Google Ads has opened Conversion Lift to Search and Performance Max campaigns as a self-serve option. Until now these tests for Search and PMax required a Google representative; advertisers can now set them up themselves in the account. The tool answers the key question: how many conversions did the ads actually create, and how many did they just take credit for. Here's who qualifies, how to set it up and how to read the results.
What Conversion Lift and incrementality mean
The usual "Conversions" column shows every purchase Google linked to your ads. But some of those people would have bought anyway, for example because they were already searching for your brand. Incrementality means only the conversions that wouldn't have happened without the ads.
Conversion Lift measures it with an experiment. Users are split into two groups: one can see ads from the selected campaigns, the other, the holdback (control) group, can't. Google then compares conversions between them. Example: the exposed group makes 120 purchases, the control group 100 under the same conditions. The ads added 20 purchases, not 120.
What changed
User-based Conversion Lift already worked for Display, Video, Demand Gen and App. Google's help page now lists Search and Performance Max too, and you no longer need a rep to start. iOS App campaigns and Travel Ads are not supported, and the tool still isn't available in every account.
Who can run a test
The bar is high. You need:
- at least 1,000 observed conversions; conversions modeled with supplementary data don't count;
- a budget of $5,000+ across the campaigns in the study;
- at least one Conversion Lift-compatible conversion action.
A campaign can be in only one study at a time: Brand Lift, Search Lift or Conversion Lift.
How to set up Conversion Lift in Google Ads
- Go to Campaigns → Experiments → Lift studies and click "+".
- Choose Conversion Lift and the "Based on users" split.
- Add campaigns from the list or by uploading a CSV of campaign IDs.
- Pick conversion goals. Choose frequent actions such as purchases or leads: rare events won't give enough data.
- Pick the main metric: conversions (iCPA) or conversion value (iROAS).
- Set the dates: 7 days minimum, Google suggests 14+, longer if the path to purchase is long.
- Set the holdback size, from 1% to 50%.
- Check the study power forecast and raise budget, duration or holdback if needed.
How to read the results
Google shows certainty from 50% to 95%. Aim for 90% or higher. Results between 50% and 90% are only directional, so don't overhaul your budget on them. The key numbers are incremental conversions, iCPA (the cost of one truly added conversion) and iROAS (the revenue the ads produced beyond what would have happened anyway).
The holdback isn't free
During the test part of your audience deliberately doesn't see ads. A bigger holdback gives more accurate results but costs more impressions and sales. Run the test when knowing whether a campaign really works is worth losing some sales.
What to test first
- Brand Search. These campaigns always look great in reports: cheap clicks, high CTR and ROAS. But someone who typed your brand often would have bought through organic results. If the control group buys almost as much, part of the brand budget can go elsewhere.
- Performance Max. PMax runs across Search, YouTube, Gmail, Display and Discover and decides where to spend. A high in-account ROAS doesn't prove PMax created those sales; a lift test checks that.
What it means for media buyers
For most solo buyers and small teams the 1,000-conversion and $5,000 thresholds put the tool out of reach. It's for e-commerce, big brands and teams running high volume to their own products. But the idea helps everyone: every platform (Google, Meta, TikTok) claims part of the same sales, so total "conversions" across dashboards exceed real purchases. Reconcile ad platform data with your tracker and actual revenue instead of trusting a single number.
FAQ
How is Conversion Lift different from attribution?
Attribution splits conversions that already happened between clicks and impressions. Conversion Lift compares groups with and without ads and shows how many conversions the ads actually caused.
How many conversions do I need for Conversion Lift?
At least 1,000 observed conversions and a $5,000+ budget across the campaigns in the study.
How long does a test run?
7 days minimum; Google recommends 14+ days, longer if the path to purchase is long.
Can I run a test without a Google rep?
Yes, for Search and Performance Max it's now self-serve if your account qualifies. Alpha and beta campaign types may still need a Google representative.
Sources: Google Ads Help, Search Engine Land, Search Engine Roundtable.