How should I think about incrementality testing for a D2C brand?

Incrementality testing measures the sales your advertising actually caused that wouldn't have happened otherwise-not just the sales that happened to pass through an ad on the way to checkout. This is the distinction that matters most, because platform dashboards conflate presence with causation.

The cleanest version for most D2C brands is a geo holdout: pick matched cities where ads keep running and comparable cities where they stop entirely for four weeks. The cities without ads show you what would have happened anyway. The difference between them is incremental revenue.

Why this matters in practice: a channel reporting 4.0x ROAS can actually be sitting at 1.7x once you account for sales that were coming regardless. That gap is not a small correction-it's often the difference between doubling a budget and holding it flat. A 4.0x channel looks like a money printer. A 1.7x channel at a 60% margin is barely breaking even, and the test tells you which decision to actually make.

The test is straightforward to run: split six to eight comparable cities into two matched groups, keep one group's ads on and pause them entirely in the other, run it for four weeks, then calculate the difference in revenue between the groups. The math is simple, but the execution matters-contamination from email or influencer campaigns reaching your "dark" cities, too short a window, or seasonality events hitting one region will all break the result quietly.

Start with the channel you least believe in. That's where the gap between reported and real ROAS is usually widest, and where one test can save you the most wasted spend.