Attribution versus incrementality: did the ad create extra sales?

An ad may receive credit for a purchase that would have happened anyway. Compare dashboard ROAS with actual sales lift.

What attribution measures

Attribution assigns credit for a conversion to advertising interactions under a particular set of rules. It helps compare journeys and campaigns within a reporting system. On its own it cannot answer whether the customer would have purchased without the ad. Brand search and retargeting are particularly likely to receive credit for demand created earlier.

What an incrementality test asks

It compares the outcome for a group exposed to the campaign with a credible control group without that exposure. With sound design, the difference estimates conversions caused by advertising. Google states that Conversion Lift intentionally sets aside standard attribution rules and measures the difference in all conversions between the groups over the study period.

A hotel example

A hotel buys ads on its own name. The dashboard shows strong ROAS because guests click and book. Some might have reached the property through an organic result or typed the address anyway. A controlled lift study can estimate what demand disappears without the ads. The design must account for OTAs, seasonality, rates and any risk to brand visibility.

Design a useful experiment

Ask one question: what happens to sales without this campaign? Choose an outcome such as confirmed bookings or contribution after refunds, and a period covering a typical buying cycle. Avoid changing prices or promotions differently across groups. Determine whether users or comparable regions can be assigned to test and control, and whether there is enough volume to learn anything.

Limits of the tools

Google Conversion Lift is not available on every account. Geographic tests need enough scale, comparable regions and reliable sales data. Newer approaches such as GeoX can support experimentation but do not erase seasonality or small-sample uncertainty. A small firm may need a simpler pilot and an honest description of what it can and cannot establish.

Use both forms of measurement

Attribution can guide day-to-day optimisation; an experiment informs how to interpret it. High attributed ROAS with low lift may justify a budget review. Strong lift with modest attribution may reveal a channel influencing demand earlier than the last click suggests. Do not assume one study’s result remains constant throughout the year.

Turn evidence into a decision

Start with consistent revenue and conversion definitions. Select the largest budget decision the test could change. Lift matters if the firm will act on it by reducing low-value spend, shifting investment or improving the offer. The goal is not to discard ROAS; it is to understand what that number cannot prove.

FAQ

Does high ROAS prove extra sales?

No. ROAS uses attributed conversions; incrementality needs a credible comparison without exposure.

Can every account run Conversion Lift?

No. Google notes availability limits; check eligibility before planning a study.

Sources

Information checked: 25 September 2026. Feature access may depend on the account.