Skip to content

Incrementality

Incrementality is the part of an outcome, such as orders or revenue, that a marketing activity caused: the extra results compared with what would have happened without it. It is a causal effect, not the number of conversions credited to a channel.

Three platforms, three answers, one budget to defend. Your GA4 export already holds the answer, and the first finding is free and never leaves your browser. Get your first finding free

By , Founder & CEOUpdated 6 min read

Run the numbers for your store: the free holdout test planner.

What is Incrementality?

The IAB and IAB Europe guidelines for incremental measurement in commerce media (November 2025) define incrementality as the causal impact of marketing: the business outcomes a campaign or tactic directly produced beyond what would have occurred if no marketing had run. Lewis and Wong, in an arXiv paper, describe it as the causal effect of showing an ad to a potential customer versus not. Google's help center uses the same idea for Conversion Lift, which it calls an incrementality tool for measuring conversions directly driven by people seeing ads.

For a statistician, this is a treatment effect in the potential outcomes framework that Cunningham's textbook, Causal Inference: The Remix, lays out: for each person, the outcome with the marketing minus the outcome without it, averaged over a defined group. Only one of the two outcomes is ever observed for a person, so the other has to be estimated from a comparison group, the counterfactual. The number depends on the group, the window and what else is running. Google says a lift study calculates the value measured campaigns create on top of any other active marketing, and IAB says incrementality is not static.

Incrementality is a quantity to estimate, and an estimate is only as strong as the method behind it. IAB sorts the methods into four families: experiments (randomized control tests, holdouts or ghost ads, matched markets), model-based counterfactuals (synthetic control, propensity models), econometric models (marketing mix modeling, time-series regression) and hybrid proxies (such as platform-reported incrementality and simple multi-touch attribution). It rates causal strength from strong for experiments to weak for hybrid proxies. Attribution is a different task. GA4's help says attribution gives credit for key user actions to the ads, clicks and other factors on the path that led to them, while IAB says attribution and ROAS describe what happened and cannot say whether marketing caused it.

The methods can disagree. A 2019 Marketing Science paper on a set of Facebook advertising experiments found that observational estimates often differed from the experimental ones even after adjusting for extensive demographic and behavioral data.

For illustration, suppose a platform credits a campaign with 1,000 orders for 4,000 euros of spend, and an experiment estimates that 800 of those orders would have happened anyway. The incremental orders are 1,000 - 800 = 200. The cost per attributed order is 4,000 / 1,000 = 4 euros, and the cost per incremental order is 4,000 / 200 = 20 euros.

Why Incrementality matters for ecommerce

Budget decisions ask what spend added, and attribution reports answer a different question. Platform dashboards and GA4 show which ads and clicks were credited with an order. Incrementality asks whether the order would have happened anyway. One documented case is branded search: in eBay's field experiments, brand-keyword ads showed no measurable short-term benefit, and paid search returned only a fraction of what non-experimental estimates implied. That result belongs to eBay and its test conditions and does not carry over to every brand or channel. Both Meta and Google say their lift results are computed differently from attributed conversions, and IAB asks for a credible counterfactual before a claim of causal impact is made.

How to use Incrementality

  1. Write the question as a with-versus-without comparison for one channel or campaign: how many orders did it add over this period, compared with not running it?
  2. Take the attributed figure from the platform or GA4 as the claim to test, not as the answer.
  3. Match the method to the decision. IAB's guidance is to use methods with stronger causal foundations for high-stakes questions such as budget allocation, and to accept lighter proxies only for fast tactical decisions, with their limits acknowledged.
  4. For the strongest evidence, run a randomized holdout: an ad platform's lift tool if your account qualifies, a geo split, or a withheld segment of a list you control.
  5. Convert the result to money: cost per incremental order is spend divided by incremental orders, and incremental ROAS is incremental revenue divided by spend. Compare them with the attributed versions of the same ratios.
  6. Repeat when conditions change. IAB says incrementality is not static, and Google notes that lift results sit within a confidence interval and encourages a regular testing plan.
  7. Check the result. Pass: a randomized comparison, an interval that excludes zero, and a similar answer on a re-test. Fail: no comparison group, an interval wide enough to include both zero and large effects, or a control group that was exposed. Then report the figure as not established.

Formula

Incrementality = E[Y1 - Y0]

Incremental orders = orders with the campaign - orders without it

Cost per incremental order = spend / incremental orders

Incremental ROAS = incremental revenue / spend

Where:
- Y1: the outcome, such as orders, for a person or region if the campaign runs
- Y0: the same outcome if the campaign does not run
- E[...]: the average over a defined group and period
- incremental orders: the estimate of Y1 - Y0 added up over the audience, taken from a comparison group
- spend: the ad spend for the same campaign and period

Only one of Y1 and Y0 is observed for any unit, so the other has to be estimated from a comparison group. A randomized holdout is the most credible way to do that, though it is difficult to scale.

Common mistakes

  1. Treating attributed conversions as incremental ones. IAB says attribution and ROAS describe what happened and cannot say whether marketing caused it, so budget moved on credited orders alone can pay for orders that would have come anyway.
  2. Calling a before-and-after change or an exposed-versus-unexposed gap incrementality. Both carry selection and seasonality. In a set of Facebook experiments, observational methods often did not reproduce the experimental effects even with extensive controls.
  3. Counting simple multi-touch attribution or platform-reported incrementality as if it were an experiment. IAB lists both under hybrid proxies, rates their causal strength as weak, and names lack of causal rigor as their weakness.
  4. Reading a null result as zero incrementality. Advertising experiments are noisy: a Quarterly Journal of Economics paper by Lewis and Rao reports that measuring advertising returns is difficult, and Google says a no-lift result does not necessarily mean the ads were ineffective.
  5. Treating one result as permanent. IAB says shifts in competition, buyer behavior and tactics can change outcomes.

Frequently asked questions

  • What does incrementality mean in marketing?
    It means the additional orders or revenue a marketing activity caused, compared with what would have happened without it. IAB frames it as the causal impact of marketing, measured against what would have occurred with no marketing activity.
  • How is incrementality different from attribution?
    Attribution assigns credit for conversions that happened to ads, clicks and other touchpoints. Incrementality asks whether those conversions would have happened anyway. IAB says attribution and ROAS describe what happened and cannot say whether marketing caused it.
  • Can you measure incrementality without a holdout test?
    You can estimate it with model-based counterfactuals, econometric models such as marketing mix modeling, or hybrid proxies, but IAB rates their causal strength below that of experiments. A study of Facebook advertising experiments found observational estimates often differed from experimental ones, so treat them as directional.
  • Is incremental ROAS the same as ROAS?
    No. Google defines ROAS as attributed conversion value divided by total spend. Incremental ROAS subtracts the control group's conversion value from the treatment group's in a lift study, then divides by ad spend.

Further reading

Your platforms guess.
We run the math.

Upload a GA4 export and see what each channel caused, next to last-click, in 1–2 minutes. The read is yours to keep.

Get your first finding free

The free finding runs in your browser. Your file never leaves it. The full read is €99, refundable within 30 days, no questions asked. Prices exclude VAT.