Turning aisle traffic density into a number your operations team can act on

The radio call came at 14:20 on day two. Aisle 400 is heaving, can somebody do something. The show director wants a decision in five minutes, and the only figure anybody can produce is a hall total from the door counters, which says 6,200 people are inside. That total cannot answer the question, because it says nothing about where the 6,200 are standing.

Aisle traffic density answers it. Count the people inside a defined aisle polygon, divide by the walkable area of that polygon, and you have persons per square metre. One division, done on a phone, and the argument becomes a grade that two people can agree on.

The denominator is where the errors live

Counting people is the part everyone worries about, and it is the easier half. Whether the count comes from an overhead sensor, a camera or two staff with clickers, you end up with an integer. The half that goes wrong is the area you divide by.

Take an aisle block drawn on the floorplan as 12 metres by 20 metres. That is 240 square metres of gross block. On the actual floor, the exhibitors either side have brought their build out to the booth line, there is a stanchion run protecting a cable tray on one side, and a coffee cart has taken two metres of frontage. The walkable width is nearer 8 metres than 12.

Same count, two denominators. At 320 people across the drawn 240 square metres you get 1.33 persons per square metre. Across the walkable 160 square metres you get 2.00. Those two numbers land in different bands and imply different actions, and the only thing separating them is how the polygon was drawn.

So fix the polygon before you fix anything else. Draw it to the walkable surface, exclude structure, and store the area as an attribute of the zone rather than recomputing it from the floorplan every time. If your zone areas come from the CAD file, check three of them against a measuring wheel on build day. Two of them will be right and one will surprise you.

Working one aisle through

The aisle 400 block is 12 metres of drawn width by 20 metres of run, and a walk of it on build day gave 10 metres of clear width once the stanchions and the cart were in. Walkable area is 10 times 20, which is 200 square metres.

The overhead count at 14:20 is 320 people. Density is 320 divided by 200, which is 1.60 persons per square metre.

Fruin worked in the reciprocal, which he called the pedestrian area module: square feet of space per person. The reciprocal of 1.60 is 0.625 square metres per person. One square metre is 10.764 square feet, so 0.625 square metres is 6.73 square feet per person.

Now compare that with aisle 700, where the radio has been quiet. Aisle 700 is 6 metres wide by 30 metres long, so 180 square metres, and the count there is 260. Density is 260 divided by 180, which is 1.44 persons per square metre, or 7.47 square feet per person.

Aisle 400 has 60 more people in it and is 11 per cent denser. Aisle 700 has fewer people and is in the same operational band. Both readings are worth acting on, and a headcount ranking would have put aisle 700 out of sight.

Grading the result against Fruin's levels of service

John J. Fruin, then at the Port of New York Authority, published the walkway level of service framework in 1971 in Highway Research Record 355, and the bands have held up for fifty years because they are grounded in observed walking speed and observed flow rather than in opinion.

His walkway bands, in square feet of area per pedestrian, are level A at 35 or greater, B from 25 to 35, C from 15 to 25, D from 10 to 15, E from 5 to 10, and F at 5 or less. Converted to persons per square metre, the same bands read:

Aisle 400 at 6.73 square feet per person is level E. Fruin described level E as the range in which virtually all pedestrians have their walking speed restricted, forward progress at the lower end is made by shuffling, and design volumes of 20 to 25 pedestrians per minute per foot of width approach the maximum the walkway can carry. He recommended designing to that level only for short peaks in the most crowded areas. Aisle 700 at 7.47 square feet per person is also level E, nearer the middle of it.

That is the sentence the show director needs. Both aisles are running at a level Fruin regarded as acceptable only in short bursts, and it is 14:20 with two and a half hours of open floor left.

Why does density beat a headcount for an operations decision?

Because every action available to the operations team changes either the numerator or the denominator, and the density is the only figure that responds correctly to both.

Opening a cross-aisle adds walkable area, which lowers density without moving a single person out of the hall. Moving a theatre session start time by fifteen minutes changes the numerator in one block and raises it somewhere else. Closing a stand's queue and giving it a timed entry converts a static blockage back into walkable area. A headcount responds to exactly one of those and misreports the rest.

There is also a safety reason to hold the density rather than the count, and it comes from the shape of the curve rather than from any single threshold. Helbing, Johansson and Al-Abideen analysed video of the crowd disaster at Mina near Makkah on 12 January 2006 and reported in Physical Review E in 2007 two sudden transitions, from laminar flow to stop-and-go flow and then to what they called turbulent flow, with the second transition producing sudden releases of pressure. The property worth carrying onto a show floor is that the change is abrupt. A crowd does not degrade smoothly from comfortable to dangerous, and a number that moves smoothly with headcount will not warn you.

What time window should the density be measured over?

An instantaneous count at 14:20 is a sample of a quantity that moves fast. Report it alone and you will get a false alarm every time a keynote empties into a single aisle for ninety seconds.

Use a short rolling window and report two figures from it. Take one-minute counts, then over a rolling fifteen-minute window report both the mean density and the highest one-minute density. Aisle 400 at a fifteen-minute mean of 1.28 and a one-minute peak of 1.60 is a busy aisle with a spike in it. The same aisle at a mean of 1.55 and a peak of 1.62 is an aisle that has been at level E for a quarter of an hour, which is a different problem and needs a different answer.

Fifteen minutes is a starting point rather than a rule. Pick the window from how long it takes your team to act. If the fastest available intervention is a stewarding move that takes ten minutes to arrange, a two-minute window generates alarms you cannot serve, and the operations lead will start ignoring them. The window that works is the one that matches the response time, and finding it is worth doing once, in a debrief, with last edition's data.

Where this stops

Density is a description of a moment inside a polygon, and it carries no information about why the moment is happening. An aisle at level E because a headline exhibitor is running a demo is a commercial success with a stewarding problem. An aisle at level E because the hall has one usable route between two halves of the floor is a floorplan defect that will recur every edition until somebody moves a feature area. The number is identical.

The measurement also degrades exactly when you need it most. Overhead counting accuracy falls as people move as a tight group, which is what a crowd at level E is, so the count that matters most is the one taken under the conditions the sensor handles worst. That error budget is what a lidar or vision counter actually delivers under load, and it belongs to C19. Treat any single reading above 1.5 persons per square metre as a prompt to send somebody to look rather than as a measurement to act on directly.

One more limit worth stating in the debrief. Density says nothing about direction. Two hundred people moving one way through an aisle and two hundred people meeting head-on produce the same figure and behave completely differently, because counterflow lowers the achievable speed at any given density. If your sensors record direction, keep it, and report the split alongside the density.

Reading a low density is a different exercise from reading a high one, and ranking the quiet aisles against the hall median is C16's subject. The hall-level version of the same question, how many people are inside the building at all, is running occupancy from entry and exit scans in C14. Comparing densities across zones of different sizes and opening hours needs the normalisation step in C17 before the comparison means anything, which is also the step that makes zone figures usable in the rest of your attendee analytics.

The first step takes an hour and needs no new hardware. Pick the four aisle blocks your stewards complained about last edition, walk them with a measuring wheel, and write the walkable area of each into your zone table. Then pull last edition's counts for those blocks, divide, and see which band each one was actually in at its worst quarter hour. If the answer is that all four were at level D and the complaints were about something else, you have learned something useful before spending a penny on sensors.