How to set up a conversion lift test, step by step

To set up a conversion lift test, first check that your account clears the platform's minimums. Then create the study in Google Ads or Meta. Pick campaigns that share one job, a purchase conversion and the share of people who sit out. Run it longer than buyers usually take to decide, and judge the result against break-even.

Google documents its setup step by step, so the steps below follow Google Ads, with Meta's rules alongside. Meta's help says you can create a conversion lift test in Experiments or through a Meta account representative. It does not publish the screens, so this guide does not guess them.

Bring three things: campaigns with enough conversions, a clean conversion signal and one question. Do these campaigns add purchases, or collect purchases that were coming anyway?

Step by step

1. Check the entry ticket. Google's user-based study asks for at least 1,000 observed conversions, a $5,000 USD minimum campaign budget and a compatible conversion action. Meta's guide asks for a campaign from the past year that spent $5,000 USD or more and logged 500 conversions. Path, in Google's words: the Lift studies tab under Campaigns > Experiments, within the Campaigns menu.
2. Clean up the conversion signal. Google recommends enhanced conversions, consent mode and Google tag gateway before launch, and runs automated diagnostics during setup. In the European Economic Area, it calls consent mode critical for recovering missing data through modelling. Meta asks for the Conversions API with an Event Match Quality score above 5, or another supported data source. Path: Google Ads > Goals icon > Settings > Enhanced conversions; Meta Events Manager > Data Sources > your dataset > Settings.
3. Pick campaigns that share one job. Google recommends including all campaigns that share the same type, audience and targeting strategy. Leave one out and the holdback group may still meet its ads, which makes lift harder to detect. A campaign can sit in only one study at a time. Path: Lift studies tab > plus button > Conversion Lift > Continue > Based on users > Continue > Campaigns > Select campaigns.
4. Choose the conversion and the scorecard. Pick your purchase conversion goal. Then choose Conversions, which puts incremental conversions and iCPA first, or Conversion value, which puts incremental conversion value and iROAS first. Google notes that conversions that happen more often raise the certainty of the result. Path: same setup page > conversion goals > type of metrics.
5. Set the dates and the holdback. Enter a holdback between 1% and 50%. A bigger holdback fills the sample sooner but gives up more sales, and a smaller one needs a longer study. Cover your conversion lag: Google typically recommends more than 14 days. Path: same setup page > start and end dates > holdback size.
6. Read the study power before you press Create. Google estimates the certainty of lift as a range from 50% to 95%. Aim for 90%; below it, Google offers budget guidance, and longer dates or a bigger holdback also raise the power. Path: Review the study power > Create study.
7. Leave it alone while it runs. Bid and budget changes are business as usual to Google. New creatives and audiences might affect the study, it warns. You can move the end date while it runs, but a study that has ended cannot be restarted. Path: Goals menu > Lift measurement.
8. Read the result against your margin. Add the Conversion Lift columns to the Lift measurement table. Read incremental conversions first, then relative lift, iCPA and iROAS. Judge iROAS against your break-even, not against the platform's ROAS. Path: columns icon > Modify columns > Conversion Lift > Apply.

A worked example

In one store's Journeys sheet, journeys with two or three touches took 12.5 days to buy. If your multi-touch buyers take that long, a two-week study would end just as the slower ones decide. For illustration, book four weeks.

Next, the sums, with round numbers invented for the example. For illustration, say your purchase campaigns spend €20,000 over those four weeks. Say the campaigns' own report credits them with €90,000 of conversion value, a platform ROAS of 4.5x.

Say the study then reports 400 incremental conversions worth €60,000, against 1,600 conversions in the control group. For illustration, that is a relative lift of 25%, because 400 divided by 1,600 is 0.25. For illustration, iCPA is €50: €20,000 divided by 400. For illustration, iROAS is 3x: €60,000 of incremental value on €20,000 of spend.

If the platform credited €90,000 and the test found €60,000, a third of the credit was sales that were coming anyway.

The floor comes from one store's Break-even sheet: with a 40% margin, break-even ROAS is 2.5x, since 1 divided by 0.40 is 2.5. If your margin is also 40%, an iROAS of 3x clears that floor, and each euro of spend brings back €1.20 of margin. The campaigns earn their keep.

For illustration, the 4.5x report would have told you to scale hard. The test says scale with care, in steps, and test again at the higher budget.

What to check when the report looks wrong

• Nothing shows after three days. Google's results start on the 4th reporting day, and processing can take up to 10 days. Give it time before you poke it.
• The lift percentage looks heroic. Relative lift divides by control conversions, so a quiet control group inflates it. Google also warns that relative lift can mislead across studies. Check incremental conversions and iCPA first.
• Certainty sits between 50% and 90%. Google calls that directional. You can still change the end date of a running study, so extend it if the budget allows.
• Lift is far below the ad report. That is the design, not a bug. Conversion Lift ignores your attribution settings and counts the difference in all conversions between the groups.
• A campaign will not go into the study. The picker marks it Needs attention if it is inactive, in another study or incompatible. Take it out of the other study first.
• Buyers keep buying after the end date. Google projects late incremental conversions for Demand Gen studies only. Meta does not recommend its post-test conversion window. Set the end date from your lag, not your patience.

What to do this week

1. Turn on enhanced conversions now. In Google Ads, go to Settings within the Goals icon and expand the Enhanced conversions section. Pass: Turn on enhanced conversions is already checked. Fail: it is not, so check it now, weeks before a study rather than halfway through one.
2. Note your break-even iROAS. In Shopify, go to Analytics > Reports, click the Category filter, choose Profit Margin and open Gross profit by product. Divide 1 by your gross margin. Pass: you have a number the iROAS must beat before the study starts. Fail: the report is thin, because Shopify reports profit only for products with a recorded cost per item.
3. Compare your lag with the study length. In GA4, click Advertising, then Key event attribution paths under the Key events dropdown. Set the path length filter to greater than 1 and note Days to key event. Pass: the figure is under 14 days, Google's usual minimum. Fail: it is higher, so set the end date further out before you create the study.

Check the homework. Your GA4 Attribution paths export already holds the evidence. Causality Engine reads that one file and shows what each channel caused next to what last-click gave it, in 1 to 2 minutes, for €99 once (excluding VAT), refundable within 30 days. Check the homework

Sources, 1 October 2026: About Conversion Lift (Meta Business Help Center); Set up Conversion Lift based on users (Google Ads Help); About Meta Events Manager (Meta Business Help Center); Understand your Conversion Lift based on users measurement data (Google Ads Help); About Bayesian methodology in Conversion Lift (Google Ads Help); Set up enhanced conversions for web using the Google tag (Google Ads Help); Profit reports (Shopify Help Center); Key events attribution paths report (Analytics Help).

Related answers

• All Incrementality questions
• What is a conversion lift test?
• What is a conversion lift test for a sports nutrition brand?
• Meta Conversion Lift explained: how to run it and trust the result
• Incremental ROAS