The visitors who don't buy: what happens between the entrance and the till

At the end of the day, almost every store knows exactly one number it can defend: the till total. How many people walked through the door to produce it, nobody knows. Anyone who bought nothing appears in no system at all.

In the same company's online shop, this has been different for twenty years. There, the visitor who buys nothing still leaves a complete session behind - with duration, sequence and the point where it stopped. Most-visited areas, peak times and the comparison against the previous period are standard reports, not projects.

That asymmetry stopped being a technology problem some time ago. This article covers what is measurable between the entrance and the till, and how each metric is defined. It builds on Measuring dwell time: four behaviour tiers, which explains how a zone reading comes about. This one is about what it changes in retail.

What the online shop has known for twenty years

The comparison is worth making because both channels ask the same question and only one of them gets an answer.

Nobody earned the right-hand column by being unusually careful. It exists because an online shop routes its visitors through a system anyway, and traces fall out of that whether you want them or not. In a physical room, that trace does not appear by itself - it appears when somebody measures.

So the gap is not a question of diligence but of equipment. And it does not close by looking harder at the till data: what is missing there was never in there.

The group that leaves no trace

Nobody in a store is less known than the people who leave without a receipt. They are in the room, they move, they stop in particular places - and all of that is measurable without anyone being identified. Why that is architectural rather than merely a matter of good conduct is covered in People counting and privacy.

Passage or pause: reading the tiers on the shop floor

How a zone visit is sorted into four tiers by its measured duration - transit, browser, shopper, committed - is covered in Measuring dwell time. This is about what the spread across those tiers means in a store.

The five-second line carries most of the meaning, and it is the one most often missing from a report. A zone with high footfall and a high transit share is not a strong spot in the store - it is a route. Judge it by footfall and you will put your best-margin range in a corridor.

The inverse is just as useful. A zone with low footfall but a high shopper and committed share is an area that works and that too few people reach. That is a routing problem, not a range problem - and the two call for opposite responses.

A third pattern is more common in retail than people expect: plenty of browsers, few shoppers. The area catches the eye but does not earn the attention. That can mean price or offer are unreadable from two metres away - a design problem, and the cheapest of the three.

The tier measures attention, not intent to buy. “Committed” means this area held somebody for two minutes. What becomes of that is decided at the till.

Reach against holding power: the four fields

Two metrics per zone answer two different questions. Capture rate says how many the area reaches. Engagement rate says how many of those it held. Both are defined in Measuring dwell time, including why capture rate can exceed one hundred per cent.

On the shop floor they only become useful once you plot one against the other. That yields four fields, and each calls for a different response:

The hidden gem is the most valuable field, because it is the cheapest to fix. An area that holds people needs no new stock and no refit - it needs a reason to walk past it. A wayfinding sign, a moved secondary placement, a sightline from the entrance.

The main aisle, by contrast, is where most bad decisions are made. It looks strong in every footfall report, which is exactly why things that need calm end up there.

Add to that the funnel across the tiers: of all zone visits to an area, how many got past transit, how many past browser, how many reached committed? The step where the funnel drops most steeply is where the area loses people.

The route through the store

The floor plan is designed, the walking route is real, and the two rarely agree. The zone visits of one session form a sequence, and many sequences form a pattern.

Three readings carry most of the weight:

• The ten most frequent paths. Not which zones are strong, but in what order they are visited. A strong zone immediately after another strong zone is a cannibalisation risk, not a double win.
• Entry and exit zones. Which area somebody sees first and which last. The exit zone is where a secondary placement still achieves something - and it is rarely the one people assume.
• Bounce per zone. The share of visits that begin and end in the same zone without a second one joining.

What is recorded is the sequence of zones: that somebody went from zone A to zone B. Positions on the floor plan are deliberately not stored for this - for data protection reasons, and the question of entry, order and exit does not need them. Zones can be renamed and recut afterwards; the analysis follows the new definition.

The campaign that was more discount than more people

After every promotion there is a number marketing likes and a number finance likes. There is rarely one that shows whether more people actually came.

The setup for it is plain: you fix a baseline window and a campaign window. Visitors and groups are compared, with daily values and a significance block.

The significance block is the part that goes missing when this is done by hand. Two weeks that differ by four per cent do not differ at all in a business that fluctuates - you can only see that if the spread is part of the calculation. A campaign review without that step reliably produces successes, because any difference looks like a result.

The comparison is only as good as the choice of the two windows. A promotion in the week before Christmas is best set against the same week a year earlier rather than the week before it - which is exactly why counts are kept for two years. And two windows of equal length with the same weekdays compare trading with trading, not Saturday with Tuesday.

Staff where the footfall is

Rotas follow habit: thin on Monday morning, full on Saturday. Whether footfall still supports that is rarely checked - until the Saturday is quiet and the Thursday evening overflows.

A recommended headcount can be derived from footfall per weekday and hour. The arithmetic behind it is deliberately conservative: it uses the 75th percentile, not the mean.

That is a decision, not a detail. A mean plans roughly half of all hours short, because roughly half of them sit above it. The 75th percentile plans so that three in four comparable hours are covered, and knowingly accepts the fourth. Cells with too little data are flagged rather than smoothed.

On top of that comes the metric that ends retail arguments fastest: visitors per employee per day. It sets the headcount you record per shift against measured footfall and makes under- and overstaffing visible against your own thresholds.

And the chain ends at the till: the queue is the moment somebody with a full basket decides whether to stay. How length, wait time, service time and throughput are measured there is covered in Measuring queues and wait times.

Measuring the refit instead of arguing about it

In most businesses, category decisions rest on sell-through and the occasional piece of market research. What actually happens on the floor - whether a new placement draws more people or merely redistributes the same ones - goes unobserved in between.

The setup that changes this is simpler than it sounds: every category becomes a zone. After that, every category carries the same four metrics as any other area - capture rate, average dwell time, engagement and visits - and every layout change turns into a before-and-after across the same zone cut.

What makes the comparison defensible are the same two rules that apply to any zone reading: zones with too few visits are flagged rather than ranked, and a zone with no previous period shows that as a state of its own rather than as zero. In category management, that decides whether a refit is judged on eleven observations or on eleven hundred.

The typical first finding from such a comparison is uncomfortable: a change that looks like a win in sell-through often did not bring more people to the category but held the same ones longer - or the reverse, drew more people and held them worse. Those are two different results with two different follow-up actions, and in revenue they look identical.

What turns this into a decision

The metrics above share one property, and it is the real difference from a till report: they describe a state you can change.

A drop in revenue says something happened. A rise in the transit share in your best-margin category says what happened - and the response to it is a floor decision, not a pricing decision.

Which is why the first sensible question to ask of an installation like this is not “how many are coming” but: where in this store are we losing people who are already here?

The order to start in

An installation delivers more readings on day one than anyone reads in the first week. The sequence that works in retail therefore runs from the most defensible number to the most interesting one, not the other way round.

1. Footfall at the entrance. One number, one definition, no precondition beyond the mounting. It is the denominator for everything that follows, and it settles the argument about whether a day was good.
2. Three to five zones, cut coarsely. Not thirty. Entrance area, two category areas, till zone. Start with a fine cut and you get many zones with too few visits - and therefore many numbers that get flagged rather than read.
3. Let it run four weeks before deciding anything. Before that the weekday comparison is missing, and without it every fluctuation looks like a trend.
4. Only then recut. By now it is visible which zone is too large and where a boundary runs straight through a walking route. The new, finer zones start their series with the new cut; footfall at the entrance runs on without a break.
5. Routes last. Paths and entry zones are the reading with the most surprises and the greatest dependence on a clean zone cut. It belongs at the end, not the start.

The most common mistake in that sequence is step two. A fine zone cut looks thorough on the plan and produces a reading in which half the areas sit below the sample threshold. Starting coarse and sharpening deliberately costs four weeks; starting fine with numbers nobody believes costs the first few months.

How we do it

The ANALYSIT Counting System counts entries by the minute and keeps them for 730 days - at any resolution from the minute to the day, with peak hour, peak day and peak month marked automatically.

Per zone it delivers average dwell time, engagement rate, visits and capture rate, plus the distribution across the four behaviour tiers with cumulative funnel, opportunity matrix, intraday curve, weekday pattern and previous-period comparison. Routes through the store are available as a zone-to-zone reading: the ten most frequent paths, entry and exit zones, bounce per zone, with export.

For promotions there is the period comparison with its significance block and export as CSV, XLSX and PDF. For planning, the staffing heatmap from the 75th percentile and visitors per employee per day. For the till, live status per queue with measured throughput.

If you want to know which of your areas is passage and which one holds, talk to us. How the zones are cut is half the outcome - and they can be recut later.