The store is a laboratory. Most retailers just never run the experiment.
Field notes on testing things in real shops with real shoppers — what to change, how to control it, and how to know whether it worked before you roll it to fifty locations.
Retail produces opinions at industrial scale. Everyone in the building knows why sales dipped on Tuesday, why the new fixture "works", why shoppers ignore the end-cap. Almost none of it has been tested. Meanwhile the same company will happily spend six figures rolling out a change chain-wide on the strength of a gut feeling and one good week that happened to coincide with payday.
The living-lab idea is the opposite habit: treat a small set of stores as a permanent test bed. Change one thing. Watch matched stores where you changed nothing. Keep the receipts. It is slower than conviction and much cheaper than being wrong at scale — and once a team runs its first honest pilot, the appetite for untested rollouts tends to evaporate on its own.
This site is a working notebook for that practice. No frameworks with trademarked acronyms, no "retail is dead" essays. Just the craft of in-store experimentation: design, instrumentation, measurement, and the organisational politics of telling a stakeholder their favourite idea lost to the control group.
The notebook
How to run an in-store pilot that proves something
One variable, matched controls, a long enough window, and the discipline to decide the success metric before launch — not after the numbers come in.
Menu boards and merchandising screens: the most testable fixture in retail
Why screens make ideal experiment hardware — instant variants, dayparts, per-store assignment — and how to design a menu board test that isolates the screen's actual effect.
Measuring a store change without fooling yourself
Seasonality, novelty bumps, cherry-picked windows and the other classic ways retail "results" evaporate under scrutiny — and the simple defences against each.
The minimum lab stack
What you actually need to instrument a test store: POS access, a counter at the door, controllable screens, and a spreadsheet you trust. Everything else is optional.
House rules
We don't publish invented case numbers, and we'd encourage you to distrust anyone who does — real store data belongs to the retailer who generated it. What we share is method: how to set up tests so that the numbers you get in your own stores are numbers you can stake a rollout on. No sponsors, no vendor ties, and occasionally opinionated about bad statistics.