Finding dead zones on a show floor before the exhibitors in them complain

The email lands on the Tuesday after the show closes. Three exhibitors from the 900 aisle, copied to each other, saying the same thing in slightly different words: nobody came down our aisle, and we want a different position next year at the same rate. The sales director forwards it and asks whether they are right.

Dead zones on a show floor are answerable, and the answer takes an afternoon if you kept door and aisle counts. The awkward part is that the answer has two halves. One half is a ranking, which tells you whether the 900 aisle really was the quietest run in the hall. The other half is a cause, which decides whether anybody can do anything about it.

The rate, and the frontage it divides by

A raw pass count per aisle is useless for comparison, because a 40 metre aisle and a 12 metre aisle are different objects. The comparable figure is traffic per linear metre of exhibiting frontage: the number of passes recorded at a mid-segment line, divided by the metres of stand frontage that segment serves.

Frontage is the part people get wrong. A double-sided aisle 30 metres long serves 60 metres of frontage. A single-sided aisle of the same length, with a wall or a feature area opposite, serves 30. Divide both by their length and you have compared two different things.

Take the 900 aisle. It is 30 metres long, single-sided, with a hall wall on the far side, so 30 metres of frontage. Over three open days the counter at its mid-point logged 3,180 passes. That is 3,180 divided by 30, which is 106 passes per linear metre of frontage.

Now the thing that makes single-sided aisles worth flagging separately. If the 900 aisle had stands on both sides, the same 3,180 passes would be spread over 60 metres of frontage, giving 53 per linear metre. Being single-sided flatters the rate. The 900 aisle looks twice as good on this measure as its traffic would justify if it were a normal aisle, and it is still the worst in the hall.

Ranking against the median, and reading the spread

The hall has 22 aisle segments. Sorted by passes per linear metre over the three days, the median segment sits at 259 and the busiest at 512.

The 900 aisle at 106 is 106 divided by 259, which is 0.409. It carries 41 per cent of the hall median. The busiest aisle carries 512 divided by 106, which is 4.8 times as much traffic per metre of frontage as the worst one.

Use the median and leave the mean alone. Aisle traffic distributions are right skewed, because a hall usually has two or three runs that everybody walks and a long tail of ordinary ones. In this hall the mean is 288 against a median of 259, so anything measured against the mean looks worse than it is, and the effect is largest for exactly the aisles you are investigating.

The next question is whether 41 per cent of the median is abnormal or merely last place, and those are different findings. Somebody has to be last. The standard tool for the distinction is the box plot rule set out in the NIST and SEMATECH e-Handbook of Statistical Methods, published by the National Institute of Standards and Technology in 2020, which defines the interquartile range as the third quartile minus the first, puts inner fences at 1.5 interquartile ranges beyond each quartile, and calls a point beyond an inner fence a mild outlier.

Run it on the 22 segments. The first quartile is 214 and the third is 338, so the interquartile range is 124. One and a half of those is 186, and the lower inner fence is 214 minus 186, which is 28 passes per linear metre.

The 900 aisle at 106 is nowhere near 28. It is the quietest aisle in the hall by a wide margin and it is comfortably inside the hall's normal spread. That is a genuinely useful thing to be able to say to three angry exhibitors, and it is the opposite of what the ranking alone implies. A hall where the worst aisle runs at 41 per cent of the median is a hall with an unremarkable amount of variation in it.

If a segment does fall below the fence, the finding is stronger and the follow-up is different. Then you are looking at something structural, and it is usually visible on a walk.

Does a quiet aisle mean weak exhibitors or a bad position?

Almost always the position, and there is a body of work that says so.

Hillier, Penn, Hanson, Grajewski and Xu published the argument in Environment and Planning B in 1993, in a paper running from page 29 to page 66 of volume 20. They set out what they called a configurational account of pedestrian movement, in which the structure of the movement grid itself privileges some routes over others and generates much of the flow pattern before any destination is considered. Their measure of integration, which describes how easily a given line of movement is reached from every other line in the system, turned out to be the strongest single correlate of observed movement rates. Their reading of retail location follows from that: shops tend to appear where movement already is, so they amplify a pattern the grid produced rather than creating it.

A hall floorplan is a movement grid with the same properties and a much simpler geometry. An aisle that is reached only by turning off another aisle, that has one open end, or that sits behind a feature area, is poorly integrated in exactly the sense the 1993 paper describes. Its traffic will be low regardless of who is standing in it.

So the diagnostic step after the ranking is a walk, ideally with the floorplan and the ranking in the same hand. The causes worth writing down are the ones that recur: a single open end, a blind corner where an aisle meets a hall wall, a run whose entrance is masked by a large stand build, a feature area or catering block that absorbs the traffic that would otherwise pass through, and a position that requires doubling back from the only door people use.

Two of those are yours to fix on the floorplan next year. The rest need a pricing answer, and pricing quiet space is its own decision with its own arithmetic in F19.

Why does the reading have to hold on every day?

Because a single-day figure picks up the conference programme, and the conference programme moves.

An aisle can read low on day one because a keynote ran late and low on day three because everybody left for flights, and neither says anything about the aisle. The 900 aisle in this example reads 41 per cent of the median across all three days, which is the finding. Per day it runs at 39, 44 and 40 per cent, which is the same finding three times.

Set the rule before you look. A segment qualifies as a candidate only if it sits below some fixed share of the hall median, say 60 per cent, on every open day. Anything that qualifies on one day and not the others goes on a separate list called timing, and that list is usually more actionable than the dead zone list, because timing problems can be fixed with a programme change.

The stability test also protects you from the sensor. A counter that dropped out for two hours produces a spectacular dead zone on one day and a normal reading on the others.

Where this stops

The whole method measures passes, and a pass is a weak proxy for what an exhibitor is complaining about. Somebody walking the 900 aisle at speed on the way to the toilets counts the same as somebody walking it slowly looking at stands. Two aisles with identical traffic per linear metre can produce very different scan volumes, and the gap between passing and stopping needs dwell data, which is a different instrument and a different post.

The second limit is that the ranking has no counterfactual in it. You know the 900 aisle carried 41 per cent of the median. You do not know what it would have carried with a different tenant mix, a different entrance or a feature area at its far end, and no amount of counting last year's traffic will tell you. The only honest way to find out is to change one thing and measure again, which means a hall where the floorplan moves is a hall you can learn from and a hall where it never moves is one you cannot.

The third is the denominator itself. Frontage metres come off the floorplan, and floorplans drift from the built hall. Check three segments with a measuring wheel on build day before you publish a ranking that will be used in a rate conversation.

None of this survives contact with a badly drawn zone map either. If your aisle segments do not have stable boundaries between editions, the ranking compares different objects year to year, and the same problem sits underneath any comparison of named zones by area and open minutes in C17. It also sits underneath the picture everyone actually looks at, which is the gridded heat map and the four choices behind it in C15. Fix the zone map once, before the show, and every measurement in the attendee analytics pack gets easier.

The first step needs last edition's counts and a floorplan. Compute passes per linear metre of frontage for every aisle segment, sort them, and write the ratio to the hall median next to each one. Then walk the bottom five on the floorplan and write one sentence about why each is quiet. If four of the five sentences say the same thing, you have found a floorplan problem rather than five exhibitor problems.