Google Ads Conversion Lift measures how many conversions your campaigns actually caused by withholding ads from a random control group and comparing it with people who could see them. As of October 2026, user-based studies cover Search and Performance Max, with a $5,000 minimum budget and at least 1,000 observed conversions.
This guide is part of the Google Ads hub. It separates what Google confirms from what the press reported, then covers eligibility, study types, setup, reading results, and how lift relates to attribution and Performance Max experiments.
How does Google Ads Conversion Lift measure incrementality?
Google Ads Conversion Lift measures incrementality by splitting an audience into a treatment group that can see your ads and a control group that cannot. The difference in conversions between the two groups is the lift: the conversions your campaigns caused that would not have happened without them during the study period.
Incrementality is the share of results that exists only because of the ads. Google's About Conversion Lift page calls the tool "an incrementality tool" and describes the method as a controlled experiment that measures "the causal impact of ads."
The study reports a small set of metrics:
- Incremental conversions (Absolute Lift): treatment conversions minus control conversions.
- iCPA: total spend divided by incremental conversions, the cost of one conversion that would not have happened otherwise.
- iROAS: incremental conversion value divided by spend, reported only when your conversion actions carry values.
- Relative lift: incremental conversions divided by control conversions.
Google's user-based reporting guide gives an example: $10,000 in spend and 800 incremental conversions is an iCPA of $12.50. It also warns that relative lift "can be misinterpreted when comparing across different studies," so compare studies on iCPA or iROAS instead.
Who can run a Conversion Lift study, and what does it need?
A user-based Conversion Lift study needs a minimum campaign budget of $5,000 USD, at least 1,000 observed conversions and at least one conversion action that is compatible with lift. Google allows studies of 7 days but recommends at least 14. Search and Performance Max became self-serve for eligible advertisers in October 2026, according to Search Engine Land.
Google's user-based setup article lists these requirements, as of October 2026:
- Budget: "a minimum campaign budget of $5,000 USD."
- Conversions: at least 1,000 observed conversions; supplementary data such as modeled store visits does not count.
- Duration: 7 days minimum, 14 or more recommended, with "up to a 17% drop in Absolute Lift" for long-lag studies under 14 days.
- Holdback: between 1% and 50% of users. If Brand Lift or Search Lift runs on the same campaigns, the holdback must be 30%.
- One study per campaign: a campaign can sit in only one Brand Lift, Search Lift or Conversion Lift study at a time.
The holdback is the real trade-off. A larger holdback gives a cleaner answer sooner, but every user in the control group misses your ads. A small one costs less but needs more time. Before launch, Google suggests a holdback size and shows study power from 50% to 95%.
What changed for Search and Performance Max in October 2026?
Search and Performance Max now appear next to Display, Video, Demand Gen and App campaigns as supported types for user-based Conversion Lift. Search Engine Land reported on October 7, 2026 that eligible advertisers can set these studies up themselves, where Search and Performance Max previously required a Google representative.
Here is what each source says, as of October 8, 2026:
- Officially confirmed. Google's user-based setup article lists Display, Search, Video, Demand Gen, App Campaigns and Performance Max as eligible, and excludes iOS-targeted App campaigns and Travel Ads. The same page still says "Conversion Lift isn't available for all Google Ads accounts" and points to your account representative.
- Reported. Search Engine Land (Anu Adegbola, October 7, 2026) reports self-serve access for eligible advertisers and notes that alpha and beta campaign types may still need a representative.
- Context. PPC Land reported in November 2025 that Google cut the minimum for incrementality tests from close to $100,000 to $5,000, while Search, Shopping and Performance Max still required a representative at that time.
No separate Google announcement of the change was found at the time of writing. The practical test is your own account: open "Lift measurement" under Goals and check whether Search or Performance Max campaigns can be selected.
User-based or geo-based lift: which study fits your account?
User-based lift splits people into exposed and control groups and suits online conversions measured by the Google tag. Geo-based lift splits regions inside one country, supports offline conversion data and Shopping campaigns, and reports iROAS with incremental cost. Most Search and Performance Max advertisers start with user-based, the type the October 2026 change opened up.
Geo-based studies use Google Marketing Areas, which are sub-country regions that Google's geo-based setup article says are built to limit people traveling between treatment and control regions.
| Criterion | User-based Conversion Lift | Geo-based Conversion Lift |
|---|---|---|
| How groups are split | Aggregated user attributes | Google Marketing Areas within one country |
| Campaign types listed | Display, Search, Video, Demand Gen, App, Performance Max | App, Demand Gen, Display, Video, Search, Shopping, Performance Max |
| Offline and store data | Modeled supplementary conversions, such as store visits | Offline conversions aggregated to ZIP or city level |
| Main metrics | Incremental conversions, iCPA, iROAS, relative lift | iROAS, incremental conversions, incremental conversion value, incremental cost |
| Planning signal | Study power from 50% to 95% | Feasibility rated High, Medium or Low |
| Best fit | Online conversions tracked in Google Ads | Store sales, CRM revenue or markets where user splits are hard |
The geo-based page also points to your Google account representative and says only holdback and go-dark designs run "in this beta."
How to set up a user-based Conversion Lift study
Setting up a user-based Conversion Lift study takes six decisions: the campaigns, the conversion actions, the dates, the holdback, a check of study power, and a freeze on major changes. Each choice affects whether the study reaches a usable certainty, so settle them before launch rather than after the first results appear.
The seven setup steps
The steps below follow Google's user-based setup article.
- Fix your tracking first. Google recommends enhanced conversions, consent mode and Google tag gateway before launch, and runs diagnostics during setup.
- Open Lift measurement. In Google Ads, go to Goals, then "Lift measurement," select the plus button and choose Conversion Lift "Based on users."
- Select the campaigns. Group campaigns that answer one business question, such as all brand Search or all Performance Max, and check that none sits in another lift study.
- Pick frequent conversion actions. Choose actions that happen often enough to produce a result; a rare final sale may need a more frequent step next to it.
- Set dates and holdback. Plan at least 14 days, longer for a long conversion lag, and accept or adjust the suggested holdback.
- Read the study power before saving. Below 90%, Google gives budget guidance; raise budget, extend dates or add campaigns until the estimate is acceptable.
- Freeze big changes during the study. Google advises changing only routine items such as bids and budgets, and a stopped study cannot be restarted.
How do you read incremental conversions, iCPA and certainty of lift?
Read a Conversion Lift result in two steps: first the certainty of lift, then the iCPA or iROAS. Certainty tells you whether the lift is likely real, and Google aims for 90%. The incremental cost metric tells you whether that lift is worth the spend, by comparing it with your break-even cost per conversion or return.
Google's certainty of lift article defines certainty as 1 minus the p-value, rounded down in 5% increments. At 90% or more there is "a very good chance that your results were caused by your ads." Between 50% and 70%, Google says to use results "directionally." Below 50%, the result shows as "No lift." Under Bayesian methodology, Google's methodology page says the true lift has "an 80% probability of being within the Credible Interval," and results start to show from the 4th reporting day. Google still recommends waiting until the study ends.
Worked example and the Lift Readout Grid
Worked example (hypothetical numbers). A Performance Max campaign spends $20,000 during a study and reports 1,000 attributed conversions, an attributed CPA of $20. The study finds 400 incremental conversions at 90% certainty. The iCPA is $20,000 / 400 = $50. If a conversion is worth $60 in margin, the campaign is profitable on incremental terms, even though it costs 2.5 times what the attributed CPA suggested.
The Lift Readout Grid below maps each certainty band and iCPA outcome to a budget action.
| Certainty of lift | iCPA below break-even CPA | iCPA above break-even CPA |
|---|---|---|
| 90% or more | Keep or scale spend; use iCPA to set your target | Cut spend or rework the campaign, then retest |
| 70% to 90% | Keep spend; retest before a large budget increase | Reduce spend in steps and retest with a longer study |
| 50% to 70% | Treat as directional; do not change budgets on it alone | Treat as directional; retest with a larger holdback |
| No lift (under 50%) | Inconclusive; rerun with more volume or time | Inconclusive; check tracking, then rerun |
If your account bids on value, the same logic applies to iROAS against your break-even ROAS, which our guide to Target ROAS shows how to calculate.
Conversion Lift vs attribution vs Performance Max experiments: which answers your question?
Conversion Lift, attribution and Performance Max experiments answer different questions. Attribution divides conversion credit across ad interactions every day. Performance Max experiments compare two setups. Conversion Lift asks whether the spend created conversions that would not have happened at all. Most accounts need all three, at different moments.
Google's data-driven attribution credits the interactions that predict a sale by comparing converting and non-converting paths. It never holds back ads, so it cannot show what happens without them. In a 2020 post, Google's John Chen called attribution "best for day-to-day, always-on measurement" and said data-driven attribution is "trained on and validated against incrementality experiments." Our guide to attribution models covers the trade-offs.
Performance Max experiments include uplift, upgrade and optimization tests, such as Performance Max asset testing.
| Criterion | Conversion Lift | Data-driven attribution | Performance Max experiments |
|---|---|---|---|
| Question answered | Did the ads cause extra conversions? | Which interactions get credit? | Which setup performs better? |
| Method | Holdout group sees no ads | Model of conversion paths | Traffic split between two setups |
| Cadence | One study, at least 7 days | Always on | One experiment over a set period |
| Output | Incremental conversions, iCPA, iROAS | Credited conversions per campaign | Difference between arms |
| Typical use | Budget levels, brand Search, targets | Daily bidding and reporting | Assets, settings, campaign upgrades |
Common mistakes with Google Ads Conversion Lift
The most common Google Ads Conversion Lift mistakes come from planning: a study that is too short or too small to reach a usable certainty, conversion actions that fire too rarely, and changes made while the study runs. Each one either wastes the holdback or produces a result that cannot support a budget decision.
- Reading a No lift result as proof the ads do not work. Google says results under 50% certainty are reported as No lift. Rerun the study with more volume or time before cutting spend, as the Lift Readout Grid recommends.
- Running fewer than 14 days with a long conversion lag. Google's user-based setup article reports up to a 17% drop in Absolute Lift for such studies (as of October 2026). Plan the length around your time to convert.
- Comparing relative lift across studies. Google warns that relative lift can mislead across studies. Compare iCPA or iROAS instead.
- Changing creatives or audiences mid-study. Google advises changing only bids and budgets. Schedule launches before or after the study.
- Replacing attribution with one study. A lift study covers one period and one set of campaigns. Use it to calibrate targets and budgets, and plan the study spend with our guide to Google Ads cost.
