Incremental attribution shows which purchases happened specifically because of an ad — not the ones that were going to happen anyway, with or without it.
Why this matters more since Andromeda
The old playbook was simple: find the ad with the highest ROAS and scale it. The problem is that approach often rewards ads that just happen to capture sales that were already going to close. Incremental attribution isolates the actual causal effect — it tells you which purchases wouldn’t have happened without that specific ad or placement.
Standard attribution overcounts
Standard models take credit for conversions that would have occurred with or without your ad, which means a campaign can show a 5x ROAS on standard metrics and a 1.2x on incremental metrics. That doesn’t necessarily mean the ad is bad — it usually means it’s retargeting people who were already close to buying rather than creating new demand.
Reading incremental results correctly
Start with the ad showing the highest incremental attribution and map where it sits in the customer’s journey — what came before and after that exposure. That view tells you whether the ad is actually starting the purchase decision or just nudging along something already in motion. High incremental lift is worth scaling. Low incremental lift, even with a great headline ROAS number, usually means you’d be pouring budget into demand that already existed.
How to actually measure the lift
Run controlled experiments — randomized exposure tests or geo holdouts — rather than relying on standard reporting alone. Google’s lift study documentation and Meta’s lift study setup guide both walk through the mechanics, and this HBR piece on experimentation is a good refresher on the underlying logic if it’s new to your team.
When the two numbers disagree
An ad with 5x standard ROAS and 1.2x incremental ROAS can still earn its budget — it’s just doing a different job, usually keeping conversion velocity moving among warm audiences rather than growing the buyer base. Ads with strong incremental lift are the ones actually expanding who buys from you, and those are the ones worth prioritizing when you’re deciding where the next dollar goes.
Before you scale anything, check this
- Look at incremental numbers first — scaling on standard ROAS alone risks just paying more for buyers you already had
- Check the ad’s position in the sequence and any prior exposures
- Use holdouts or randomized tests rather than assuming lift
- Weigh incremental data alongside creative and audience signals, not in isolation
Related reading on this blog:
- Claude Builds Ads Dashboards Fast
- Anatomy of a High-Converting Meta Ad
- How the 1% Advertiser Wins Consistently
- One-Click Claude Connector for Google & Meta Ads
- Claude + MakeUGC + Seedance Hyper-Realism
Check incremental numbers before every scaling decision, and your budget starts going toward ads that actually grow revenue — not just ones that show up first in the standard reporting.