Google Ads Conversion Lift: User-Based Incrementality Testing

Causality vs. Correlation: The End of Algorithmic Attribution

Most B2B and e-commerce brands scale based on algorithmic attribution. If a user clicks an ad and buys, Google takes 100% of the credit. But what if that user was going to buy anyway? Without incrementality testing, you are likely paying a "growth tax" by serving ads to your existing, high-intent brand loyalists.

The Secret Weapon: User-Based Measurement (Not Cookie-Only Attribution)

In a world heavily impacted by iOS ATT and browser cookie deprecation, standard attribution is breaking. Google describes user-based Conversion Lift as a study type where test groups are created from aggregated user attributes, while geo-based studies use aggregated geographic data and can support offline data. The practical takeaway: the lift study is not asking "who clicked last?" It is asking whether exposed users converted at a higher rate than a comparable holdout.

How the Treatment & Control Methodology Works

Google calculates lift by splitting your target audience into randomized, mutually exclusive cohorts:

• The Treatment Group (Exposed): Users who are eligible and actively served your ads.
• The Control Group (Holdout): Users who fit your exact targeting criteria but are intentionally suppressed from seeing your ads.

By comparing the conversion rates of these two isolated groups, Google calculates the "Absolute Lift" (the raw number of net-new conversions) and the "Relative Lift" (the percentage increase over the baseline).

What Is Conversion Lift Study Power?

Study Power, also called Conversion Lift feasibility, estimates the likelihood that your experiment will produce conclusive lift results. Google says Study Power is shaped by historical campaign data, estimated lift, selected conversion actions, daily campaign budget, study duration, and traffic split or holdback percentage.

The dashboard above turns those inputs into a practical readiness score. It is not a replacement for Google's feasibility tool, but it gives you a fast gut check before you ask for a formal study.

• Budget: More media spend increases the chance that the study has enough observations to detect causal lift.
• Conversion volume: Google notes that conversion actions with more data can increase certainty.
• Duration: Longer experiments help when the average conversion lag is longer than a few days.
• Holdout percentage: Larger holdouts increase control sample size, but they also increase opportunity cost.

Choosing Holdout Size Without Burning the Business

Google allows holdback sizes from 1% to 50% for user-based Conversion Lift studies. A larger holdout gives the control group more data, which can make lift easier to detect, but it also means more eligible users are intentionally prevented from seeing your ads.

A smaller holdout lowers the opportunity cost, but it may require a longer experiment to collect enough control conversions. This is why the calculator includes a Holdout Opportunity Cost estimate: finance teams need to see the cost of clean measurement before they approve the experiment.

Which Conversion Actions Should You Measure?

Bottom-funnel actions like purchases and qualified leads matter most, but Google recommends considering upper- and mid-funnel actions when the final conversion does not have enough volume. Pageviews, lead form submissions, add-to-cart events, and other leading indicators can help reveal whether ads are creating real demand before revenue catches up.

For a high-consideration B2B funnel, that might mean measuring form submissions and sales-qualified lead movement. For ecommerce, it might mean measuring add to cart and purchase value side by side so you can see both intent lift and revenue lift.

How Long Should a Conversion Lift Study Run?

Google allows studies as short as 7 days, but typically recommends more than 14 days, especially when the business has a longer conversion lag or higher-value conversion. If your average conversion lag is 10 days, a 7-day test will likely undercount delayed conversions and make the campaign look weaker than it really is.

Use a simple rule of thumb before launch: the study duration should be long enough to capture the average lag between impression and conversion, then add extra time for the experiment to accumulate a stable sample.

Bidding, Attribution, and Creative Inputs That Change Lift

Incrementality is not only a reporting framework. It is also a diagnostic for your campaign architecture. Google notes that conversion-based or conversion-value-based bidding strategies such as Target CPA, Maximize Conversions, and Target ROAS have been associated with higher lift than manual CPC or target CPM strategies.

Google also notes that data-driven attribution is calibrated using incrementality signals. That does not make in-platform attribution the same thing as causal lift, but it does mean your bidding and attribution settings should be aligned before you judge the experiment.

• Attention: Hook the user quickly enough for the ad to be noticed.
• Branding: Introduce the brand early and repeatedly, especially in video.
• Connection: Give the audience a reason to care, not just a feature list.
• Direction: Ask for the next action clearly so lift has somewhere to go.

Reading the Report: Cost per Incremental Conversion

Use the calculator above to run your own scenarios. If your standard Google Ads dashboard claims a $45 CPA, but your Conversion Lift study reveals a Cost / Incremental Conv. of $110, your unit economics are broken. You aren't acquiring new customers; you are subsidizing organic traffic.

Official Google Ads Conversion Lift References

For implementation details, eligibility, and current setup language, use Google's own documentation as the source of truth before launching a study.

• About Conversion Lift
• Set up Conversion Lift based on users