---
title: "How to A/B test in a physical store | GoTrack"
description: "A/B testing in a physical store: before-and-after or two matched stores, one change at a time, rates not totals. A practical method with product data."
url: https://gotrackgo.com/en/in-store-ab-testing
language: en
updated: 2026-09-30
publisher: GOTONOM Yazılım Teknolojileri A.Ş.
---

# How to run an A/B test in a physical store

Websites test everything; stores mostly guess. How to test a placement, a fixture or screen content in a real store with product interaction data, and how to read the result honestly.

## The short answer

**Change one thing, measure the same products before and after, or in two matched stores at the same time, and compare rates: pickups per day, put-back rate and hold time.** Keep everything else still, give it enough days, and treat the result as a change you observed, not as proof.

## Two ways to test

| Design | How | Good for | Watch out for |
|---|---|---|---|
| Before and after, one store | Measure a period, make the change, measure the same length again | Placement, facing, fixtures, screen content | Seasons, campaigns and weather change behaviour on their own |
| Two matched stores, same period | Change one store and leave the other as it is | Separating the change from the season | Stores are never identical: compare each with its own baseline too |

## Step by step

1. **Write the question**: One change and what you expect it to move. For example: moving a model to eye level raises its pickups.
2. **Choose the measure**: Pickups per day, put-back rate or hold time for the products involved, decided before you start.
3. **Record a baseline**: The same products in the same places, for a normal period.
4. **Make one change**: Move, re-face or swap content, and change nothing else in that area.
5. **Measure the same length again**: The same days of the week if you can; avoid campaign days unless the campaign is what you are testing.
6. **Compare rates**: Per product and per store, and check whether neighbouring products lost what the moved one gained.

## Reading the result

- **Small differences need many pickups.** A few pickups more or less is noise; look for a difference that holds day after day.
- **Look next door.** A product that gains pickups may take them from its neighbours; check the whole area.
- **A change, not a cause.** A before-and-after shows what changed; it does not prove your change caused it. GoTrack shows the change; it does not run statistical tests for you.
- **Join with sales for revenue.** GoTrack does not read POS data; combine its exports with your sales.

## Questions

### Can GoTrack tell marketing which products are examined, and for how long?

Yes. It reports pickups per product and variant, average hold time, put-back rate, products compared together and never-touched products. It measures how long a product is held, not how long someone stands in front of it.

### Can GoTrack show whether sales went up?

Not on its own. GoTrack measures what happens at the shelf — pickups, hold time and put-backs — and does not read POS data. To relate interest to sales, join GoTrack's CSV exports or webhook events with your own sales data.

### Can we test a new product placement with GoTrack?

Yes, as a before-and-after comparison: move a product to a new shelf zone or fixture and compare its pickups, hold time and put-backs with the period before. GoTrack shows the change; it does not run statistical tests for you.

### What is a good put-back rate?

There is no universal benchmark: it depends on the category, the price and the store. The useful comparisons are a product against itself over time, and against similar products in the same store and zone. GoTrack reports the rate per product, colour, zone and store so those comparisons are possible.

## Have a change to test?

Tell us what you want to try; we will show you what GoTrack would measure.

[Request a demo](https://gotrackgo.com/en/demo) · or write to info@gotonom.com

---

GoTrack · by GOTONOM · https://gotrackgo.com/en/ · info@gotonom.com