Conversion Lift Tests

A conversion lift test is an experiment run by the ad platform itself to measure how many conversions your ads caused. Google describes Conversion Lift in Google Ads as an incrementality tool that estimates how many conversions your ads directly drive, and says it is not available for all accounts, so you contact your Google account representative. Meta describes its Conversion Lift as a type of lift test where one group has the opportunity to see your ads and another does not, set up with a Meta account representative. Neither is a button every advertiser can press today. Google's page links to separate guides on comparing lift types, setting up each study type, and the methodology behind the results.

The method is a randomized split. Google divides the audience into a treatment group, who see your ads, and a control group, who do not, and calls the difference in conversions the lift, the increase in conversions caused by the presence of the ad. It offers two designs: user-based, built from aggregated user attributes, and geography-based, built from aggregated geographic data and supporting offline data. Meta's page says tests that begin measuring standard web events on or after October 1, 2021 require the Conversions API. Because the platform decides who is held out, it can keep the control group clean in a way an advertiser cannot. Tell your account representative which conversion event matters most, since the metrics you can see depend on conversion volume.

Results are reported as incremental numbers. For user-based Google studies, you can see incremental conversions, relative lift, incremental conversion value, incremental cost per action, and incremental return on ad spend, with the return figure appearing only if you assign values to conversion actions. Geo-based studies report incremental ROAS, conversions, value, and cost. The metrics available depend on your conversion volume and how the groups are split. Meta reports a confidence percentage, and says that for lift tests 90 percent or higher represents a statistically reliable result. Conversion volume and test length decide whether you reach it. Do not change targeting or large budget levels while the test runs, as changes can blur the difference between the groups.

There are two limits to remember. First, the platform grades its own work: it runs the test, controls the groups, and reports the result. That is useful, but it is not independent. Second, a platform lift test covers only that platform. If Google ads and Meta ads both lead to the same purchase, each test measures its own channel in isolation, so cross-channel overlap stays invisible. Another mistake is running a test with too little budget or too few expected conversions, which produces a wide, inconclusive range. Changes to targeting or budget during the test can also undermine it. Use the result to set a target cost per action for the channel, and recheck it after a major creative or audience change.

Run a conversion lift test before a major budget decision, such as doubling spend on a channel whose attributed results look strong. Define the conversion event and the minimum lift that would justify the budget before the test starts. After it ends, compare incremental cost per action with the cost per action in your normal reports. The ratio between them tells you how much of the attributed result the ads actually caused, and you can use that ratio to adjust targets for future spend. Remember that a lift test result applies to that campaign, period, and audience, and may not carry over to a new offer.