Lift

Lift is the change in an outcome that your marketing actually caused, measured by comparing a group that was exposed to it against a comparable group that was not. The unexposed group is the control, or holdout — a slice of otherwise-similar people you deliberately withhold the marketing from — and the difference in outcomes between the two groups is the lift. If the exposed group converts at a higher rate than the control, the difference is conversion lift; if brand awareness or favorability rises more in the exposed group, that is brand lift; if purchases climb, that is sales lift. The defining feature is the comparison. Lift is not what a tracker credits or what a model assigns after the fact — it is a measured difference against a counterfactual, an estimate of what would have happened without the marketing.

Lift matters because it answers the only question that justifies spend: did the marketing make a difference. Plenty of conversions credited to a channel would have happened anyway — the customer was already going to buy, and the ad merely got the click on the way. Lift strips those out by holding up a control that shows the baseline, so you see the extra outcomes the marketing genuinely produced on top of what would have occurred regardless. That makes lift the honest measure of effect, and it is why lift studies underpin serious decisions about whether a channel, a campaign, or a whole budget is worth it. A campaign can look busy in attribution reports and still show little or no lift, which is exactly the case worth knowing about.

Lift and attribution answer different questions, and treating them as the same is a costly mistake. Attribution assigns credit for conversions that were tracked to the touchpoints along the path — it distributes existing, observed conversions among channels. Lift measures how many conversions the marketing caused that would not have happened otherwise, by comparing exposed and unexposed groups. Attribution can hand a channel a heap of credit for demand it merely intercepted; lift asks whether that channel actually created demand. A branded-search or retargeting campaign often wins big under attribution while showing modest lift, because it captures people who were already going to convert. So attribution tells you which tracked touch preceded the sale, while lift tells you whether the touch was the reason the sale happened at all.

The two are complementary rather than rival, if you keep their jobs straight. Attribution is cheap, continuous, and useful for tactical routing — deciding which creative or keyword to lean into day to day. Lift is more expensive and periodic, because it requires withholding marketing from a control, but it is the truer read on causal impact and the right tool for big allocation calls. Lift is also the concept at the heart of incrementality testing, which is the discipline of designing controlled experiments — holdouts, geo splits, randomized exposure — specifically to measure lift. So incrementality is the practice, and lift is the number it produces. Use attribution to steer within a channel and lift to decide whether the channel deserves the budget in the first place.

Measure lift with a genuine control. The whole method rests on the exposed and unexposed groups being comparable in every way except the marketing, so randomize exposure where you can, or use well-matched geo splits where you cannot. Pick the outcome that matters — conversions, sales, or a specific brand measure — and size the test so the difference you care about is detectable rather than lost in noise. Run it long enough to capture the real effect, especially for purchases that take time. Report lift as the difference between exposed and control on that outcome, with a plain statement of confidence, and use it for the decisions that deserve the cost of a holdout: whether a channel or campaign is genuinely worth the spend.

The failures usually break the comparison. A control group that differs from the exposed group in some other way — geography, seasonality, audience quality — produces a difference that is not lift but confounding. Too small a test cannot detect real lift and gets waved away as null when the effect was simply undetectable. Confusing lift with attribution credit leads teams to think a high-credit channel has high impact when its measured lift is thin. And declaring lift from a study with no control at all — just a before-and-after — mistakes a trend for a causal effect. Protect the control, size the test, measure the outcome that matters, and keep lift separate from the credit attribution hands out.

Lift — the incremental outcome a campaign causes over an unexposed control — is the causal measure at the heart of incrementality testing, distinct from the credit that attribution distributes.

Etymology: source.

Search interest for this term over the last five years:

View interest-over-time on Google Trends →

Areas of marketing where lift is a core concern: