Quick answer
Incrementality is the additional outcome (conversions, revenue, sign-ups) caused by a marketing activity, measured as the difference between a group exposed to it and a comparable group that was not.
Formula: Lift = (exposed rate − holdout rate) / holdout rate; incremental conversions = (exposed rate − holdout rate) × exposed users
What is incrementality?
Incrementality is the part of a result your marketing actually caused: the conversions, sign-ups or revenue that would not have happened without it. It is measured by comparing a group that saw the activity with a comparable group that did not.
A retargeting campaign may show 500 attributed purchases. If 400 of those people would have bought anyway, its incremental effect is 100.
Incrementality vs attribution
| Attribution | Incrementality | |
|---|---|---|
| Question | Which touches came before the conversion, and how should credit be split? | Would the conversion have happened without this activity? |
| Method | Rules or models over user paths | A controlled comparison with a holdout |
| Output | Credited conversions per channel | Lift and incremental conversions |
| Typical bias | Over-credits channels that reach people already about to buy | Needs a clean holdout and enough volume |
They answer different questions; use attribution for day-to-day comparisons and incrementality tests to calibrate how much to trust it.
A sample holdout test
A retargeting campaign is shown to 100,000 users; 10,000 randomly chosen users are held out (sample numbers, $25,000 spend, $50 average order value):
| Group | Users | Converted | Rate |
|---|---|---|---|
| Exposed | 100,000 | 3,000 | 3.0% |
| Holdout | 10,000 | 240 | 2.4% |
- Lift = (3.0% − 2.4%) / 2.4% = 25%
- Incremental conversions = (3.0% − 2.4%) × 100,000 = 600
- Incremental revenue = 600 × $50 = $30,000
- Incremental ROAS = $30,000 / $25,000 = 1.2
If the ad platform credited itself with all 3,000 purchases, it would report a ROAS of 3,000 × $50 / $25,000 = 6.0; even half that credit (1,500 purchases) would show 3.0. The incremental view says each dollar returned $1.20.
Ways to test
- User holdout: randomly exclude a share of the audience. Needs control over who is targeted.
- Geo test: turn the campaign off in matched regions and compare.
- On/off (time-based): pause the channel for a period; weaker, because other things change over time.
- Platform lift studies: some ad platforms run their own conversion-lift tests.
Pitfalls
- Holdout too small. Small differences in conversion rates need large groups; plan the size with minimum detectable effect in mind.
- Non-random holdouts. Comparing people who did not see an ad because they were not targeted is selection bias, not a holdout.
- Contamination. Holdout users see the same message through another channel.
- Too short. Effects that take weeks (brand search, word of mouth) are missed.
What TapCub can do
TapCub's attribution analysis credits conversions to channels, UTM sources, campaigns, referrers or landing pages using sessions in a lookback window (30 days by default). It does not run holdout tests or control ad delivery. For a geo or on/off test, use annotations to mark the test period and compare the goal's conversion rate by country, or before and after, with the period comparison. For a user holdout run in your ad platform, send the conversion as a server-side event with the user's group as a property, and compare the two groups' rates in a segment-based analysis.
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.
Growth Strategist
Explores acquisition channels, activation, and conversion experiments, turning data insights into actionable growth hypotheses and evaluation methods.