Geo-test feasibility planner
Incrementality testing
How Long Does It Take to Set Up a Geo Experiment?
With usable data access, feasibility and test design often take 2–3 weeks; the in-market observation period then depends on the customer timeline, outcome volume, intervention, and comparison-market readiness. This planner is a directional starting point, not a substitute for historical analysis.
How long should a geo test run?
Long enough to cover how long your customers take to buy, not a generic countdown. Four active weeks plus a two-week cooldown is a starting point for many DTC sales and revenue tests. Treat it as a place to start the conversation, not a rule. Size it with MDE and power on your own sales history. Four weeks is my rule-of-thumb minimum. Shorter tests give only a directional read. The purchase cycle matters most. Your sales volume and how much you change spend play a part. So does how well the markets track each other. Upper-funnel tests often need longer.
Not sure your test can see the lift? Start with a geo test power analysis.
Start with your context
What kind of lift are you looking for?
Choose the closest business outcome for this test.
What the planner can—and cannot—tell you
A test duration has two clocks.
The first is the customer’s clock: how long it takes to move from the tactic you are trying to the outcome you want to measure. A new conversion channel can create a useful directional read quickly. A brand-awareness strategy measured against a long-consideration purchase can take months before the curve even begins.
The second is the data’s clock: how much ordinary movement exists in sales or another KPI, how much you change the treatment, and how much outcome volume reaches the candidate markets. A larger, credible intervention can reveal a signal sooner; a small footprint or low volume can require more time.
A browser form cannot replace the historical geo-level analysis that chooses markets, validates a comparison, and estimates the design. It can make the trade-offs clear before you ask for that work.
Before day one
Make sure the comparison can earn its role.
A geo test creates a parallel-universe question: what happened where the media changed, compared with what would likely have happened without the change? Candidate treatment and comparison markets need to behave predictably enough in history for that comparison to be credible.
Check more than whether two groups have similar averages. The relationship should hold through normal hills and valleys and across relevant time windows. If it does not, choosing a longer duration will not fix the wrong comparison.
Read matched markets vs. synthetic controlHave a daily GA4 export? Run the three pre-launch checks on your own data.
Use the right claim
A short read can be useful without being the whole story.
For a high-volume conversion change, a short test may be enough to make the next practical move. Label that result as directional. For upper-funnel work, lower-funnel purchases, long consideration cycles, or a decision that will govern an always-on budget, allow the test and observation period to match the claim you want to make.
Do not stop a fixed-duration test simply because an interim chart looks promising. But a pre-agreed safety rule can protect the business: if a major holdout clearly harms the operation, restore the business before preserving more data.
Statistical power
How Do You Know You Can Trust a Geo Test?
You cannot guarantee statistical significance before a test begins. You can design for enough statistical power: a reasonable chance of detecting a change that is real and large enough to matter to the decision.
That work starts with the client’s own history. We define the intervention and outcome, estimate the smallest lift or loss worth acting on, measure ordinary movement in the outcome, and test candidate market comparisons against historical data. From there, we can estimate whether the proposed spend, duration, and market footprint are likely to produce a useful answer.
More sales volume can make a signal easier to detect. But the question matters too. Testing whether a conversion channel changes sales is different from testing whether upper-funnel video changes a later purchase. A smaller intervention, noisier outcomes, or a longer consideration cycle can require more time, more spend, or a broader treatment footprint.
If the available data cannot support a credible design, that is a useful finding. We can adjust the question or design, or decide a geo test is not the right next measurement step.
FAQs
Questions teams ask before a geo test.
Is four weeks always enough?
No. Four active weeks is a useful starting point for many DTC sales and revenue tests, not a universal rule. The consideration cycle, outcome, treatment footprint, and historic market behavior decide whether it is enough.
Why add an observation or cooldown period?
A purchase or other outcome can occur after the active media change ends. An observation period helps capture that delayed effect and separates the treatment window from the final read.
What should I bring to a feasibility review?
Bring the decision, the KPI, historical sales or outcome data by geography, channel spend history, the proposed intervention, major calendar events, and the team’s appetite for a holdout or heavier spend.