Glossary /
Incrementality
Efficiency & cost
Incrementality
The only honest question in marketing measurement: how many conversions would not have happened without the ad?
Attribution & Measurement
Unit Economics
Data Governance & Nomenclature
Creative & Delivery
Audiences & Targeting
Mobile & Privacy
Incrementality is the measure of conversions that were genuinely caused by a marketing activity — the lift above what would have happened anyway. It's established through controlled experiments like holdout or geo tests, not attribution math.
?
?
What is Incrementality?
Attribution assigns credit to touchpoints a converter happened to see. Incrementality asks a harder, more useful question: of the people who converted, how many converted because of the ad versus despite it? You answer it by withholding the ad from a randomized control group and comparing. This matters because a large share of platform-attributed conversions are people who would have bought anyway — loyal customers, branded searchers, retargeted carts. Counting them as ad-driven inflates ROI and rewards channels that simply intercept existing demand. The gold standard is a randomized experiment: split your audience or geographies into test and control, run the campaign on test only, and measure the difference. Geo tests and ghost-ad holdouts are the most practical designs. Done right, incrementality is the closest marketing gets to a clean causal read.
Incrementality formula
Incremental Lift = (Test Conversions − Control Conversions) ÷ Control Conversions
Incremental Lift = (Test Conversions − Control Conversions) ÷ Control Conversions. You run the campaign on a test group and withhold it from a comparable control group. The gap between the two, expressed as a percentage of the control baseline, is the lift your marketing actually caused.
Worked example
A geo holdout test: test markets saw ads, control markets did not.
Test Conversions
1,200
Control Conversions
1,000
Incremental Lift
20%
(1,200 − 1,000) ÷ 1,000 = 20%
Only 200 of the 1,200 conversions were truly caused by the ads. The other 1,000 would have happened anyway.
Incrementality formula
Incremental Lift = (Test Conversions − Control Conversions) ÷ Control Conversions
Incremental Lift = (Test Conversions − Control Conversions) ÷ Control Conversions. You run the campaign on a test group and withhold it from a comparable control group. The gap between the two, expressed as a percentage of the control baseline, is the lift your marketing actually caused.
Worked example
A geo holdout test: test markets saw ads, control markets did not.
Test Conversions
1,200
Control Conversions
1,000
Incremental Lift
20%
(1,200 − 1,000) ÷ 1,000 = 20%
Only 200 of the 1,200 conversions were truly caused by the ads. The other 1,000 would have happened anyway.
How Incrementality differs across ad platforms
Meta (Conversion Lift)
Meta's Conversion Lift uses a randomized holdout and is methodologically sound — but Meta designs the test, picks the conversion window, and reports the result. They're grading their own homework, and the grade is reliably generous.
Google (Conversion & Brand Lift)
Google offers conversion lift and brand lift studies. Same structural problem: the platform that sells the media also runs the experiment that judges the media. Useful directionally, not as an independent verdict.
Independent geo / holdout testing
A geo or holdout test you design and measure across channels — outside any single platform — is the only setup with no incentive to flatter a given channel. This is the version that survives scrutiny in a budget review.
Common Incrementality misconceptions
Platform-reported conversions are already incremental.
They're almost never incremental. Platform conversions are attributed — credited to ads people saw — and a big chunk of those people would have converted anyway. Incrementality strips out that baseline; the gap is usually large.
Running a lift study in the platform settles the question.
A platform's own lift study is better than its attribution, but the platform still controls the test design, the window, and the reporting. Independent cross-channel testing is the only way to compare channels on equal, unbiased terms.
Related Terms
Frequently Asked Questions
What is Incrementality in simple terms?
Incrementality is the number of sales your advertising actually caused — not just sales it got credit for. The test is simple: if you turned the ad off for a comparable group, how many of those conversions would disappear? That disappearing portion is the incremental part.
How is Incrementality calculated?
Why does Incrementality differ across ad platforms?
How does Clarisights report on Incrementality?
Attribution & Measurement
Unit Economics
Data Governance & Nomenclature
Creative & Delivery
Audiences & Targeting
Mobile & Privacy
Incrementality is the measure of conversions that were genuinely caused by a marketing activity — the lift above what would have happened anyway. It's established through controlled experiments like holdout or geo tests, not attribution math.
?
?
What is Incrementality?
Attribution assigns credit to touchpoints a converter happened to see. Incrementality asks a harder, more useful question: of the people who converted, how many converted because of the ad versus despite it? You answer it by withholding the ad from a randomized control group and comparing. This matters because a large share of platform-attributed conversions are people who would have bought anyway — loyal customers, branded searchers, retargeted carts. Counting them as ad-driven inflates ROI and rewards channels that simply intercept existing demand. The gold standard is a randomized experiment: split your audience or geographies into test and control, run the campaign on test only, and measure the difference. Geo tests and ghost-ad holdouts are the most practical designs. Done right, incrementality is the closest marketing gets to a clean causal read.
Incrementality formula
Incremental Lift = (Test Conversions − Control Conversions) ÷ Control Conversions
Incremental Lift = (Test Conversions − Control Conversions) ÷ Control Conversions. You run the campaign on a test group and withhold it from a comparable control group. The gap between the two, expressed as a percentage of the control baseline, is the lift your marketing actually caused.
Worked example
A geo holdout test: test markets saw ads, control markets did not.
Test Conversions
1,200
Control Conversions
1,000
Incremental Lift
20%
(1,200 − 1,000) ÷ 1,000 = 20%
Only 200 of the 1,200 conversions were truly caused by the ads. The other 1,000 would have happened anyway.
How Incrementality differs across ad platforms
Meta (Conversion Lift)
Meta's Conversion Lift uses a randomized holdout and is methodologically sound — but Meta designs the test, picks the conversion window, and reports the result. They're grading their own homework, and the grade is reliably generous.
Google (Conversion & Brand Lift)
Google offers conversion lift and brand lift studies. Same structural problem: the platform that sells the media also runs the experiment that judges the media. Useful directionally, not as an independent verdict.
Independent geo / holdout testing
A geo or holdout test you design and measure across channels — outside any single platform — is the only setup with no incentive to flatter a given channel. This is the version that survives scrutiny in a budget review.
Common Incrementality misconceptions
Platform-reported conversions are already incremental.
They're almost never incremental. Platform conversions are attributed — credited to ads people saw — and a big chunk of those people would have converted anyway. Incrementality strips out that baseline; the gap is usually large.
Running a lift study in the platform settles the question.
A platform's own lift study is better than its attribution, but the platform still controls the test design, the window, and the reporting. Independent cross-channel testing is the only way to compare channels on equal, unbiased terms.
Frequently Asked Questions
What is Incrementality in simple terms?
Incrementality is the number of sales your advertising actually caused — not just sales it got credit for. The test is simple: if you turned the ad off for a comparable group, how many of those conversions would disappear? That disappearing portion is the incremental part.
How is Incrementality calculated?
Why does Incrementality differ across ad platforms?
How does Clarisights report on Incrementality?
See Viewable Impression across every channel in one report
See Viewable Impression across every channel in one report
See viewability and vCPM across every platform—display, video, programmatic— in one normalized report, instead of reconciling vendor numbers by hand.
Book a demo