Using footfall to choose where to put a cash machine
A cash machine earns from the people who pass it, when they pass. How to compare spots by footfall, busy hours and nearby ATMs, and what footfall cannot tell you.
Part of the guide Location intelligence automation
To choose a cash machine site with footfall, measure how busy a small circle around each candidate spot is, when in the day and week it is busiest, and how many cash machines already stand within a short walk. A spot that is busy at the hours people need cash, with few machines nearby, goes first. Footfall cannot tell you how many people will withdraw money, so it ranks sites against each other; it does not forecast transactions.
This is written for ATM network planners at banks, independent ATM operators and the retailers and landlords who host machines. It explains what footfall measures, which readings matter for a cash machine in particular, a made-up comparison of four spots, the flow that produces it, and the traps.
What footfall measures here
The Mobility Footfall block reports footfall as people-hours of presence around a place, scaled to the whole population. A person who spends two hours in a cafe beside your candidate spot counts twice; a person who walks past in a minute counts a small fraction. For a cash machine that skew matters: dwell areas such as food courts look busier than they are for passing trade, and corridors such as station exits look quieter than the stream of people they carry.
Two more limits. Each day's figures arrive about ten days later, so a seven-day window covers the seven days that ended ten days ago. And the data begins in mid-October 2026 and covers most countries but not yet Australia, so long windows hold fewer days for now and a spot in Sydney or Melbourne returns no footfall.
The readings that matter for a cash machine
- Footfall per day at a small radius. Up to 200 m the block reads the spot's own small block of the map, about 0.1 square kilometres; that is roughly the area a person would cross to use a machine.
- The busiest hour and how the day splits. Cash demand peaks at lunch, after work and late on weekend nights in many cities. A spot that is busy only at 10:00 on weekdays serves offices, not shoppers.
- Weekday against weekend. Run the block twice, once for weekdays and once for weekends, to see which spots hold up on a Saturday night.
- Machines already there. Nearby places counts places of the type ATM, and Bank or ATM, within a short walk of each spot, with the nearest one and its distance.
- The trend. The last seven days against the seven before shows a spot getting busier, useful once a site has a few weeks of data.
A made-up comparison of four spots
These figures are invented to show how the comparison reads; they are not real footfall. A bank in Istanbul has four candidate spots offered by landlords: a metro exit, a shopping street, a mall food court and a hospital entrance.
| Spot | Footfall a day | Busiest hour | Evening share | Cash machines within 300 m |
|---|---|---|---|---|
| Metro exit | 38,000 | 18:00 | 41% | 1 |
| Shopping street | 52,000 | 19:00 | 46% | 7 |
| Mall food court | 61,000 | 13:00 | 33% | 4 |
| Hospital entrance | 24,000 | 11:00 | 19% | 0 |
Made-up figures, to show how the columns are read. Footfall is in people-hours a day.
The food court looks busiest, but much of its footfall is people sitting down, and four machines already serve it. The shopping street is busy in the evening but crowded with seven machines. The metro exit has a strong evening peak and only one machine nearby, so it leads. The hospital entrance is quiet in people-hours, yet it has no machine at all and steady daytime demand from visitors; the team keeps it as a second choice for a different reason: service, not volume.
The flow that does the comparison
- Put the candidate spots in a table in My data, one row each with a name and position, or send them in from your own system.
- Add Mobility Footfall on Each incoming place, radius 200 m, window the last 30 days, returning Footfall, Busy hours and Busiest days.
- Add Nearby places on Each incoming place with the cash machine types, within 300 m, passing on each place with counts.
- Add a Calculated field for footfall per existing machine.
- Add Sort and limit on that field, highest first.
- Add an xMap LLM block that writes two sentences per spot from the figures, including the busiest hour and the nearest machine, and says when a spot had no footfall data.
- Send it with Email to the network team, with a map.
Run it once to compare the candidates. Then keep it running monthly on your existing machines as well: a machine whose surrounding footfall falls month after month is a candidate to move.
Where footfall misleads cash planning
- Reading people-hours as people. A crowded cafe terrace scores high on presence and low on passing customers. Look at the busy hours, not only the total.
- Comparing spots at different radii. Keep the same radius for every candidate; a wider circle always adds footfall.
- Treating no data as no people. A spot with no footfall data says so in footfall_note. It is unmeasured, not empty.
- Forgetting what cash is used for. Markets, taxi ranks and nightlife streets lean on cash more than offices do, and that is local knowledge the data does not hold.
In cities such as Mumbai or Lima, where the busiest streets change street by street, run the same flow on a dozen points along a corridor rather than three or four, and let the map show where the busy stretch begins and ends.
Questions
Can footfall predict how many withdrawals a machine will make?
No. It measures how busy the area is in people-hours. Use it to rank spots against each other, then compare with your own machines' transaction data.
How recent are the figures?
Each day arrives about ten days later, so the most recent days in any window are around ten days old. The data starts in mid-October 2026.
Which radius should I use for a cash machine?
Small: 100 to 200 m. That is the area a person would cross to reach the machine, and up to 200 m the block reads the spot's own block of the map.
Does it work in every country?
Footfall covers most countries but not yet Australia. Where a spot has no data, the block says so instead of returning zero.