How to Measure Footfall Before Investing in Public-Realm Improvements

Start from the decision, not the count

Before choosing equipment, decide which decision the number will feed: does the average weekday flow justify widening a sidewalk, adding seating, opening a passage, or building a plaza? Different decisions need different metrics. Flow ('how many people cross an entrance per day'), occupancy ('how many are present right now'), dwell time ('how long do they stay'), and queue length ('how many are waiting') are measured differently, and a system that counts entries accurately can report occupancy badly.

Set the go/no-go threshold before fieldwork: at what average figure will you invest, and below what will you not? That converts the count from a study into an engineering input. For cities that have never counted people, public-life programs emphasize that the first pedestrian counts are the most important, because they establish the baseline for every comparison that follows — capture it even if a single day's number seems unremarkable.

• Flow: crossings of an entrance or screen line per period
• Occupancy: people present on the space at one time
• Dwell: how long people remain and what they do
• Queue/wait: congestion at a gate or narrow point

Define the counting zone and your screen lines

On a map, fix the physical boundaries of the study area: every entrance and exit, mid-block locations, intersections, and magnets such as transit stops, shops, schools, and parks. For a plaza or park this means its perimeter and main approaches; for a street, run a screen line across the sidewalk at mid-block, where flow is stable rather than distorted by the intersection.

Separate transit from stay: who merely passes through versus who stops, sits, stands, and interacts. Then define what counts as a 'visitor.' In practice the largest disagreement between a system and a manual audit is usually staff, couriers, and through-travelers who never stop — a definition problem rather than a sensor problem. Write the rule into the protocol so surveys taken in different years remain comparable.

Choose a counting method and know its bias

Manual in-field counts by trained observers with clickers or an app are accurate and cheap for a short baseline, and they double as the reference for validating automated equipment. Automated methods — infrared beam, time-of-flight (ToF), thermal, stereo or camera vision, and Wi-Fi/BLE probing — give continuity but each carries a systematic bias: people walking abreast can register as one; an oblique camera angle stacks people behind each other; backlit entrances at sunrise and sunset degrade video analytics.

Wi-Fi and Bluetooth sensors count devices rather than people, so a phone left at home or switched off is an error, and device identifiers are treated as personal data in many jurisdictions. A thermal sensor that outputs only a count is anonymous by design. The governing rule is validation: compare any automated counter against a manual count in your busiest hour and worst lighting, because accuracy under those conditions is the number that matters, and re-validate after the camera or unit is moved.

• Infrared beam: cheap and simple, no direction, weak with groups
• Time-of-flight overhead: accurate directional counting, narrow coverage
• Thermal: works in darkness, inherently anonymous, weaker in heat
• Stereo/camera vision overhead: accuracy plus context, needs mounting
• Wi-Fi/BLE: low hardware cost, counts devices, not people

Sample representatively, then extrapolate honestly

One count on a single afternoon is close to useless because flow swings by day of week and time of day. Plan surveys in the same slots — a working weekday and a weekend, morning and evening peaks plus an off-peak period — and repeat them across a season or several typical weeks, recording weather, holidays, and events. Because activity patterns differ sharply between a central business district and residential or mixed neighborhoods, tune the schedule to the local rhythm.

When resources are limited, take a short series of surveys at the same points and state plainly that this is a sample, not an annual figure. Archive the raw data, conditions, and protocol: it is the comparability of repeat surveys on the same days and hours, not the absolute precision of any single day, that makes the data usable for evaluating the effect of an improvement.

Read beyond the count: latent demand, context, and models

Current pedestrian volumes almost always understate real demand because poor conditions suppress it: gaps in the network, unsafe crossings, and uncomfortable streets deter people who would otherwise walk. So alongside counting, assess potential. Plot the destinations that attract pedestrians — transit stops, building entrances, schools, shops, parks — and the density and mix of uses, since proximity of destinations and dense mixed land use are among the strongest drivers of walking.

For areas you cannot survey physically, use a sketch-plan approach or a network movement model: estimate likely journeys between origins and destinations, then calibrate the estimate against real field counts on a limited set of streets. Practice shows that once calibrated on a few dozen segments, estimates can be extended to hundreds or thousands of streets, including not-yet-improved ones where few people walk precisely because of the current environment.

• Sketch map of destinations and likely routes between them
• Network or agent-based pedestrian movement model, calibrated to field counts
• Land-use density and mix as drivers of trip generation
• Barrier audit: network gaps, crossings, lighting, comfort

Validate, document, and plan the repeat

Validate automated counters against manual counts under your worst real conditions — busiest hour, worst light, widest entrance — and repeat the check after any repositioning. Keep a field log: dates, hours, weather, special events, and any change on site such as temporary construction, closures, or a market, because without that log an anomalous day will masquerade as a trend.

Schedule the post-occupancy repeat before construction even starts, and run it on the same days of the week and hours as the baseline, ideally in the same season, so seasonal swings do not masquerade as project impact. It is the paired before-and-after under an identical method — not any single flattering number — that shows whether footfall rose, dwell lengthened, and how life on the space actually changed.

Turn the numbers into the investment call

Collapse the results against the threshold set at the start: average weekday and weekend flow, share of people who stop, dwell time, and congestion at pinch points. Benchmark these against comparable streets or plazas in your city to judge whether the figure is normal or exceptional. Respect the limitation that a count does not explain 'why': quality, safety, comfort, and the reasons people come are revealed by observation, stationary mapping, and short user surveys, not by a single sensor.

Finally, treat the figures as part of a broader business case. Rising footfall and dwell generally track the vitality of retail and the attractiveness of a district, but they do not guarantee it. Combine quantitative measurement with a qualitative place audit and genuine engagement with residents and businesses, and the decision about where to direct the public-realm budget will rest on verifiable data rather than intuition or lobbying.