Predictive analytics in retail stores is a phrase that sells data platforms to chains, and underneath it is one forecast an independent store can make with its own numbers on a sheet of paper: how many shoppers will come through the door next week, by day and by hour. Everything a store plans, the rota, the fills, the order to the wholesaler, is sized to that number, and most stores size it to last week's habit instead. This page is written for the owner or manager of a single store: where the footfall count comes from, how to turn a few months of it into a forecast for next week, what to adjust for and what not to, and how the forecast feeds the rota. It publishes no industry model and no benchmark; the forecast is worked from your own count.
The count comes first, and it is probably in the till
A door counter is the direct measure. Without one, the till gives a close proxy: transactions a day, divided by the share of visitors who buy, is the visitors. The conversion share is the store's own figure, taken by standing at the door for a few busy hours and counting who comes in against who pays, and it is stable enough over months to use. Twelve weeks of daily transactions is enough to start. The count by hour is the more valuable series, because the rota is written by the hour and a daily total says nothing about whether Saturday's problem is the morning or the afternoon.
Turning a few months of counts into next week's forecast
Take the same day of the week over the last several weeks and average it, weighting the most recent weeks more heavily. That gives a base for each day. Then apply what you know that the average does not: a holiday in the week, a local event, a school term starting, weather if your trade depends on it, and last year's count for the same week if the store is seasonal. That is the whole method, and it is what the platforms do with more data and less local knowledge. The forecast is wrong every week by some amount; the point is that it is wrong by less than a copy of last week, and that the error is visible and shrinks as the counts accumulate.
From the forecast to the rota and the fills
Footfall times conversion is the transactions the week will generate. Transactions divided by what one associate can serve in an hour is the selling labor the week needs, by day and by hour, and the rota is built from that plus the cover the doors need plus the back of house hours. On the stock side, the same forecast times the units per transaction for a fast line is the units the shelf has to hold between fills, which is what the planogram's depth is for. One forecast, two plans, and both are checked afterwards against what actually happened, which is how the next forecast gets better.
What not to forecast
Do not forecast conversion; measure it, because it moves slowly and a guess about it multiplies through everything. Do not forecast by line for slow lines, where a week's sales of three units carries no signal; forecast the category and let the planogram share it out. Do not buy a platform to forecast a single store's footfall until the store has a door counter and a year of hourly counts, because the platform is only as good as the series it is given, and a store that has the series can do the arithmetic on the free rota planner without it.
Questions people ask about predictive analytics in retail stores
How accurate should a footfall forecast be?
Within about ten percent for a normal week once a few months of counts are in, and worse in holiday weeks until there is a prior year to read. The measure that matters is whether the rota planned on the forecast ran short or long, and by how much, and that is a number to write down each week.
Do we need a door counter?
It helps, and the cheap ones are cheap. Without one the till count divided by conversion is a serviceable proxy for planning the rota, and it is what most independent stores use. Where the counter earns its price is in seeing visitors who did not buy, which is the beginning of understanding why.
Does Footfally forecast for us?
No. The rota planner takes the footfall you expect as an input and works the hours, the cost and the labor percentage from it. The forecast is yours, from your own counts by the method on this page; Footfally Pro keeps each week's expected and actual footfall against the rota so the error is visible and the next forecast starts from real numbers.