What is a matched market test?

A matched market test changes your marketing in some regions and leaves look-alike regions alone, then compares sales in both. If the two sets of markets moved together before the test, the gap that opens during it is usually what the change caused. It suits channels you can switch on or off by region.

Picture the classic version. A brand airs a new radio spot in one city, keeps a similar city as its yardstick and compares the tills. Google's researchers describe so-called matched market tests as ones that may compare a single control region with a single test region. Modern designs use more markets and let software do the matching. The idea stays the same: build a fair "what if" from places that did not get the change.

What one store's data shows

One store's anonymised GA4 export, 1 January 2024 to 21 August 2026. It holds shares of revenue only: no ad spend, no order counts.

Start with credit. In that store's Channels sheet, the touched column sums to 110.4% of revenue, because a journey that touched two channels counts in both. That is attribution doing its job: sharing out sales that already happened. It cannot say what any one channel added. A matched market test shares nothing out. It changes one thing in some markets and counts every sale there, from every channel, Direct included.

Then the blind spot. Organic Video sits at 0.0% of revenue in all three views of that store's Channels sheet. Either none ran, or it ran and no tracked visit came from it, and the export cannot tell which. A channel people watch rather than click can earn next to nothing in click credit. A matched market test can price it if you can buy it by region: local radio, posters or paid social aimed at chosen cities. Organic posts reach everyone everywhere, so no market stays clean of them.

Then the clock. In that store's Journeys sheet, one-touch journeys carry 79.5% of revenue and take 0.5 days to buy. In that store's export, journeys of 10 or more touches carry 3.0% and take 16.0 days. So in that store, most revenue comes from quick buyers, with a thin tail that takes about two weeks. Google's GeoX glossary calls the period after a test, used to check for delayed impact, a cooldown. Your markets have to stay matched through it.

What the export cannot show is geography. It holds no regions, so it cannot tell you which markets match or what a test would find. That needs daily sales by region, from your own orders.

Why do so many matched market tests mislead?

The usual advice is to pick two similar cities, run the ads in one and compare. Each part of that advice hides a trap.

• Similar in size, not in rhythm. Two cities with the same population can have different sales cycles. Meta's GeoLift guide says to match test and control markets on the exact outcome you will measure. Google's GeoX design groups markets by sales volume, trend and seasonality.
• Matched and judged on the same weeks. A pair chosen because it tracked well last year will always look good on last year. Google's GeoX FAQ says that kind of in-sample fit often underestimates the noise, and the study can end up underpowered. Its fix is to check each design on weeks it did not use, and to run placebo tests.
• The test side picked by hand. If you choose which city gets the ads, your hunch rides along into the result. In Google's design, chance assigns markets to test or control inside each matched group.
• Ads that cross the border. People commute, travel and search for places they are not in. By default, Google Ads also reaches people who showed interest in your locations. Google's geo test guide says to select Presence to prevent leakage, and Meta has its own box to untick.
• Scored on the platform's numbers. GeoX asks for raw, unfiltered sales rather than attributed conversions, so attribution rules cannot skew the result. For a store, that means sales by region from Shopify.

What can a matched market test not tell you?

• A small lift. Google's GeoX FAQ says the smallest change a test can detect depends on how volatile and how well matched the markets are. More budget does not lower that floor. A channel that adds a little may never show up.
• A quick, cheap answer. Every sale in the region counts, attributed or not. Google's GeoX FAQ says that makes geo tests noisier, so they often need more money and time than user-level lift tests.
• Anything beyond the change you made. Switching a channel off values the budget you run today. Adding budget in test markets, which Google calls a heavy-up, values the next euro. Neither speaks for other markets, seasons or creative.

So when is it the right tool? When a platform cannot run a user-level test for you, when sales happen offline, or when you want one referee for several platforms. Google's GeoX FAQ calls geo tests the primary method for offline and omni-channel impact.

What to do this week

1. Lay two candidate markets side by side in Shopify. Go to Analytics > Reports and open Total sales over time. In the Dimensions menu, set the time unit to week and add Billing city with ⊕. Then filter Billing city is one of your two cities. Pass: the two lines rise and fall in the same weeks, peaks included. Fail: one city has spikes the other lacks, so it is not a match.
2. Check your Google reach setting. Go to Campaigns in the Campaigns menu and select the Settings icon next to the campaign you would test. Expand Locations, then Location options. Pass: Presence is selected, or you will switch to it for the test. Fail: it reads Presence or Interest, the default, which also reaches people interested in your locations.
3. Check the reach box on Meta. In Ads Manager, open an ad set, go to Audience and select Locations. Pass: Reach more people likely to respond to your ads is unticked, so ads stay where you aimed them. Fail: it is ticked, which Meta may do by default for a city or region, so untick it before any test.

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: Estimating Ad Effectiveness using Geo Experiments in a Time-Based Regression Framework (Google Research); GeoLift Best Practices (Meta); Welcome to GeoLift (Meta); Meridian GeoX: Design methodology (Google for Developers); Meridian GeoX: FAQs (Google for Developers); Meridian GeoX: Prepare your pretest data (Google for Developers); Meridian GeoX: Glossary (Google for Developers); Meridian GeoX: Types of experiments (Google for Developers); About advanced location options (Google Ads Help); Implement campaigns for geo experiments (Google Ads Help); Use location targeting (Meta Business Help Center); Sales reports (Shopify Help Center); Filtering and editing your reports (Shopify Help Center); Setting and comparing time ranges for your reports (Shopify Help Center)

Related answers

• All Incrementality questions
• How to run a matched market test, step by step
• How do I run a geo holdout test?
• Incrementality testing, explained