Measuring the true impact of digital ads is challenging because ad targeting means the people who see ads are very different from those who don’t.
Randomized controlled trials (RCTs) are the best way to measure ad impact, but they are costly because you have to hold back potential customers from seeing ads to create control groups.
The researchers tested a model that combines a small number of RCT results with standard industry metrics like last-click conversion counts to predict ad impact.
They found proxy metrics like last-click tend to systematically over- or under-estimate ad impact, but in a predictable way.
This means advertisers can use just a few RCTs to calibrate their standard metrics and then accurately predict ad impact for campaigns without doing full RCTs.
Advertisers likely already have usable data from proxy metrics, so they may not need extensive new RCTs to make good predictions.
Full article :
