What Is Incrementality Test (Lift Study)?
An incrementality test (or lift study) is a controlled experiment that measures how much of your paid conversions would not have happened without the ad. A matched group sees the ad (exposed), a comparable group does not (holdout), and the difference in conversion rate between the two groups is the true incremental lift. Reported ROAS assumes 100 percent of tracked conversions are caused by the ad; incrementality tests routinely show the real number is 40 to 75 percent.
Formula
Exposed CVR is the conversion rate in the group that saw the ads. Holdout CVR is the conversion rate in the matched, non-exposed group. Multiply exposed group size by lift % to get incremental conversions. Divide incremental revenue by ad spend to get true incremental ROAS (iROAS), which is the only number that should drive scaling decisions.
Geo holdout on a Dhaka DTC brand
A Meta account reported 4.2x ROAS. A 4-week geo holdout paused ads in Sylhet and Rajshahi while keeping Dhaka and Chittagong live. Total blended revenue in holdout regions dropped only 22 percent versus expected baseline, not the 40 percent that reported ROAS implied. True incremental ROAS came in at 1.9x, and the account rebuilt its budget mix accordingly: 30 percent shifted to a Google Search brand-defence campaign that was measurably additive.
Benchmarks
- Typical incremental share of reported Meta conversions: 40 to 75 percent (Nielsen / Meta MMM meta-analyses, 2023 to 2025).
- Brand keyword incrementality (Google Search): 10 to 35 percent; the rest would have arrived organically.
- Minimum spend to detect a 10 percent lift with 90 percent confidence: roughly USD 15k to 40k over 3 to 4 weeks (varies with baseline CVR).
- Recommended cadence: one platform-level lift test per quarter per major channel; one geo holdout every 6 months.
Why it matters
Every ad platform reports the conversions it can claim credit for; none report the ones that would have happened anyway. That gap is the single largest source of budget waste in performance marketing. Incrementality is the only measurement discipline that answers the actual business question, which is not 'what did the platform report?' but 'what did the spend cause?'. Brands that skip it consistently overspend on retargeting and brand search and underspend on true prospecting.
Common mistakes
- 1.Running a lift test for one week. Weekly seasonality alone can swing results more than the effect being measured; minimum 3 to 4 weeks.
- 2.Using unmatched geos. Comparing Dhaka to a rural district is not a control; use synthetic control or matched-market methods.
- 3.Testing during a promo, sale, or major PR event. External shocks contaminate the holdout and produce false negatives.
- 4.Treating platform-native Conversion Lift as ground truth. It measures lift within the platform's attribution frame; it cannot see cannibalisation of Google or organic.
- 5.Killing the test the moment holdout regions show a dip. Small dips are within noise; act on statistical significance, not day 3 nerves.
Put Incrementality Test to work
Free calculators
Related services
FAQs about Incrementality Test
What is the difference between attribution and incrementality?
Attribution assigns credit for tracked conversions across touchpoints; it always adds up to 100 percent even if the ads did nothing. Incrementality measures whether the conversion would have happened without the ad. Attribution answers 'who gets credit?', incrementality answers 'did it work?'.
How long should a lift test run?
3 to 4 weeks minimum for most direct-response accounts. Shorter windows are dominated by weekly seasonality and day-of-week noise. B2B and long-consideration purchases often need 6 to 8 weeks.
What is a good incremental ROAS?
For paid social prospecting, an incremental ROAS of 1.5x to 2.5x is healthy on most DTC accounts. Brand-keyword Google Search often shows iROAS under 1.0x, which is the classic sign of paying for traffic that would have arrived organically.
Can I run a lift test with a small budget?
Below roughly USD 15k of monthly spend the statistical power to detect anything under a 25 percent lift is very low. Small accounts get more value from a 2-week creative-off holiday or a channel-off week and comparing blended MER, not a formal geo test.
Do I trust Meta's built-in Conversion Lift study?
It is a useful directional signal, but it only measures lift within Meta's own attribution frame. It cannot see if the incremental Meta sales cannibalised Google Search or organic. Pair it with a blended MER read across the same window before making budget decisions.
Related terms
Total revenue ÷ total ad spend, the blended, attribution-free ROAS.
How credit for a conversion is assigned across ad touchpoints.
Turning ads on/off by region to isolate true incremental lift vs organic baseline.
Estimated conversions Google or Meta report when direct measurement is blocked by consent or ITP.
Regression model that attributes revenue across channels using aggregated, cookieless data.
Meta's automated ecommerce campaign type with minimal manual controls.