“Data-Driven Thinking" is written by members of the media community and contains fresh ideas on the digital revolution in media.
Today’s column is written by Eric Berry, co-founder and CEO at TripleLift.
The idea that most consumers directly engage with digital advertising – even if it resonates with them – is a fiction built purely for attribution. Consumers don’t click on TV commercials, radio ads, stadium sponsorships or – by and large – digital ads, yet advertisers often demand this incongruous attribution.
Questionable behavior by numerous players in the digital ad ecosystem and the fact that it’s simply easier than the alternatives have made post-click attribution the de facto metric, but it is so deeply flawed as a measure of effectiveness that it might as well be random.
The ultimate goal of any attribution program should be to measure the effectiveness of ads in a way that aligns the goals of the advertising vendor being measured with those of the brand. Instead, last-click attribution has created a number of perverse incentives.
On the other hand, last-click means the vendors only get credit for users with a predisposition for clicking, which in turn undervalues the impact of digital advertising and hurts the brand’s overall ROI. Both diverge significantly from the advertiser's overall interest, and thus leave significant room for improvement.
Billboards: A Real-World Example
Physical billboards represent the real-world analog for how digital ad attribution should work. Measurement doesn’t involve tracking drivers who see a billboard and then immediately drive to the store, nor does it involve a competition where billboards are placed ever closer to a mall.
Instead, analysis of billboard effectiveness considers the relative, aggregate sales among exposed and unexposed consumers. Online ads ought to be measured in a similar fashion, but instead are predicated on the false notion that there is a single instance of an ad that drives the conversion, that that particular impression is signaled by a click and that the clicked ad is the only advertising responsible for driving that conversion.
To effectively measure online ads, one should consider their impact as similar to physical billboards. This means the right question should be: To what degree do the users exposed to ads outperform those not exposed, as measured by the advertiser’s KPIs? Analysis of attribution must also reflect the realities of online advertising, including bot traffic, non-viewable ads and potentially unscrupulous counterparties.
Align Attribution To Brands’ KPIs
To effectively conduct this study, brands must be willing to partner with a third party, such as a data-management platform, that will separate the eligible, targetable user pool among the total number of competing ad tech vendors, plus one control group, each with representative cross sections. Vendors should only be able to target their allocated users, and credit should only be given for conversions following viewable ads, as measured by a third-party accredited vendor.
This would drive vendors to focus on high-quality, viewable impressions – with a concern for reach and frequency – that meaningfully increase the performance of the brand, instead of optimizing to game the attribution.
In the ideal case, each vendor would be analyzed by considering how many of the users in their user pool have been exposed to viewable impressions and then converted based on full, cross-device attribution, and how many of their users have converted relative to the control. This automatically removes questions of conversions subsequently driven by retargeting and other funnel analyses.
Ultimately, online advertising can be, and often is, very effective for brands. But through the lens of inadequate analysis, any form of advertising can be made to appear underperforming. Hopefully, by considering a representative view of impacts of their advertising efforts, brands can realize the true impacts of their advertising efforts – and ultimately optimize toward those partners that truly deliver meaningful ROI.