In-store conversion rate: the number the store pretends to have

Conversion rate is the number of customers who buy divided by the number of visitors who entered. The formula is simple. The problem is the denominator: the physical store almost never counts who came in.

E-commerce measures this with its eyes closed. It knows how many visited, how many added to cart, how many paid. The physical store, mostly, knows only how many paid.

What it measures well, when it measures

When there is foot-traffic counting at the door, conversion tells the hard truth: of every hundred who enter, how many leave with a bag. It is the number that separates the store that sells from the store that only receives visits.

For deciding layout, staffing, and display, it is the indicator that matters. More than revenue, which hides whether the store converts or just depends on those who already walk in decided.

What it hides

Without a traffic counter, the store’s “conversion” is invented from transactions. There is no real denominator, so the number is blind to whoever entered and left empty-handed. The store thinks it converts well because it does not see the customer who gave up in line or could not find the product.

Shelf-out makes this worse. Whoever enters, does not find the item, and leaves shows up nowhere. Conversion with no traffic erases exactly the lost sale that matters.

The pair that is missing

Conversion alone, without measured traffic, says nothing. The pair that makes it honest is foot-traffic counting at the entrance, and alongside it the average basket.

Read together, they locate the problem. High traffic and low conversion: the problem is inside the store, in shelf-out, price, or service. Low traffic and high conversion: the problem is bringing people in, not converting. Without the pair, the store invests in the wrong front.

The implementation note

Any in-store AI project, queue, flow, layout, point-of-sale recommendation, needs the truth of who entered. Optimizing conversion without counting visitors is optimizing in the dark, and the model learns from a number the store invented.

It is the same blindness as measuring service level at the dock and assuming the shelf is full. The right number lives where it is hard to measure.

Take the store-performance metric that has sat on the slide for six months. Does it have a real denominator, or is it a transaction pretending to be conversion? If it pretends, it is healthy and blind.

Send me its name and how it is measured today. In one hour I will send back a one-page audit: the measurement problem, the pair that should sit alongside, and the size of the work if the gap confirms. If it confirms, the two-week Diagnóstico scopes the rebuild.