Incrementality is the additional business result caused by a marketing activity. It separates outcomes created by the activity from the baseline outcomes that would have happened without it.
If 1,000 customers purchase after seeing an ad, the ad platform may claim all 1,000. If a comparable control group shows that 700 would have purchased without the ad, the incremental result is 300 purchases.
Incremental conversions and lift
Three numbers appear in most incrementality reports:
- Incremental conversions: treatment conversions minus the estimated baseline conversions
- Absolute lift: treatment conversion rate minus control conversion rate
- Relative lift: the absolute difference divided by the control conversion rate
The denominator matters. Some tools report the incremental share of treatment conversions instead of relative lift against the control. Keep the formula beside the number.
Incremental ROAS
Incremental return on ad spend, or iROAS, uses the revenue caused by the activity:
incremental revenue / marketing spend
Platform ROAS uses revenue credited under the platform's attribution rules. That can include people who were already likely to buy, especially for retargeting and branded search. iROAS removes the estimated baseline and gives budget owners a stricter view of return.
A simple example
A retailer runs paid social in a group of matched regions and withholds the campaign in the control regions.
| Treatment | Control | |
|---|---|---|
| Customers | 50,000 | 50,000 |
| Purchases | 2,750 | 2,500 |
| Conversion rate | 5.5% | 5.0% |
The campaign produced an estimated 250 incremental purchases. Absolute lift is 0.5 percentage points. Relative lift is 10%.
This example assumes the regions were comparable and nothing else changed differently between them. Real tests need pre-test matching, a power analysis, and checks for promotions, inventory changes, and overlapping campaigns.
What incrementality does not tell you
One test does not explain every touchpoint in the customer journey. It estimates the effect of a specific treatment during a specific period. Results can change when creative, audience, season, price, or channel conditions change.
Incrementality also does not replace operational reporting. Marketing teams still need daily campaign data and attribution models to inspect where conversions came from. Tests act as a causal check on those reports.
How to measure incrementality
The cleanest method is a controlled experiment. Assign comparable users or regions to treatment and control, expose only the treatment group to the activity, then compare a predefined business outcome.
Read the full incrementality testing guide for user holdouts, geo experiments, formulas, and test planning.
Teams also use marketing mix modeling to estimate contribution across a broad media portfolio. MMM can incorporate experiment results as calibration inputs, but the model itself remains an estimate based on historical variation and assumptions.
Start with first-party outcomes
A test needs reliable conversion data from both groups. If browser restrictions remove purchases unevenly across devices or channels, the observed lift can move even when the real lift does not.
Hardal's marketing measurement product collects web, mobile, and server events as first-party data before sending the permitted signal to analytics and advertising destinations. arabam.com measured 80.1% more PageView events and 22.8% more purchases after replacing browser-only delivery for critical events.
Use that outcome data to run an incrementality test on the budget decision with the largest downside. Write the decision rule before seeing the result.