This piece originally argued that the physical store needed an operating system. We have since rewritten it. The architecture claim was never wrong. It just wasn’t the point, and it quietly let us off the hook on the part that matters. What follows is the corrected version.
Every digital business runs on an operating system: a platform that connects data, logic, and action. For years the physical store had none. Decisions ran on gut feeling, spreadsheets, and scattered data.
So we said the obvious thing: physical retail needs an operating system.
It’s true. It’s also the least useful sentence we could have offered a store manager. An operating system is architecture: plumbing, correctly built, invisible when it works. Nobody buys plumbing. And more importantly, describing the product that way let us stop at the easy half of the job.
The easy half is seeing
Most in-store analytics (ours included, in its first form) is very good at showing you things. Which zones earn attention. Where dwell collapses. How long the queue was on Saturday at 15:00. This is genuinely valuable, and it is genuinely incomplete.
Because a dashboard hands you an observation and walks away. What a store manager on a Tuesday morning actually needs is not a hundred observations. It is one sentence: do this, in this zone, this week. And then, four weeks later, a second sentence that almost nobody in this industry is willing to write: here is whether it worked.
That is a loop, not an operating system. Three steps, in order.
See
Behaviour on the floor is measured continuously, at zone level, through your existing cameras. Every visitor is classified by how they behave, not who they are: walk-bys who pass, short lingerers who slow, and Clear Lingerers who deliberately stop and consider: the physical equivalent of a click.
This is the decomposition that online retail was born with and physical retail never had. A store counts people at the door and receipts at the till. Between those two numbers sits the entire commercial life of the shop, and it is measured by nobody. Split the visit into reach, engagement and conversion, and you can finally tell a reach problem from an engagement problem, which matters enormously, because they need opposite fixes.
Act
Measurement that doesn’t narrow to a decision is just a more expensive way of being uncertain.
So the platform ranks zones by the euros within reach: the realistic value of a small, achievable improvement, benchmarked against the next gear up rather than some unreachable ideal. Your own best weeks first. Then your best comparable zone. Then the peer cohort. Best-in-class stays a horizon, not a target, because a goal nobody can hit is a goal nobody acts on.
Out of that comes one move. Not a list. One.
Prove
This is the step the industry skips, and it is the only one that compounds.
When you make the change, you log it with its date. A website can split its visitors in two. A store cannot. So a change you log is measured against the same floor before it: the same number of days either side, at least four weeks, one rule for every change. The verdict can read improved, declined, mixed or no change. Season, promotions and a second change on the same zone move the numbers too, so you decide whether it was the change. The euro result comes back labelled honestly: measured, modelled, or still building. We report proved euros only where a tracked action has real sales behind it. Everything else stays an estimate and says so.
Sometimes the verdict is no change. We report that too. A tool that only ever confirms its own advice isn’t measuring anything.
Three gauges, not three forces
The store state itself is expressed through three gauges, each doing a different commercial job:
- ATTRACT: grow. How much of the demand already on your floor becomes real engagement, and what the jump to the next gear is worth.
- SERVE: protect. Service and waiting at checkout and advice points, led by minutes against target. Any euro is sales at risk (est.), a modelled range that is never summed.
- DEPLOY: save. Staffing efficiency: idle or over-staffed capacity you could redeploy to where demand actually is.
A note on that last one, since this post previously said something different. We used to call the third gauge FLOW and define it as movement, congestion and bottlenecks. That was a mistake: it double-counted with SERVE, which already owns the queue story. DEPLOY measures staffing efficiency only. Congestion is not its job.
Behaviour is measured; money is modelled
One discipline holds all of this together, and it is worth stating plainly because the industry is careless about it.
Behaviour is measured. Money is modelled. Any euro figure is an estimate unless it comes from a tracked action with sales entered against it, and we label the difference every time. We measure realised impact. We do not promise a return, and you should be sceptical of anyone in this category who does.
What changes when the loop closes
A store that only sees has a report. A store that sees and acts has a busy manager. A store that sees, acts, and proves has a method: it makes its next move on evidence, because it accumulates a record of what actually works on its own floor rather than what works in a case study about someone else’s.
The operating system is still down there, doing its job. Edge devices, behavioural models, zone analytics, all of it running quietly on infrastructure you already own. But it was never the promise.
The promise is the loop: see the floor, act on it, prove what worked. Most tools stop at the first step. The whole point is the third.
