The short answer
For a geo incrementality test, 30 days after treatment is a practical starting point for measuring delayed sales. Establish your company's window by checking when the sales effect stops accumulating, then use it consistently across comparable tests. Shorten or extend future windows when evidence supports it, and include any later reversal in the result.
This article is part of Asked by Marketers, a series answering real questions from marketing leaders.
Sellforte's team holds more than 1,450 meetings each year with marketing leaders in Ecommerce and Retail about Marketing Mix Modeling and incrementality testing. Each week, we anonymize at least one question from those conversations and answer it in depth, based on what marketers are actually struggling with, not what keyword tools suggest. About the series →
Why marketers ask this
The campaign has ended, spending has returned to normal, and the team needs a result for its next budget decision. Yet customers may still be deciding whether to buy. Others may have bought earlier because of the campaign and now have less reason to purchase. A short measurement window can miss either effect. Waiting longer also creates complications when the next campaign reaches the same customers.
What post-treatment window should we start with?
For a geo test without an established follow-up window, I would plan to measure sales for 30 days after treatment ends. Use that first analysis to learn whether 30 days captures the remaining effect. Use the evidence to set a company-specific window for comparable incrementality tests.
- Confirm when the intervention actually ends. Check the media data for the date the tested spending difference stops. If restoring the campaign takes several days, the scheduled end date may be wrong.
- Collect daily sales and media data for the original test and control groups through the follow-up. Keep the final sales date separate from the reporting date: transactions through that cutoff may arrive later.
- Use earlier comparable tests to choose the primary window before launch. If you have no such evidence, plan the 30-day follow-up and a 14-day sensitivity check. Agree what would justify collecting a longer follow-up, including an effect that is still continuing at day 30, and what would invalidate the comparison.
- Once you have established how long the effect lasts, use that window for future comparable tests. A shorter or longer standard should reflect your own evidence. Document exceptions before launch when a campaign's purchase cycle or delivery schedule differs.
A frequent-purchase retailer and a business with a long consideration cycle may need different windows. Purchase-delay data and evidence of advertising carryover help you plan, although click-to-purchase lag does not establish the full causal effect. Delayed-effect estimates from Marketing Mix Modeling are also useful inputs to check against experimental evidence.
For a platform conversion-lift study, confirm before launch whether the report includes observed conversions after treatment, modeled delayed conversions, or neither. The platform's available follow-up may differ from your preferred window and from the attribution window in ordinary campaign reporting. Keep modeled extensions separate from sales observed during follow-up.
How do we know whether 14, 30, or 60 days is enough?
Check whether the estimated daily sales effect has faded and the cumulative effect has settled. Use sales against the same valid control or counterfactual, which estimates what would have happened without the tested change. Total sales after the campaign include purchases that would have happened anyway.
| What you see | What to do |
|---|---|
| The effect has faded by day 14 and stays flat afterward | Use the later observations to confirm that 14 days captures the effect. That evidence can support a 14-day standard for future comparable tests. |
| Sales are still being gained or lost at day 14 | Extend the data through day 30 and reanalyze the same groups. You usually do not need to run the intervention again if the original comparison remains valid. |
| The effect still accumulates at day 30 | Examine a longer follow-up, such as 60 days, and check when it settles. If the comparison remains valid, use that evidence to revise the standard window. Sixty days is not a universal upper limit. |
| The cumulative effect reverses during follow-up | Include the reversal in the net result. Investigate purchases brought forward or delayed until advertising resumed, as well as other business changes. |
A settled cumulative line can remain above or below zero because it includes sales gained or lost earlier. Inspect the uncertainty: a few flat days do not establish that the effect has ended, and a late movement can be noise. Check promotions, stock availability, tracking changes, and control comparability; see assessing an incrementality test's credibility.
Keep the original primary result when extending an analysis, and label the longer window as a follow-up analysis. Learning the duration from a completed test should inform the plan for the next one. Stopping at the highest iROAS or extending until lift becomes statistically significant makes the estimate depend on the desired result. Report the planned sensitivity checks, too. Once little additional signal remains, more days can add uncertainty without improving the decision.
How this looks in practice
Consider a hypothetical retailer testing a temporary increase in advertising. Before launch, the team chooses 30 days of follow-up based on earlier tests, with a 14-day sensitivity check. Assume the groups remain comparable, later advertising returns to business as usual, and the experiment causes no further spend difference after treatment. All figures below are illustrative, and the sales definition is unchanged across periods.
| Measurement cutoff | Cumulative incremental sales | Additional spend | iROAS |
|---|---|---|---|
| Treatment ends | $80,000 | $20,000 | 4.0 |
| 14 days later | $100,000 | $20,000 | 5.0 |
| 30 days later | $90,000 | $20,000 | 4.5 |
The first 14 days add $20,000 to the estimated sales effect. The remaining follow-up contributes a negative $10,000, leaving $90,000 in total. That pattern could reflect some purchases brought forward, but the analyst must also investigate other causes. The primary result is 4.5, accompanied by its uncertainty interval. The earlier 5.0 remains a sensitivity result; it does not replace the agreed endpoint because it is higher. Uncertainty intervals are omitted here to keep the arithmetic visible, but they are necessary for an actual decision.
Related questions
Can we still use older tests that had no post-test window?
Yes, if the design is sound and the result is useful for its measured horizon. Record the missing follow-up when using the evidence in MMM calibration, and assess whether delayed or reversed effects could change the decision. Reanalyze the original groups if suitable data exists; otherwise, do not add a guessed percentage to imitate a longer test.
Does a longer follow-up capture customer lifetime value?
It captures additional observed sales only while the comparison remains valid. It does not establish the full future value of customers acquired during the test. Keep observed sales and any projected customer value separate, and avoid counting the same purchases twice; see measuring the full sales impact of mid-funnel and top-of-funnel campaigns.
How Sellforte helps
Sellforte's experiment analysis shows sales against a counterfactual, cumulative treatment effects, and the associated spending change. Your team can use those views to examine the chosen follow-up period and keep the measured horizon explicit when applying the result to budget decisions or MMM calibration. Book a demo.
Authors
Lauri Potka is the Chief Operating Officer at Sellforte and has over 15 years of experience in Marketing Mix Modeling, marketing measurement, and media spend optimization. Before joining Sellforte, he worked as a management consultant at the Boston Consulting Group, advising some of the world's largest advertisers on data-driven marketing optimization. Follow Lauri on LinkedIn, where he is one of the leading voices in MMM and marketing measurement.