By Jason Taylor
A key element to marketing is to accurately measure its effectiveness, so that we can, with precision, optimize and improve it. Attribution is a measurement process that consists of looking at individuals, identifying what marketing they were exposed to, and tracking that against some marketer-defined conversion event. There are many ways to attribute an impression to a conversion event, however, to do it properly involves two key concepts: multi-touch attribution and incrementality.
In the concept of multi-touch attribution, we want to look at all ad impressions that an individual might have seen. This includes different types of media like display, audio, and video as well as impressions across all campaigns running during that time period. If we do not do this, it’s impossible to truly understand whether or not the advertising is working.
An individual can be exposed to two campaigns seeing one or more impressions from each; if we measure those separately a single purchase will be credited to both campaigns resulting in double counting. We also want to spread credit across the impressions to avoid overemphasizing lower funnel activity like search or retargeting, which can tend to happen nearer to the purchase.
One way that can be used is equal weight to distribute the credit across all campaigns for each conversion event. If an individual makes a purchase, we look back at all the impressions that they have seen over the last 14 days and give credit to all those impressions equally. So if the individual made a $5 purchase and they saw five impressions, each impression would get $1 credit. This multi-touch approach gives fair credit across different types of marketing and does not double count a single conversion event across multiple campaigns.
The second key concept is incrementality. Let’s say our conversion event of interest is an online purchase. Individuals buy things for many different reasons: they might like the color, they may buy something out of habit, or that purchase could have been influenced by marketing. To measure the effect of advertising, we want to isolate the effect of the advertising from other factors. We need to understand the incremental effect of marketing on our purchase, or said another way, if we had not run the marketing what would our sales have been?
A great way to measure incrementality is a methodology known as Ghost Bidding. It is a test/control setup, where we randomly choose individuals that qualify for a campaign as the test group, but we do not win the bid. This means our test and control groups are made up of the same audience, they are both part of our bidding network, and both qualify for the bid. We then look at the difference between our control group conversion rate and that of our test group, which tells us the incremental value of the marketing.
If we put those two things together, we can properly attribute conversions to impressions and measure the incremental effect of our paid media. This gives marketers an accurate picture of what their media campaigns accomplished, and allows for informed decisions on future marketing campaigns. Without this we cannot truly evaluate effectiveness – we could just be wasting media dollars on people who would have purchased without the influence of marketing.