Conversion Lift in Google Ads: Users vs Geo, Setup and Results

Everything here comes from Google Ads Help: About Conversion Lift, Comparing lift types, the setup pages for users and geography, the results pages for users and geography, Bayesian methodology, certainty of lift and Implement campaigns for geo experiments. Checked on 12 November 2026. For testing a campaign change on a split of traffic, see Google Ads experiments; for brand metrics such as ad recall, see Brand Lift.

What Conversion Lift measures

Google calls Conversion Lift an incrementality tool that helps you measure the number of purchases, site visits, and any other conversions directly driven by people seeing your ads. It splits your audience into a treatment group that sees your ads and a control group that does not, and the difference in conversions between the two is the lift.

That is a different question from normal conversion reporting. Attributed conversions in Google Ads follow each action's attribution settings, such as a 7-day click window (see conversion tracking). Conversion Lift intentionally ignores the standard conversion tracking settings and attribution rules and counts the conversions that would not have happened without the ads.

Like Brand Lift, it is not open to every account: To use Conversion Lift, contact your Google account representative.

Based on users vs based on geography

Based on users

• Split: users held back from your ads vs users who see them.
• Breakdowns: conversion action and category, age, gender, country.
• Offline: store sales and offline leads through modeled supplementary conversions.
• Budget: set from your historical conversion data, minimum $5,000.
• Metrics: incremental conversions, relative lift, iCPA, incremental value, iROAS.

Based on geography

• Split: regions that show your ads vs regions that do not.
• Breakdowns: no demographic slices.
• Offline: your own offline conversions aggregated to ZIP or city level.
• Budget: usually higher than users-based.
• Metrics: iROAS, incremental conversions, value and cost.

Google's comparison page sums up the trade-off. Users-based studies give detailed reports but rely mostly on opt-in traffic, so they can be affected by measurement gaps. Geo studies are more transparent and can measure across channels and offline, but the split is noisier.

Which campaign types? Google's pages differ here. The comparison page lists users-based studies for Video, Discovery and Demand Gen, with App, Display, Search, Shopping and Performance Max through your account representative. The users setup page lists Display, Search, Video, Demand Gen, App and Performance Max as supported, but not App campaigns targeting iOS or Travel Ads. Geography-based studies cover App, Demand Gen, Discovery, Display, Video, Search, Shopping and Performance Max.

Requirements for a users-based study

• Conversions: at least 1,000 observed conversions. Conversions with supplementary (modeled) data do not count towards it.
• Budget: a minimum campaign budget of $5,000 USD.
• Conversion action: at least one conversion action compatible with Conversion Lift.
• One study per campaign: a campaign can be in only one Brand Lift, Search Lift or Conversion Lift study at a time.

To strengthen the data, Google recommends setting up enhanced conversions for web and for leads, consent mode (ideally advanced consent mode) and Google tag gateway before launch. Setup runs automatic diagnostics; you can save a study despite warnings, but Google advises fixing them first. Our enhanced conversions section covers the first of these.

Study power

Study power, also called feasibility, estimates how certain your result will be. It is shown as a range from 50% to 95% in 5% steps and depends on the conversion actions you pick, daily budget, study duration, holdback percentage, the account's history and the expected lift. Below 90%, Google shows budget guidance to reach 90%. Results between 50% and 90% are directional.

Google's example: a study power of 70% to 80% means that if lift is detected, its certainty will be between 70% and 80%. Longer studies, higher budgets, a larger holdback and more conversion actions all raise power. If you focus on conversion value and values vary widely, power drops.

Setting up a users-based study

1. Open Lift studies under Campaigns > Experiments and select the plus button.
2. Choose Conversion Lift, then "Based on users".
3. Name the study and select campaigns, from a list or by uploading campaign IDs in a CSV. Campaigns already in another study cannot be selected.
4. Keep the campaign's conversion goals or choose specific ones. Supplementary conversions carry a warning icon in the picker.
5. Pick the report focus: Conversions (incremental conversions and iCPA) or Conversion value (incremental value and iROAS). The other metrics stay in the report.
6. Set start and end dates and the holdback, from 1% to 50%. If you also run Brand Lift and Search Lift, leave the holdback at 30%; all users-based lift types share the same holdback group.
7. Check study power, then Create study.

Campaigns, holdout and duration

• Campaigns: Google recommends including all campaigns that share the same type, audience and targeting. Measuring a subset shows what those campaigns add while the rest run, and it lowers the chance of detecting lift, because held-back users can still see your other ads.
• Holdout: setup suggests a size. Bigger holdouts (up to 50%) collect data faster but cost more conversions you would otherwise have won. Smaller ones need a longer study.
• Conversions: upper and mid-funnel actions such as page views or lead forms show higher lift. Google suggests measuring them next to bottom-funnel actions.
• Bidding and attribution: Google says campaigns on conversion-based bidding (Target CPA, Maximize Conversions, Target ROAS) and data-driven attribution are associated with higher lift.
• Duration: at least 7 days, but Google recommends 14 or more. It reports up to a 17% drop in absolute lift for studies with a long conversion lag that ran under 14 days. Products with long buying cycles need longer.
• How often: every study gives up some conversions in the holdout. Google says most advertisers run about 1-2 studies a year, often ahead of budget decisions.

Geography-based studies

Geo studies split a country into Google Marketing Areas (GMAs): sub-country regions built to limit "contamination", where someone sees an ad in a treatment area and converts in a control area. Contamination makes the measured lift smaller. Requirements and setup:

• Campaigns must target a single country.
• Conversion sources: the Google Ads conversion tag, Firebase, DV360 Floodlight, or your own offline data aggregated to ZIP or city level and shared with your account team.
• In this beta you can run only holdback and go-dark studies.
• Setup: Lift measurement, plus button, Conversion Lift under "Based on Geo", then campaigns, dates and conversions.
• Feasibility shows High, Medium or Low. Google does not recommend going ahead on Low; raise the budget or include all campaigns of a channel or goal.

Google also documents open-source geo experiments that you run by changing campaigns yourself: holdback (new strategy in test regions only), go-dark (switch off spend in test regions) and heavy-up (extra budget in test regions). Each uses Presence location targeting to avoid leakage and individual, not shared, budgets. When a budget-capped campaign drops test regions, Google notes that unspent budget flows into the control regions, so the daily budget has to be cut to the control regions' share. Location settings are covered in our location targeting guide.

Reading users-based results

• Incremental conversions (absolute lift) = treatment conversions - control conversions.
• iCPA = total ad spend / incremental conversions. Google's example: $10,000 and 800 incremental conversions give $12.50.
• Relative lift = incremental conversions / control conversions. With 90 incremental conversions for every 10 control conversions it is 900%. Google warns that relative lift is easy to misread across studies, because a small control group inflates it.
• Incremental conversion value = treatment value - control value. It is more volatile than counts, so intervals are wider.
• iROAS = incremental conversion value / total ad spend. Google's example: $20,000 treatment and $10,000 control revenue on $5,000 spend gives 2, or 200%. Normal ROAS divides all attributed value by spend.

Delayed incremental conversions ("cooldown"): from 14 April 2026, Demand Gen-only studies that started on or after 2 March 2026 include modeled conversions expected after the study ends. They are in the report by default; the results tooltip shows figures without them.

Intervals: Google is moving Conversion Lift to Bayesian methodology, which also uses historical campaign data and reports 80% credible intervals: an 80% probability that the true lift lies in the range. Certainty of lift reads the same way as for Brand Lift: 90% or more is strong, 50-90% directional, and below 50% shows as no lift.

Reading geo results

Geo results are under Goals > Measurements > Lift measurement, in the study's Details page. iROAS is incremental conversion value divided by incremental cost: an iROAS of 2 means $2 of new conversion value per $1 spent. Results come with a confidence interval; Google's example is an iROAS of 2.2 with a range from 1.3 to 3.5. An optional cooldown period after the test keeps collecting data and is recommended when your conversion cycle is longer than a few weeks.

Statuses are "Significant Positive iROAS" (new revenue exceeds spend), "Not enough data" (common early; check budget for High feasibility, over-targeted audiences and location exclusions) and "No significant lift detected" (look at conversion slices).

When results appear

From the Bayesian methodology page:

• Results start showing on the 4th reporting day, and conversion processing can make that up to 10 days.
• They show only once certainty of lift is above 50%.
• The minimum is 150 conversions from the treatment group and 65 from the control group.
• No results below $5,000 USD spend.

Google recommends waiting until the study ends before acting, because results move while it runs.

Changes during a study

• You can add or remove campaigns and change the end date. Pausing means moving the end date; an ended study cannot be restarted.
• You cannot add Brand Lift or Search Lift to a running Conversion Lift study; end it and create a new one.
• Bid and budget changes are fine. Changing creatives or audiences mid-study makes the results harder to learn from.

Our view

This is our own opinion, not Google's guidance. Conversion Lift answers the question attribution cannot: would these sales have happened anyway? It is most useful before a large budget shift, for example deciding whether brand search or remarketing deserves its spend. Get conversion tracking, enhanced conversions and consent mode right first, measure all campaigns of one type together, run at least 14 days, and judge the result on iCPA or iROAS next to your normal CPA and ROAS, not on relative lift.

Frequently asked questions

What is the minimum budget for Conversion Lift in Google Ads?

For users-based studies Google lists a minimum campaign budget of $5,000 USD and at least 1,000 observed conversions. The budget actually needed is set during setup from your historical conversion data. Geo studies usually need more.

What is the difference between Conversion Lift and Google Ads experiments?

An experiment compares two versions of a campaign to see which performs better. Conversion Lift compares people or regions that saw your ads with ones that did not, to measure how many conversions the ads caused at all.

How long should a Conversion Lift study run?

At least 7 days, but Google recommends 14 days or more, and longer for products with long buying cycles. Results start showing from the 4th reporting day.

What is iROAS?

Incremental return on ad spend: incremental conversion value divided by ad spend. It counts only the value that would not have happened without the ads, while normal ROAS uses all attributed value.

Why can't I see Conversion Lift in my account?

Conversion Lift is not available to every account. Google says to contact your Google account representative to use it.

For Demand Gen setup, see our Demand Gen campaigns guide; for Performance Max, see Performance Max.

Want landing pages that turn more of that incremental traffic into customers? See our website development service and all our services, or email [email protected].