autos is five weeks.
• The length of time it takes to attribute 90% of conversions varies by 250%; each product requires its own distinct attribution model.

Ultimately, no two campaigns are alike, and what drives their success is tough to predict at the outset, but incredibly valuable to understand.

“With more powerful measurement and analytical tools now available to marketers, the emerging best practice is to create an attribution model specific to the item advertised and use the model to measure and pay media partners based on their respective value creation,” said Sandro Catanzaro, VP of Analytics & Innovation at DataXu. “Current attribution models misalign value creation with value capture.”

These results have many implications for display campaigns. Similar to the analysis in our previous MarketPulse, a key conclusion is that one size does not fit all when it comes to attribution models. At DataXu, we offer a range of unique analytics and technical capabilities to help our customers uncover the real value drivers of their campaign across its lifetime, from use of third-party data to impression-level optimization to attribution models. By putting these powerful insights to work, our customers can achieve superior media efficiency and ROI.