This article was originally published here.
Incrementality testing, when set up correctly, is the most accurate way to measure the real, incremental value created by marketing activities. In March 2021, Adam Lovallo, founder of Thesis (now a Barrington Media Group Company) wrote a blog post that looked at view-through conversions versus click-through conversions for 10 US-based brands.
Ad accounts with the below criteria -- and view-through conversions at 40%+ of default conversions -- would benefit from an incrementality test, as it's likely that the real, incremental CAC from prospecting is at least 5x+ what the pixel is reporting.
- The ad account's ad sets are using an optimization window that includes 1-day view (most often 7-day click + 1-day view)
- Performance is evaluated/decisions are made using a default attribution window that includes 1-day view (usually 7-day click + 1-day view)
- View-through conversions account for a large percentage of attributed conversions from the prospecting campaigns (again, when using a default attribution window)
Today, with many changes to the social advertising landscape (iOS14.5, Facebook's switch from date-of-click to date-of-conversion attribution logic, and 28-day click attribution having disappeared, for now), he re-ran these numbers to get an updated benchmark.
For the analysis, Adam looked at 70 US-focused ad accounts loosely categorized as eCommerce. These accounts spent roughly $12m in a 30-day period. He calculated the percentage of attributed purchases using the default attribution window (7-day click + 1-day view) that were attributed as 1-day view-through conversions.
On average, 34% of conversions were attributed to the 1-day view window, and about 12 of the brands studied were even higher. In many instances, Adam reports that he has seen view-through conversions be very much validated by incrementality testing, but in this case, any one of these brands making decisions based solely on Facebook's default attribution window is probably in for a rude awakening when they run an incrementality test.