What is incrementality in advertising?

Short answer. Incrementality is the share of your conversions that your ads actually caused, as opposed to conversions that would have happened anyway. You measure it by comparing a group that saw the ads with a comparable holdout group that did not; the difference is the incremental effect.

Definition

Singular describes incrementality as "a measure of the increase on conversions that ads have on driving the marketers desired outcome," whether that is awareness, installs or paid subscriptions. Tenjin's glossary puts the point sharply: "attribution credit is not the same as causal impact." Attribution tells you which ad a converting user touched last; incrementality tells you whether the ad made the difference.

Formula

Two numbers come out of a test, and Singular defines both:

• Lift is "the increase the ad spend caused from Group A to Group B." As Tenjin writes it: lift = (test conversion rate - control conversion rate) / control conversion rate.
• Incrementality is "the percentage of Group B which converted as a result of advertising."

Singular's own example: a control group with no ads produces 100 conversions, the exposed group 130. Lift is 30%, and incrementality is "30 ad-related conversions / 130 total conversion = 23%."

Worked example (illustration)

These are made-up round numbers. A brand splits regions into two matched groups. The exposed regions record 1,200 purchases; the holdout regions, scaled to the same size, record 1,000. Lift = 200 / 1,000 = 20%. Incrementality = 200 / 1,200, about 17%. If the platform attributed 900 purchases to the campaign, it claimed far more than the 200 the test says it caused.

What a good or bad reading tells you

Tenjin summarises the readings: positive lift "confirms that the campaign is generating genuine additional value," while "a low or near-zero incremental lift suggests the campaign is largely taking credit for organic conversions." Tenjin flags retargeting and brand campaigns as the places where attribution most overstates impact, and new channels as the place a test pays for itself before a big budget commitment.

How it is used in app, game and brand advertising

App teams use holdouts and geo tests to check channels and retargeting. Brands adding a new channel next to Meta use it to answer whether the new channel brings new customers or re-labels existing ones. Singular describes the core method: "audiences are randomly segmented into test and control cohorts." A geographic split, where whole regions are held out, is the other common design.

Common mistakes

Tenjin lists the ones we see most:

1. Too small a sample, so the test "will not reach statistical significance."
2. Too short a window, which captures noise.
3. A contaminated control group that sees the campaign through other channels.
4. Reporting a point estimate without the confidence interval.

We would add a fifth: changing creative, budget and audience at once, so nobody knows what caused the result.

Flixr's view

Our view, from our guide to incrementality testing for ad creative: run a bake-off first (native creative against repurposed, same channel, same window, only the creative changes), then a geo or audience holdout once spend justifies it, and read purchases in your own data. Platform ROAS is for running a campaign, not for judging a channel. For the channel question, see how to prove a new channel adds demand.