Footfall Counting Accuracy: What MAPE and Confidence Intervals Really Mean

Every people-counting vendor advertises a footfall counting accuracy figure — 98%, 99.5%, "market-leading" — and almost none of them tells you what it means or how they got it. For a shop owner acting on the number, that headline percentage is close to useless on its own. This guide explains, in plain language, the three ideas that actually describe accuracy — a stated error, MAPE, and confidence intervals — so you can tell a meaningful accuracy claim from a marketing one.

Because the underlying method is the same across every brand built on this engine, the full technical methodology lives once on our hub, StreetProof, and this page is canonical there. Here we translate it for the capture-rate question you care about. Start from the capture rate guide if you have not.

Footfall counting accuracy: why "98%" tells you almost nothing

A single accuracy percentage hides everything that determines whether it is true for your footage:

• Accurate under what conditions? Bright daylight and a thin, tidy pavement is easy. Dusk, rain, and a dense crowd where bodies overlap is hard. A number measured on the easy case does not carry to the hard one.
• Accurate at what? Total count over a day can look excellent while a specific busy ten minutes is off, because errors cancel out over long periods. The headline averages away the moment you might care about most.
• Measured against what? "Accurate" only means something compared to a true count. If nobody hand-counted the same footage to check, the percentage is an assertion, not a measurement.

This is why a good footfall report leads with evidence you can inspect — an annotated clip, raw counts — rather than a single hero number. If a claim gives you only the percentage, audit it before you lean on it.

MAPE: a more honest way to say "accurate"

MAPE stands for mean absolute percentage error. In plain terms: on average, how far off are the counts, as a percentage of the true value? A MAPE of 5% means that, across the checked periods, the count typically misses by about 5% — sometimes over, sometimes under.

MAPE is more honest than a headline accuracy badge for two reasons. First, it describes typical error, not a best-case demo. Second, it is symmetric about the real answer — it counts an overcount and an undercount as errors, whereas a friendly "accuracy" figure can quietly net them off. When you compare footfall providers, ask for MAPE and the conditions it was measured under. A provider who can give you that is describing their product; one who only has a shiny percentage is describing their marketing.

Key takeaways

• A single accuracy percentage is meaningless without the conditions, the metric, and the true count it was checked against.
• MAPE (mean absolute percentage error) describes typical error and counts over- and under-counts honestly.
• A confidence interval turns a sampled count into a range; more observed footage means a tighter range.
• Honest reports lead with checkable evidence and error bars, not a hero "99%" badge.

Confidence intervals: why your number is a range

There are two different uncertainties in any footfall figure, and it helps to keep them apart.

• Counting error — how well the software counted the people it could see. MAPE describes this.
• Sampling uncertainty — how well the footage you observed represents the period you want to know about. A confidence interval describes this.

The second one is the reason your estimate comes as a range rather than a single number. If we count what crossed your line over sampled footage and then project a typical week, we widen the figure to reflect that we watched part of the week, not all of it. That widened range is a 95% confidence interval: roughly, the band the true weekly figure is likely to fall within. Observe more footage and the band narrows; observe less and it widens — and we show the band rather than hiding it.

This is exactly why a one-day reading is labelled a spot reading. Its counting error might be small, but its sampling uncertainty for a whole month is huge — so honestly, no monthly projection is offered from a single day. Reading your audit shows where the interval and the confidence label appear on the report.

What this means for your capture rate

Your capture rate is a ratio of two counted numbers, so both kinds of uncertainty flow into it. In practice:

• Over a longer period, the rate is solid. Small counting errors on passers-by and entries largely cancel across a week, so a seven-day capture rate is trustworthy.
• On a short clip, treat it as directional. A single busy hour gives a real-feeling capture rate, but it is one hour in one condition — a signal, not a settled fact.
• Compare like with like. The fairest test is your own street before and after a change, over matched periods, so both readings carry similar uncertainty and the difference is the signal.

None of this is a reason to distrust counted footfall — it is the opposite. A number that shows its error bars is far more trustworthy than a naked "99%," because it is telling you the truth about what it knows and what it does not.

Measure it and see the ranges yourself

The best way to understand error bars is to look at your own. Get a $99 Capture Rate Audit and read the confidence label on it, or see pricing for a seven-day audit whose weekly estimate comes with a proper interval. For the full accuracy methodology and how it is validated, read the canonical StreetProof guide.

Frequently asked questions

What does footfall counting accuracy actually mean? It means how close a counted number is to the true number of people. A single "98% accurate" badge is nearly meaningless without the conditions it was measured under. A more honest expression is MAPE — the average size of the miss — plus a confidence interval around any projection.

What is MAPE in footfall counting? MAPE is mean absolute percentage error: on average, how far off the counts are, as a percentage. A MAPE of 5% means the counts miss by about 5% on average. It is more useful than a headline accuracy figure because it describes typical error, not a best case.

Why is my footfall estimate shown as a range? Because a projection from sampled footage is uncertain, and an honest report shows that uncertainty as a confidence interval. A narrower range means more footage was observed; a wider range means less. A single point value with no range is hiding its margin of error.