Footfall going up or down? How to read a trend without fooling yourself

A footfall trend is only worth acting on when it survives four checks: it is bigger than the site's usual week-to-week swing, it is not shared by every other site in the same city, nothing obvious happened in those dates, and it repeats the following week. Most single-week moves fail one of these. The ones that pass, a store whose surroundings get steadily quieter while its neighbours hold steady, are exactly the early warnings a network manager wants.

This piece is for regional and network managers, landlords tracking their centres, and analysts who receive a weekly footfall table. It explains what the trend figure compares, the four checks in detail, an example across a network in Buenos Aires, and an AI employee that applies the checks before anything lands in your inbox.

What the trend figure compares

Mobility Footfall gives each place footfall_trend: the average daily footfall of the last 7 days against the 7 days before, as words ("+6% on the week before") and as a number in footfall_trend_pct. It needs at least 14 days of data. Both weeks end about ten days before the run, since each day arrives late.

Footfall is people-hours of presence around the place, so a rise means more time spent nearby by everyone: shoppers, workers, residents. It does not mean more customers, and a fall does not mean fewer. It is a signal about the street, which you then compare with your own sales.

Four checks before you believe it

1. Is it bigger than the usual swing?

Every place wobbles. A corner that normally moves between minus 4% and plus 5% a week has not changed when it shows plus 5%. Keep a few weeks of each site's trend and judge a new figure against that range, not against zero.

2. Is the whole city moving?

If every one of your sites in a city fell by a similar amount, the cause is probably city-wide: weather, a holiday, school holidays, a transport strike. A drop at one site while the others are flat is the signal. Comparing each site with the median of its city takes this out.

3. What happened in those dates?

Check footfall_from and footfall_to and ask what fell inside: a public holiday (see how footfall changes on holidays), a festival, a big match, a road closure, building works on the block, a neighbour's opening or closing. A one-off reason means the trend will likely reverse.

4. Does it repeat?

A real shift, such as a large employer moving out or a new transport stop opening, shows up two, three and four weeks in a row. Wait for the second week before you change staffing or stock, unless the cost of waiting is high.

An example across a network in Buenos Aires

A made-up case. A chain of eyewear shops has six stores in Buenos Aires and gets the weekly trend for each. Suppose one Monday the table reads (invented for illustration):

Invented figures for illustration; not real figures for Buenos Aires.

Every store fell, so the first reading is a city-wide week; the median is about minus 6.5%. Store 5 fell three times as much, far outside its range. The manager finds the street outside was closed for works during most of those dates. Next week Store 5 is back within its range, and nothing needed changing. Store 1, a little below the median, is worth watching: if it is still lagging in two weeks, something local is going on.

Mistakes to avoid

• Reacting to one week. The commonest mistake, and the most expensive.
• Comparing trends across circle sizes or windows. Keep every site on the same radius and the same weekdays or weekends setting.
• Treating the trend as sales. Footfall can rise while sales fall, if the new crowd is not your customer.
• Ignoring the lag. The weeks compared ended about ten days ago. A change you can already see on the street will reach the figures later.
• Places with no data. "No footfall data here yet" is not a fall to zero. Footfall does not cover Australia yet, and some places in covered countries have gaps.

An AI employee that applies the checks

1. Start with Schedule, every Monday.
2. Bring in your stores with My data, with their city.
3. Add Mobility Footfall with Footfall and Trend, 200 m, the last 14 days, the same day setting for all.
4. Add xMap LLM set to work on the whole list, asked to compare each store's trend with the median of its own city and name any store well outside it, with the dates covered.
5. Add Branch: when the report names an outlier, send it; otherwise send a one-line all-clear.
6. Send it by Email. In the inbox: the city medians, the stores outside them, how far, and the dates to check.

Keep each week's figures in a sheet as well, so the usual range for every store builds up on its own and the second-week check is a glance.