What is incrementality? Measuring the additional impact of advertising
Incrementality asks how much an outcome increased compared with what would have happened without an advertising intervention. Conversions attributed to an ad are not necessarily all additional conversions. Google's lift measurement explanation
The business question is whether continuing or increasing an investment creates additional outcomes. This guide distinguishes methods and explains preparation and interpretation.
How it differs from ordinary CPA and ROAS
| Approach | Question |
|---|---|
| Attributed outcomes | Which ads received credit for conversions? |
| Incrementality | Which outcomes would not have occurred without the intervention? |
| Before-and-after comparison | How did observed metrics change over time? |
Someone already planning to buy may click an ad; another person may discover the offer through that ad. Those cases have different implications for spending. This illustration does not justify automatically stopping branded or remarketing campaigns. Test your own conditions.
Better measurement is different from additional sales
On September 10, 2026, Google announced measurement updates and the general availability of Meridian GeoX. Recovering reported conversions and establishing causal business impact serve different purposes. Official announcement
More conversions recorded after a tag or import fix do not automatically mean higher actual sales. See the offline conversion guide to prepare downstream opportunity measurement.
A/B tests, lift studies, and geographic experiments
| Method | Comparison | Main decision |
|---|---|---|
| Settings A/B test | Current and changed advertising settings | Which setting to adopt |
| User-based lift study | Advertising delivery versus a withheld control | Additional outcomes from advertising |
| Geographic experiment | Different interventions across regions and control conditions | Incremental impact of a tactic or budget change |
| Marketing mix modeling | Time-series evidence and model assumptions | Estimated channel contributions and budget allocation |
Google's experiment center explanation distinguishes settings tests from lift studies. Comparing two headlines does not establish the impact of advertising versus no advertising.
Google lift measurement is not available to every account. Check the official setup guidance.
Meridian GeoX is an open-source library for geographic experiment design and analysis. It can support MMM or independent experiments. Official overview
Using a library does not automatically create the advertising intervention. Data preparation, delivery changes, and analysis remain necessary.
Prepare a study brief
Choose one spending decision and the difference that would change it.
| Check | Define |
|---|---|
| Business question | Whether the result changes continuation, expansion, or reduction |
| Outcome | Purchases, opportunities, or revenue you can record |
| Intervention | Advertising presence, budget changes, or a specific tactic |
| Control | How users or regions receive comparison conditions |
| Timing and data | Conversion delays, weekdays, and seasonality |
| Interference | Other channels, promotions, and purchases across regions |
| Feasibility | Detectable effects, spending, and acceptable operational impact |
Geographic experiments need historical daily outcomes by region. Designs changing existing budgets also need historical spending. GeoX workflow
Do not subtract raw outcomes from differently sized regions and label the result incremental. The comparison needs pre-intervention trends and an appropriate design. When evidence or statistical power is insufficient, prepare measurement before starting.
Read iCPA and iROAS carefully
iCPA = Advertising cost evaluated by the experiment / Estimated additional outcomes
iROAS = Estimated additional revenue / Advertising cost evaluated by the experiment
Align the cost scope with whether you compare advertising presence or additional spending. If estimated additional outcomes are zero or negative, do not interpret iCPA as an ordinary positive acquisition cost. Revenue-based iROAS is not a profit margin.
Review uncertainty, scope, and observed versus predicted outcomes alongside point estimates. An inconclusive difference is not proof of zero effect.
When only before-and-after data exists, report it as an observed change. Use the marketing review guide to avoid overstating causality.
An AI preparation template
Review preparation for measuring advertising incrementality.
Spending decision the result should change:
Target intervention:
Available measurement and experiment features:
Outcome definitions and records:
Historical data and its granularity:
Possible user or geographic separation:
Other ads, promotions, and sales changes:
Spending and operational constraints:
Return:
- Whether this is a settings test, incrementality study, or before-and-after review
- Additional evidence needed before experimentation
- Conditions that could invalidate the comparison
- Areas requiring specialist design or analysis
- Reasons to defer the experiment
Do not invent numbers or required budgets.
Do not present before-and-after changes as causal effects.
Do not substitute AI guesses for statistical analysis.
Start by identifying the spending decision you would change. Shoryaku AI chat can help organize questions using connected advertising and GA4 metrics. It does not design or execute lift studies or Meridian GeoX experiments for you.