Marketing attribution is a powerful tool for evaluating ad tech vendors. Attribution helps you compare which vendors contributed to a conversion, and in what way: how many touchpoints did it take for a user to convert? How many impressions were garnered over how long a time period? How many net conversions did each vendor generate?
All of this information is important to deciding how and where to spend media budget. The number at the bottom of your report is going to say that Vendor X performed better than Vendor Y. But it’s also the product of a set of algorithms, not human analysis. Numbers at the bottom of a report tell a story, but not the whole thing. And like any data set, attribution results can be made to tell different stories depending on the narrator. Individual channel owners with a stake in the answers may interpret straight numbers in varying, and self-serving, ways.
As a result, it’s critical to validate your attribution reports with observed data. This is also known as a gut check. By adding the human perspective to the data models, you can ensure that your results intuitively make sense, and that they aren’t skewed by different channel owners’ agendas.
Three types of analysis are particularly effective for attribution validation: frequency analysis, post-source analysis, and time analysis.
1. Frequency analysis. Let’s say that your attribution tool reports that Vendor A took 100 impressions to convert one user, while Vendor B took only ten. At first glance, both converted the same number of users. If you’re paying by impression, however, it’s quickly clear that Vendor A is more expensive. Your attribution solution will serve up both of those data points.
But deeper analysis of the frequency of impressions yields additional insight. Vendor A is essentially beating users into submission – not a good look for your brand. Vendor B, on the other hand, is hitting higher quality traffic with a higher propensity to convert and greater interest in your offering. This makes Vendor B a much better choice at scale for your brand, your budget, and your conversion rates.
2. Post-source analysis. All touchpoints are not created equal. Vendor A and Vendor B may both report that a user clicked a display ad. But what happened next? Say the next most common touchpoint for Vendor A was another display impression, while Vendor B’s was a click to your website. The former is a push tactic, the latter a pull tactic. Vendor A’s user still isn’t actively engaging with your brand, while Vendor B’s user is actively reaching out and moving through the funnel, indicating that Vendor B’s approach garners greater influence on the user. Your attribution solution should tell you that Vendor B is performing better. The post-source analysis adds the “why.”
3. Time analysis. An in-depth analysis of observed time data can add another dimension to the “why” of vendor performance and comparison. Your attribution solution reports that Vendor B is better – but how much better, and with what impact to the business? Look at your time-to-order (TTO). With Vendor A, it took a user 30 days to convert. With Vendor B, it only took three. That’s 27 days in which you weren’t advertising for your competition, or spending money on a user who proves to be an undesirable lead (no one wants to spend $50 to get someone to buy a $5 widget.)
Shorter TTO also makes testing easier. Vendor A not only takes a month to convert a user, but you can’t analyze the results of your media placement for a month in order to optimize. As a result, you’re getting hit twice: paying too much per conversion, and losing the ability to adjust mid-stream for greater efficacy. Again, your attribution report will make it clear that Vendor B is better. But validating that result against the observed data paints a clearer picture of what “better” really means.
Attribution is a powerful tool in the media practitioner’s toolbox, but it’s only useful if you can trust the data. That starts with a platform that can crunch the right numbers, in the right way, to meet your business objectives. With that in place, bring back the humans. A thoughtful analysis of your observed data will add the context to make attribution data truly insightful and actionable.