Can AI accurately predict marketing incrementality without running a randomized controlled trial?

In this episode, Pranay Piyush and Sundar Swaminathan break down a new marketing science paper introducing Predicted Incrementality by Experimentation (PI) and discuss what it means for marketers, ad platforms, and measurement teams.

They cover:

How PI predicts incrementality using historical experiments
Why the model achieved an 0.88 R²
Whether AI can reduce the need for randomized controlled trials (RCTs)
What this means for Meta's incremental attribution
Why benchmarks can be misleading
Which companies can benefit from predictive incrementality today
Why RCTs remain the gold standard

If you're a CMO, growth leader, marketing scientist, or performance marketer, this episode offers a practical breakdown of one of the most interesting developments in marketing measurement this year.