Google just made Conversion Lift self-serve for Search and Performance Max. Here is how to set up a test and learn how many sales your ads truly cause.
Your Google Ads report says a campaign drove 400 conversions. Would those people have bought anyway? Brand searches, returning customers and people already halfway through a purchase all get counted, and standard reporting cannot separate them from sales your ads genuinely caused.
Google has now opened user-based Conversion Lift to Search and Performance Max campaigns on a self-serve basis. Before, those campaign types needed a Google representative to set the test up. If you spend enough to qualify, you can now measure incrementality yourself. This guide covers whether you are eligible, how to set a study up, and how to read the result.
What Conversion Lift actually measures
A lift study splits your audience into two groups. One group can see your ads. The other, a holdback group, is kept from seeing them. Google then compares conversions between the two. The gap is the number of conversions that likely would not have happened without your advertising.
That is a different question from attribution. As Google's results documentation explains, incremental conversions ignore attribution windows and count all conversions over the full study. Attribution tells you which touchpoint gets credit. Lift tells you whether the ad changed behaviour at all.
Check whether you qualify
This is not for small accounts. According to Google's setup guide, you need:
- At least 1,000 observed conversions (conversions with supplementary data do not count toward this)
- A minimum campaign budget of $5,000 USD for the study
- At least one conversion action that is compatible with Conversion Lift
- An eligible campaign created before you set up the study
Supported campaign types now include Display, Search, Video, Demand Gen, App and Performance Max. iOS-targeted app campaigns and Travel Ads are not supported, and campaigns in alpha or beta still need confirmation from your Google representative. Also, a campaign can only sit in one lift study (Brand Lift, Search Lift or Conversion Lift) at a time.
Google also notes that the tool is not available for every account. If you do not see it, your representative is the route in.
Fix your tracking first
A lift study is only as good as the conversion data feeding it. Google recommends strong data before launch: enhanced conversions for web and for leads, consent mode and advanced consent mode, and Google tag gateway. Setup diagnostics flag gaps, and you can save a study despite warnings, but this is the wrong place to cut corners.
If your tracking is patchy, a clean web analytics setup comes before any test. Otherwise you will spend $5,000 or more to measure noise.
How to set up the study, step by step
Per Google's documentation, the flow looks like this:
- In Google Ads, go to Campaigns, then Experiments, then Lift studies, and click the plus button.
- Choose Conversion Lift, then "Based on users".
- Name the study and pick your campaigns, either from a filtered list or by uploading campaign IDs in a CSV.
- Choose your conversion goals, or keep the defaults.
- Pick a focus metric: Conversions (you get incremental conversions and cost per incremental conversion) or Conversion value (incremental value and iROAS).
- Set start and end dates.
- Set the holdback, anywhere from 1% to 50%. If you run Brand Lift or Search Lift alongside, keep the 30% holdback.
- Review the study power, then create the study.
Get the design right
Study power
Study power is shown from 50% to 95%. Aim for 90% or higher. Below that, Google offers budget guidance, and results between 50% and 90% should be treated as directional only.
Duration
Studies can run for as little as 7 days, but Google recommends at least 14. Businesses with long conversion lags, such as expensive B2B purchases, need longer. Google says studies under 14 days with long lags have shown up to a 17% drop in absolute lift, which means a short test can make your ads look worse than they are.
Keep the campaign stable
You can change bids and budgets mid-study, but changing creatives or audiences can affect results. You also cannot restart a study once it has ended, and you cannot add another lift type to a running one. Plan the test before you press go.
How to read the results
Results appear on the Lift measurement page. The numbers that matter:
- Incremental conversions: treatment conversions minus control conversions.
- Relative lift: incremental conversions divided by control conversions. It can look huge when the control group has few conversions, so do not compare it across studies with different volumes.
- Cost per incremental conversion (iCPA): total spend divided by incremental conversions. Google's example: $10,000 spent with 800 incremental conversions gives an iCPA of $12.50.
- iROAS: incremental conversion value divided by ad spend. Conversion value is volatile, so expect wider confidence intervals.
Resist peeking. Cumulative results show up during the study, but Google recommends waiting until the end for the most accurate read. Some conversions are also modelled when they cannot be linked directly to an ad interaction, so better cross-device tracking improves accuracy.
What to do with the answer
Compare the iCPA with the cost per conversion your dashboard shows. If the gap is large, your reporting has been flattering the campaign, and budget is probably better spent elsewhere. If the two are close, you have evidence to scale with confidence. Either way, you now have a number you can defend to a finance team.
Brand search and retargeting are the usual suspects when results disappoint, because they harvest demand that already existed. Test them separately from prospecting campaigns so one does not hide the other. If you run both, a clear Performance Max and Search structure makes the test cleaner.
Key takeaways
- Conversion Lift is now self-serve for Search and Performance Max, but you need 1,000+ conversions and a $5,000+ campaign budget.
- Fix enhanced conversions and consent mode before you launch.
- Run at least 14 days, aim for 90% study power, and leave creatives and audiences alone.
- Judge campaigns on cost per incremental conversion, not just reported cost per conversion.
- Test brand and retargeting separately from prospecting.
If your account is smaller than the thresholds, the same thinking still applies: ask what your spend caused, not what it was credited with. Binatec's PPC management and conversion rate optimisation teams run these measurement questions for clients regularly. If you would like a second opinion on your setup, you can get a free quote.
John Doe
Supreb Information. Thanks
15 July 2014 View replies . 1