Every Monday our incrementality model runs against every client account. It asks
By Youval Peltier · November 20, 2026 · Curated by George's Blog
Every Monday our incrementality model runs against every client account. It asks one question about every search term: did spend move purchase share, or did it just catch the share that would have happened anyway?
The hardest terms to call are the ones with a ROAS above 10 on a small branded budget. The reports say they’re working. The model often disagrees.
Here’s why.
When a branded campaign maxes out early each day, two things are true:
→ The clicks it caught were the easiest conversions of the day
→ The clicks it never served were the ones worth measuring
The ROAS report shows the first group. The model watches purchase share on the term, and purchase share tells a different story.
When the model flags a term, our ad managers run the controlled cut: a share of the budget off for four to six weeks, every other variable steady, and the organic purchase share chart overlaid on top. If organic share holds, the ad spend was buying attribution. If it drops, the spend was incremental.
Across the client accounts we’re running this on, roughly 10 to 30% of ad budgets caught sales that were going to happen anyway. On a $1M ad program, that is $100K to $300K pointed at the wrong place.
That’s the question we built a model to ask. Every term, every account, every Monday.