Footfall Sampling: How 7 Days of Video Becomes a Defensible Estimate

The most reasonable question a sceptical tenant can ask is: "you didn't film every second for seven days, so how is this number real?" The answer is footfall sampling done honestly — the engine counts every pedestrian in the footage it is given, scales the observed count up to the full period only where coverage has gaps, and prints exactly how much was projected and how confident that projection is. This guide explains, in plain language, how observed video becomes a defensible footfall estimate, so landlords and brokers can defend the number instead of hoping nobody asks.

Key takeaways

• Every pedestrian in the supplied footage is counted; sampling only ever fills gaps in coverage.
• Observed counts are scaled to the full period and travel with a confidence interval that reflects how much was seen.
• The interval models counting noise; representativeness (weather, seasonality) is handled by the confidence label, not hidden.
• A well-observed seven-day window projects tightly and reads high confidence; a short clip is labelled an indicative spot reading.
• Disclosure is the whole trick — a projection you can see the limits of is one a counterparty will accept.

Footfall sampling: counting first, scaling second

Start with what is not a sample: the count itself. For every frame of footage you provide, the engine detects and tracks each pedestrian silhouette and records each line crossing. Within the observed footage, nothing is estimated — it is a direct count of what crossed the line.

Sampling enters only when the observed footage does not cover the whole seven-day period continuously. If you filmed heavily on some hours and lightly on others, or missed the small hours entirely, the engine does not invent those gaps. It scales the observed count by the share of the period that was actually observed and reports the result as an estimate, not a measurement. That distinction — counted where seen, scaled where not, and always labelled which is which — is the backbone of a defensible certificate.

Why one day cannot speak for a week

Footfall is not a constant you can read once and multiply. It swings by:

• Time of day — a lunch peak, a commuter wave, a dead mid-afternoon.
• Day of week — an office street roars Monday to Friday and empties on Sunday; a leisure pitch does the reverse.
• Weather and season — a wet Tuesday and a sunny one are different streets.

This is exactly why a Footfall Certificate is built on seven days rather than a convenient afternoon. Seven days spans both weekend and weekday and averages out a single bad-weather blip, so the projected week reflects the address's real rhythm instead of one lucky or unlucky slice. It is also why we are blunt about short clips: a great-looking count from a single busy hour is not a week, and pretending otherwise is the oldest trick in footfall marketing.

The confidence interval, in plain language

Because part of the number is projected, the certificate attaches a confidence interval — the range you see around the estimate. Think of it as the honest margin: the count has natural variation, and the interval expresses how much the projected total could reasonably move given how much footage was actually counted.

Two properties matter for reading it:

1. More observation narrows it. A large, representative sample projects with a tight interval. A thin sample projects with a wide one. The width is information, not noise — it tells you how firm the centre figure is.
2. It models counting variation, not the calendar. The interval captures the statistical noise in the count. It does not, by itself, promise that the observed week is a perfectly typical week of the year — that is what seasonality caveats and the observation period on the header are for. We keep those two ideas separate rather than blur them into one falsely reassuring number. The underlying formula and what it does and does not cover are set out in footfall accuracy, MAPE and confidence intervals.

The confidence label and the spot-reading guard

On top of the interval sits a plain confidence label, and it is the single most important honesty feature in the whole product:

• Observe a full, well-covered seven-day window and the certificate reads high confidence, with daily and monthly projections shown.
• Observe only a modest window and it reads medium.
• Observe just a short clip and it reads low — presented as an indicative spot reading, with the daily projection withheld and no monthly figure at all.

That last rule is deliberate: the system refuses to turn seconds of footage into a projected day. If you ever see "spot reading — footage too short to project a day," that is the product protecting the certificate's credibility, and yours. A landlord who understands this can promise a tenant something rare — that the number was not inflated by projecting further than the footage allows. To see where these labels surface on the document itself, read how to read a Footfall Certificate.

A worked intuition (illustrative only)

Suppose the observed footage over a week captures a large, representative count of crossings. Scaling that to the full period gives the headline estimate, and the interval around it is roughly proportional to the square root of the count — so bigger, better-observed counts produce proportionally tighter intervals. The practical upshot for you is simple: the more continuous the coverage you supply, the narrower and more confident your certificate's numbers become. This is an illustration of the mechanism, not a figure for any real property — every certificate carries its own counted numbers and its own interval.

What this means for you

You do not need to be a statistician to present a certificate well; you need to know three things. First, the count is real for what was filmed. Second, the projection to a full week is disclosed and bounded, not invented. Third, the confidence label stops the document from overreaching. Together, those make the estimate defensible — the property word that matters when a tenant, a bank or a buyer starts asking questions. That is also the theme of why you should never trust an unverified footfall number, the ones with no sample disclosed at all.

Ready to produce a certificate whose estimate you can stand behind? Start from the Footfall Certificate guide and run one from the pricing page; brokers, your first listing is $149.

Frequently asked questions

Does the certificate count everyone or sample? It counts every pedestrian in the supplied footage, then scales observed counts up to the full period only where coverage has gaps — with the projected share and a confidence interval disclosed.

How can a projected number be trustworthy? Because it is disclosed, bounded and labelled: coverage scaling, a confidence interval sized to the count, and a confidence level that flags thin windows as indicative.

Why is more observation better? An hour cannot represent a day and a day cannot represent a week; more observation captures real variation, centring the estimate and narrowing its interval.

What is a spot reading? A very short observation. The system labels it low-confidence and indicative and withholds daily and monthly projections, so a tiny clip is never sold as a certified week.