Incrementality testing
How to measure whether your ads caused the sale, or just tagged a sale you were going to get anyway.
Incrementality is the gold-standard measurement of ad effectiveness. It answers: "If I turned this ad off, would the revenue go away?"
The method
Geo-based holdout. We randomly split your target regions into test and holdout. Test regions see ads as normal; holdout regions see no ads from the channel being tested. After 2-4 weeks, we compare revenue.
Minimum requirements
- Sample size: at least $50k in test-period spend for statistical power at 95% confidence.
- Duration: 14 days minimum, 28 days recommended.
- Geography: need at least 8 addressable regions (US state-level works for most brands).
What you learn
Most brands discover one or two channels where incrementality is 40-60% of what last-click attribution claimed. That gap is your overpaid-for revenue.
The flipside
You also often discover a channel that was undercredited. Incremental lift bigger than reported ROAS means the channel is driving demand that other channels then capture. And pausing it would be expensive.