How to pick an incrementality test length, step by step

Usually in seven moves. Read your purchase lag in GA4 and your weekly sales swings in Shopify, then draft at least two whole weeks that cover the lag. Check the platform's power at those dates, check your history, add a cooldown for late buyers, and lock the end date before launch.

This sets the length of a test you have already chosen: a regional go-dark or a platform lift study. Each step names the screen where its number lives. You need GA4, Shopify and Google Ads; on Meta, the same logic runs in Experiments.

Step by step

1. Read your purchase lag in GA4. Choose your purchase key event and filter Path length to greater than 1 touchpoint. Note Days to key event for those paths: they are the buyers a short test misses. Path: GA4 > Advertising > Key events dropdown > Key event attribution paths > Path length filter.
2. Measure how much a normal week swings. Group a year of sales by week and note how far a typical week lands from its neighbours. A lift smaller than that everyday swing needs many weeks to stand out. Path: Shopify admin > Analytics > Reports > Category filter > Sales > Total sales over time > Group by.
3. Draft the length in whole weeks. Take the longer of your multi-touch lag and Google's recommended minimum of 14 days. Round up to whole weeks, so both groups see the same mix of weekdays and weekends. Path: Google Ads > Campaigns menu > Campaigns > Experiments > Lift studies tab > plus button > Conversion Lift.
4. Check the power at those dates. Choose Based on users, add the campaigns, enter the start and end dates and read the study power. Google says to aim for 90%, and that a longer study, a bigger holdback and more conversion actions all raise it. On Meta, some tests show a power estimate first, and Meta typically suggests 80% or higher. Path: Lift studies tab > plus button > Conversion Lift > Based on users > study power.
5. Check your history for a geo test. For a regional go-dark, Google's geo guide wants pre-test data of at least 3 times the test. Count back from your start date and check those weeks hold no launch or site outage. Path: Shopify admin > Analytics > Reports > Total sales over time > date range.
6. Add a cooldown for late buyers. After the end date, the ads go back to normal but you keep counting sales from the test. A simple rule: make the cooldown at least as long as your lag. Google recommends a cooldown date range for geo studies when your conversion cycle runs longer than a few weeks. Meta's Conversion Lift counts conversions between the test's start and end dates, so on Meta, stretch the test instead. Path, to read it: Google Ads > Goals > Measurements > Lift measurement > study name > Details.
7. Lock the end date and read the result once. Results appear while a study runs, but Google recommends waiting until the end. You can pause a user-based study by changing its end date, and once it ends it can't restart. When you read it, leave out recent days with under 90% of conversions reported, as Google advises for its experiments. Path: Google Ads > Campaigns menu > Campaigns > Experiments > Lift studies tab.

On Meta, Conversion Lift runs in Experiments. As a guide, Meta wants a campaign from the past year with $5,000 USD or more of spend and 500 conversions.

A worked example

For illustration, with round invented numbers. Say you sell something people mull over, and GA4 shows 33 days to key event on your multi-touch paths.

Step 4 then checks the power. Suppose the study power reads 75% at five weeks with a 10% holdback. Say a 20% holdback lifts it to 90%: you pay with a bigger slice of buyers who miss your ads, not with more weeks.

Now see what cutting it short would cost. For illustration, say the channel truly adds €10,000 of sales on €4,000 of spend over the test. That is 2.5x, exactly break-even at a 40% margin: 1 divided by 0.40, the same sum as one store's Break-even sheet.

Suppose someone stops the test on day 10 because the line looks flat. Google found up to a 17% drop in absolute lift in long-lag studies shorter than 14 days. If the full 17% goes missing, the test reports €8,300 of added sales, about 2.1x. Against the break-even line, that reads as a loss, and a channel that pays its way gets cut.

Five weeks of patience, plus five of counting, would have kept it.

What to check when the plan looks wrong

• The power estimate barely moves. Google says measuring only a subset of campaigns lowers your chances of detecting lift. Test the whole channel, or add a more frequent action, such as page views, as a secondary metric.
• The early result flips from week to week. That is normal while data builds. Meta's advice for a test whose groups look alike is to wait until it has finished, because outcomes can change. Your locked end date is doing its job.
• One region spikes on its own. A launch, a pop-up or a site outage moved it. Google's geo guide says to drop regions with major launches and to consider removing outage dates from the analysis.
• The tested channel buys slower than the store. A store-wide lag can hide one slow channel. In the paths report, switch the table to Campaign and read Days to key event for the campaigns under test.
• Orders keep landing after the cooldown. Your slowest paths run longer than the average suggested. Next time, size the cooldown on the top-revenue paths in GA4, not the store average.
• Meta's lift numbers differ from Ads Manager. Meta counts lift between the test's start and end dates, while Ads Manager credits conversions inside attribution windows. Judge the test on the lift numbers.

What to do this week

1. Check GA4's lookback window. In Admin, under Data display, click Events, then Attribution settings. Pass: purchases use a window longer than your lag, such as the default 90 days. Fail: someone set it shorter than your lag, so lengthen it now; changes apply going forward, not to past paths.
2. Find a quiet window in last year's sales. In Shopify, open Total sales over time for the weeks you plan. Compare them with the same weeks last year. Pass: no sale, launch or spike sat inside the test or its cooldown. Fail: one did, so move the test to a calmer stretch.
3. Clear your campaigns for the study. In Google Ads, open the Lift studies tab and select the plus button. Choose Conversion Lift and add the campaigns you will test. Pass: they show as Eligible. Fail: they show Needs attention, meaning inactive, in another study or incompatible, so fix that before the start date.

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: Key events attribution paths report (Google Analytics Help); Sales reports (Shopify Help Center); Set up Conversion Lift based on users (Google Ads Help); About confidence in your tests and experiments (Meta Business Help Center); Implement campaigns for geo experiments (Google Ads Help); Understand your Conversion Lift based on geography measurement data (Google Ads Help); Differences between Conversion Lift test results and other reporting tools (Meta Business Help Center); About Conversion Lift (Meta Business Help Center); Test your bid strategy (Google Ads Help); Similar performance between test and holdout groups in a test (Meta Business Help Center); Select attribution settings (Google Analytics Help).

Related answers

• All Incrementality questions
• How long should an incrementality test run?
• Holdout test planner
• Incrementality testing, explained
• Lift test, in the glossary