Key takeaways
- Google Ads Conversion Lift holds back a control group from your ads and measures the difference in conversions, so it shows what your campaigns actually caused, not what they were credited with.
- User-based studies need at least 1,000 observed conversions, a US$5,000 minimum campaign budget and a compatible conversion action. Not every account has access yet.
- Google’s November 2026 setup documentation leans hard on data quality: enhanced conversions with hashed customer data, 7 to 14 days of clean pre-study data, and no GA4-imported web conversions in user-based studies.
- Run it when a real budget decision depends on the answer. Plan for at least 14 days and aim for 90% study power.
Google Ads Conversion Lift is worth running when you’re about to make a significant budget decision and you don’t trust attributed conversions to tell you what your ads really drive. It compares people (or regions) who could see your ads with a held-back group that couldn’t, and reports the incremental difference. Here’s when a Google Ads Conversion Lift study is worth it, how to set it up and how to read it.
What is Google Ads Conversion Lift?
Conversion Lift is Google’s built-in incrementality test. It splits your audience into a treatment group that can see your ads and a control group that can’t. The gap in conversions between the two groups is the lift. That gap estimates the conversions your campaigns caused, rather than the ones they were credited with.
Google offers two versions, as described in About Conversion Lift:
| Based on users | Based on geography | |
|---|---|---|
| How groups are split | Aggregated user attributes | Sub-country regions (Google Marketing Areas) |
| Main metrics | Incremental conversions, relative lift, incremental conversion value, iCPA, iROAS | iROAS, incremental conversions, incremental conversion value, incremental cost |
| Offline data | Only with hashed customer data | Supported (aggregated offline data via your account team) |
| Readiness signal | Study power (aim for 90%+) | Feasibility status (High, Medium, Low) |
Conversion Lift ignores your attribution settings, so Google’s results guide says its numbers won’t match your standard reports. That’s the point. For the attribution side of that gap, see why GA4 and Google Ads numbers don’t match.
When is a Conversion Lift study worth running?
Run a study when the answer will change a decision worth more than the test costs. Good triggers: you’re asked to justify or cut a big Performance Max or Demand Gen budget, brand search looks suspiciously efficient, or a client wants proof that spend drives net-new sales. If nobody will act on the result, skip it.
Good reasons to run one:
- Automated campaigns you can’t fully inspect. Performance Max and Demand Gen report attributed conversions that may include people who would have bought anyway. See our guide to Performance Max controls for what you can steer before testing.
- Setting a realistic target. An incremental CPA (iCPA) is a better input to a break-even target than an attributed CPA.
Reasons to wait:
- You’re below or barely at the volume thresholds.
- You plan to change creative or audiences mid-test. Google warns that can skew results (routine bid and budget changes are fine).
- Your conversion tracking is shaky. A lift test built on broken data gives you a confident-looking wrong answer.
What do you need before you can set up Conversion Lift?
For a user-based study you need at least 1,000 observed conversions, a minimum campaign budget of US$5,000, and at least one Conversion Lift-compatible conversion action. A campaign can sit in only one lift study at a time, and access still varies by account.
Those rules come from Google’s Set up Conversion Lift based on users page, as of November 2026. The details that matter:
- Observed conversions only. Conversions from supplementary data don’t count toward the 1,000.
- The budget is in US dollars. If you bill in Canadian dollars, convert it yourself and leave headroom for exchange-rate moves.
- Supported campaigns. User-based studies support Display, Search, Video, Demand Gen, App and Performance Max. App campaigns targeting iOS and Travel Ads aren’t supported.
- One study per campaign across Brand Lift, Search Lift and Conversion Lift.
- Access. Google’s help pages still say Conversion Lift “isn’t available for all Google Ads accounts.” Check the Lift studies tab in your account; if Conversion Lift isn’t offered, ask your Google representative.
Geo-based studies have no published conversion minimum; the geo setup page shows a feasibility status instead, and Google advises against running one rated Low.
How do you prepare your conversion data for a lift test?
Get enhanced conversions working cleanly for 7 to 14 days before launch. Google’s updated setup page says user-based Conversion Lift relies on hashed customer data, so click-ID-only offline imports and GA4-imported web conversions won’t work. Fix the data first; the test can’t fix it for you.
Google expanded this part of its setup page in November 2026. The checks to work through:
- Turn on enhanced conversions for web and for leads.
- Stop relying on GCLID-only offline imports. Imports using only GCLID, GBRAID or WBRAID can’t be used, because people in the control group never clicked an ad. Send hashed customer data instead, and don’t send BRAID alongside it; Google says that blocks the customer data.
- Don’t use GA4-imported web conversions in user-based studies. They’re listed as unsupported.
- Format customer data correctly. Emails lowercased, trimmed and SHA-256 hashed. Phone numbers in E.164 format, which for Canada means starting with +1. Addresses with all four required fields.
- Upload API conversions within 24 hours and include hashed customer data on at least 30% of pings, ideally with email.
- Send dynamic conversion values and a valid currency code (CAD for most Canadian accounts) if you want iROAS.
- Keep web and app conversions in separate conversion actions. Google says mixing sources in one action drops all of its data from the study.
- Check Google tag gateway coverage. Google flags routing less than about 75% to 80% of web traffic through a first-party domain as a risk.
Advanced consent mode is listed as a requirement for EEA, UK and DMA traffic. Canada isn’t on that list, but you still need a consent setup that works under Canadian privacy law. Our guide to Google Consent Mode in Canada covers setup and testing, and our hub on conversion tracking in Canada covers consent, CASL and server-side tagging.
How do you set up a user-based Conversion Lift study?
In Google Ads, go to Campaigns, then Experiments, then the Lift studies tab. Create a Conversion Lift study based on users, pick your campaigns and conversion goals, choose a primary metric, set dates and holdback, then check study power before you create it.
- Open Campaigns > Experiments > Lift studies and click the plus button.
- Select Conversion Lift, then Based on users.
- Name the study and choose campaigns, either from the list or by uploading campaign IDs.
- Choose conversion goals. Supplementary conversions are flagged with a warning icon.
- Pick a primary focus: conversions (incremental conversions and iCPA) or conversion value (incremental value and iROAS). The other metrics still appear in the report.
- Set dates. The minimum is 7 days, but Google recommends more than 14, especially if your conversion lag is long. It says short studies with long lag can see up to a 17% drop in absolute lift.
- Set the holdback between 1% and 50%. Keep it at 30% if you’re also running Brand Lift or Search Lift.
- Review study power. Aim for 90% or higher. Between 50% and 90%, treat results as directional. If power is low, the tool suggests budget changes.
- Click Create study.
For a geo-based study, Google’s page points to Goals > Lift measurement and “Based on Geo”, with a feasibility check before saving. Once live, leave creative and audiences alone. An ended study can’t be restarted.
How do you read Conversion Lift results?
Focus on incremental conversions, iCPA or iROAS, and the certainty of lift. Google aims for 90% certainty. Results between 50% and 90% are directional, and anything under 50% is labelled “no lift,” which means the test couldn’t detect an effect, not that the ads did nothing.
The core formulas, from Google’s results guide:
- Incremental conversions = treatment conversions minus control conversions.
- iCPA = total ad spend divided by incremental conversions.
- iROAS = incremental conversion value divided by total ad spend.
Illustrative example: a campaign spends $20,000 CAD during a study and shows 1,000 attributed conversions, a $20 CPA. The lift study finds 400 incremental conversions, an iCPA of $50. If your break-even CPA is $40 (see how to calculate target CPA from your margins), the attributed number says “scale” and the incremental number says “this campaign is losing money at the margin.”
On certainty, Google’s About certainty of lift page gives these bands:
| Certainty of lift | How to treat it |
|---|---|
| 90% or higher | Very likely caused by your ads. Safe to act on. |
| 70% to 90% | Good chance, but randomness is still possible. |
| 50% to 70% | Directional only. |
| Below 50% (“no lift”) | Inconclusive. Re-test before drawing conclusions. |
Google is moving Conversion Lift to a Bayesian methodology. Under it, results can show from the fourth reporting day (processing can take up to 10 days), and you’ll see an 80% credible interval: an 80% probability the true lift falls inside that range. Google recommends waiting until the study ends before acting. For long sales cycles, geo studies offer an optional cooldown period that keeps collecting data after the test.
Does Conversion Lift work for Canadian advertisers?
Yes, but smaller Canadian accounts hit the volume limits sooner than large US ones. The 1,000-conversion and US$5,000 thresholds are the same everywhere, so a regional business may need to group campaigns or run longer to qualify. Geo-based studies also require each campaign to target a single country.
- Cross-border campaigns. If one campaign targets Canada and the US together, it isn’t eligible for the geo method. Split by country first.
- Bilingual accounts. If you run separate English and French campaigns, include both so the test measures the whole program. See our guide to bilingual Google Ads campaigns.
- Privacy. User-based lift depends on hashed customer data. Under PIPEDA and Quebec’s Law 25, make sure your privacy notice and consent cover sharing that data with Google for measurement. This is not legal advice; check with your privacy lead.
Conversion Lift checklist
- Write down the decision and the iCPA or iROAS that would change it.
- Confirm 1,000+ observed conversions, a US$5,000+ study budget and Conversion Lift in your Lift studies tab.
- Get enhanced conversions running with correctly formatted hashed data for 7 to 14 days.
- Remove GA4-imported and click-ID-only conversions from the goals you’ll measure.
- Freeze creative and audience changes for the test window.
- Plan more than 14 days and check study power is 90% or higher before launch.
- Compare iCPA or iROAS to break-even and report the certainty band.
Frequently asked questions
Is Conversion Lift free?
Google doesn’t charge a separate fee for the study, but it isn’t free. The holdback group doesn’t see your ads, so you give up some exposure during the test, and you need a campaign budget of at least US$5,000.
Can I use offline or CRM conversions in a lift study?
Yes, with conditions. User-based studies need enhanced conversions for leads with hashed customer data such as email; imports that rely only on GCLID, GBRAID or WBRAID can’t be used. Geo-based studies can use aggregated offline conversions, arranged through your Google account team.
What does “no lift” mean in Conversion Lift?
It means certainty of lift was below 50%, so Google couldn’t separate your ads’ effect from chance. It doesn’t prove the ads had no impact. Usual causes are too few conversions, a small effect or thin segments. Google recommends re-testing.
What’s the difference between Conversion Lift and a campaign experiment?
A campaign experiment compares two versions of a campaign, such as two bid strategies, and both versions show ads. Conversion Lift compares ads with no ads, so it tells you whether the campaign drives extra conversions at all. Use experiments to choose how to run a campaign, and lift studies to decide whether it’s worth running.
Sources: Google Ads Help: About Conversion Lift; Set up Conversion Lift based on users; Set up Conversion Lift based on geography; Understand your Conversion Lift based on users measurement data; Understand your Conversion Lift based on geography measurement data; About certainty of lift; About Bayesian methodology in Conversion Lift
Platform details as of November 2026.
