Pedestrian Counting for City Centres and High Streets

A Saturday street market closes the main shopping street to traffic. The business improvement district reports footfall up 14% against the previous Saturday. Two independent retailers on the side street say their tills were quieter than a wet Tuesday. The landlord of the corner unit asks for the data to support a rent review. Everyone is quoting the same single number, taken from one footfall counter mounted outside the station, and nobody can say which of them is right.

That is the normal state of pedestrian counting in most town centres: one or two sensors, one headline figure, and a dozen stakeholders reading into it whatever they already believed. The hardware is rarely the problem. The problem is treating a street network like a shop with one door.

One counter at the station is a statistic, not a strategy

A city centre has entry points, through-routes, destination streets and dead ends. A single counting point tells you how many people passed that spot. It cannot tell you whether they turned left towards the department store or right towards the car park, whether the new food hall pulled people from the old high street or added to it, or how far along a street footfall drops off before the vacant units start.

The deployments that actually settle arguments use a network of counting points chosen deliberately:

• Gateways – station exits, bus interchange, multi-storey car park lifts, the footbridge from the retail park. These give you total arrivals.
• Decision points – the junction where the street splits, the entrance to the covered arcade, the mouth of the side street. Bidirectional counts here show route choice.
• Decay points – 50, 100 and 150 metres along the secondary street. This is where you learn precisely where a street stops working, which is the number landlords and planners most need and least often have.

With that structure, the market-day question answers itself. Gateway counts rose, the decision-point count at the side street fell, and the two independents were right. The BID was also right. Both things were true, and now there is a basis for rerouting the market stalls next time rather than a shouting match.

What a people counting system has to survive outdoors

Indoor counting is a controlled environment: fixed lighting, a ceiling to mount on, a doorway of known width. Streets offer none of that. Sensors have to deal with rain on the lens, low winter sun straight into the optics at 3 pm, umbrellas and hoods that change the shape of a person, cyclists and pushchairs that should be classified differently, and a counting zone that might be twelve metres wide with no walls to channel anyone.

Modern 3D and AI sensors from manufacturers such as Xovis are built for exactly this, and VemCount is sensor-agnostic so the same platform can mix outdoor weather-rated units on wide squares with standard indoor devices at the arcade entrances. But the device is the easy part. Three practical issues catch out almost every first-time deployment:

• Lamp column power is usually switched. Many street lights are fed through a photocell or a central timer, which means the column has power at night and none during the day – precisely the hours you want to count. Any implementer who has done this twice will insist on an unswitched feed or a separate supply before a single bracket goes up. Agreeing this with the highways authority takes longer than installing the sensors.
• Counting lines near bus shelters and benches. Draw the virtual line through a waiting area and you will count the same person every time they shuffle. Lines go across the walking channel, not across the place people stand.
• Mounting height and vibration. Columns sway in wind and shake when buses pass. Sensors need rigid brackets and a mounting height that matches the manufacturer's field-of-view specification for the street width, or the count area quietly shrinks and you under-report the kerbside flow.

Honest accuracy, and knowing when a sensor has gone quiet

Vemco commits contractually to a minimum of 96% counting accuracy. In practice, installations typically run at 98 to 99%, but only where conditions allow – good lighting, a sensible mounting position and reasonably predictable pedestrian behaviour. Outdoor sites test those conditions harder than a shop door does, which is why the contractual floor matters more here than the typical figure. Anyone selling you a flat 99% for a rainy, unlit square in January has not done one before.

The bigger risk in a street network is not a sensor counting slightly low. It is a sensor counting nothing for three weeks because a contractor cut the power, and nobody noticing until the quarterly report shows a street that apparently emptied. VemCount's sensor health monitoring flags devices that stop reporting or whose counts drift outside their normal pattern, so the gap is spotted on day one rather than at the board meeting. For an IT director responsible for a hundred devices across a town centre, that alerting is the difference between a manageable estate and a permanent trust problem with the data.

Who uses high street footfall data, and for what

The same count feeds very different decisions depending on who is reading it.

Retail directors want street footfall alongside their own door counts and POS data. If the street outside is up 8% and transactions are flat, the problem is capture rate – window, signage, entrance – not demand. If the street is down 8% and transactions are flat, the store is outperforming its location and the regional manager deserves to know. Without the street layer, both scenarios look identical on a sales report.

Mall operators and landlords use decay-point counts in leasing. A unit 120 metres from the main flow with a quarter of the passing footfall of the anchor pitch cannot sustain the same rent, and having the number removes the emotion from the conversation. Over time it also shows whether a new tenant or a public-realm scheme has actually shifted the decay curve.

Facility managers schedule against hourly profiles. Cleaning rounds, bin collection and security patrols timed to actual peaks rather than a fixed rota is an unglamorous but immediate saving, and occupancy data for covered arcades and squares supports event capacity decisions.

IT directors care about the plumbing: whether the platform is hosted or on private cloud, how the counts reach the BI tool, who owns the data when a BID contract changes hands, and that no images leave the sensor. Visitor counting on public streets attracts scrutiny, and being able to show that the sensors output only anonymous count data – not video – shortens most privacy conversations considerably.

A deployment checklist that holds up

• Map gateways, decision points and decay points before choosing hardware.
• Confirm unswitched power and connectivity at every column – in writing, with the authority that owns it.
• Validate each sensor with a manual count across at least one wet day and one bright day.
• Set health alerts and name the person who receives them.
• Agree the reporting cadence and the definitions – especially what counts as "the high street" – with every stakeholder before the first dashboard goes live.

Frequently asked questions

How is pedestrian traffic counted? Most city-centre schemes now use 3D or AI camera sensors mounted on lamp columns or building facades, which detect each person crossing a virtual line and record direction without storing images. The counts are sent to a platform such as VemCount, which aggregates them by location, hour and direction and checks that every sensor is still reporting. Older methods – manual clicker surveys, pressure mats and infrared beams – are still used for spot checks but struggle with wide streets and groups walking side by side.

If you are planning a footfall counter network for a town centre, high street or outdoor retail district and want to get the sensor positions, power and reporting structure right first time, talk to Vemco Group about a site review before the brackets go up.