How To Measure Incremental Lift

Introduction

Incremental lift is the extra conversions a marketing activity causes, measured against a matched group that didn't get it: for illustration, if regions where your Meta ads kept running made 1,150 orders in two weeks and matched regions with the ads paused made 1,000, the lift is 15%. As a formula, lift = (conversions with the activity − conversions without) ÷ conversions without. In that example, 150 of the 1,150 orders were incremental; the rest would have happened anyway.

Why Measure Incremental Lift?

Attribution models share out credit by last-click or linear rules, whether or not a channel caused the sale. Lift asks what would have happened without the channel. It tells you:

• which channels cause sales, not just which ones show up before them;
• where spend is buying sales you'd have had anyway;
• which channels deserve more budget and which can take a cut.

The Causal Approach

You can't watch the same buyers with and without a channel at the same time, so a holdout stands in for the missing half: regions (or people) kept away from the channel, chosen to track the rest closely. The gap in conversions between the two is the channel's causal effect.

Steps to Measure Incremental Lift

Five steps for a regional holdout on your own sales data:

1. Pick one channel and one outcome, such as Meta prospecting and orders by region from your Shopify store.
2. Split regions into test and holdout. Choose regions whose orders moved together over the past few months, with enough orders in each to see a change.
3. Pause or cut the channel in the holdout regions for at least two weeks, longer if your customers take weeks to decide. Keep prices, promotions and other channels the same everywhere.
4. Compare against the weeks before. If the test regions normally make 1.15 times the holdout's orders and made 1.32 times as many during the test, the channel lifted orders by about 15% (1.32 ÷ 1.15 = 1.148).
5. Check it isn't noise. If that ratio wobbled as much from week to week before the test, the result is inconclusive, not zero. That's a rough statistical significance check; a formal test does the same with more rigour.

Interpreting Results

A lift that clearly beats the normal wobble means the channel adds sales you wouldn't otherwise have had. A lift near zero means it mostly collects credit for sales that would have happened anyway. A negative result deserves a re-run before anyone acts on it: check that the regions were well matched and nothing else changed.

Practical Use

Move budget in steps toward channels whose lift holds up, and re-test after big changes in spend, creative or season, because lift isn't fixed. For how lift fits with marketing attribution, see the guide to incrementality testing.

Conclusion

Rule of thumb: budget on lift, not on reported ROAS, because every platform grades its own homework, and read every lift against the weeks before the test.

Related Resources

Incrementality Testing: The Only Way to Know If Your Ads Actually Work

How to Run a Holdout Test on Meta Ads Without Killing Your Revenue