Everything About People Counting Solutions & Features

retail footfall analytics — Procurement Guide for Retail Footfall Analytics | Vemco Group

Written by Admin | Sep 22, 2026, 5:06:51 PM

A shopping centre operator we spoke with had run a tender for 140 entrances. The winning bid was 30% cheaper per sensor than the runner-up. Eighteen months later the sensors were fine, but the software licence was tied to that hardware, the data export was a CSV emailed weekly, and the marketing team had quietly gone back to guessing. The saving on hardware was gone within the second year, and replacing the platform meant replacing the sensors too.

That is the most common failure in retail footfall analytics procurement: the tender is written around devices, and devices are the least important part of the decision. This guide covers what to specify instead, how to phrase the clauses that actually protect you, and how to score bids so the cheapest sensor does not win by default.

Start with the questions the data must answer

Before anyone opens an rfp template, write down the five decisions the counts will inform. They differ sharply by sector, and they determine which requirements matter.

  • A fashion retailer wants conversion by store by hour, which means counts must join to POS transactions at the same granularity.
  • A shopping centre wants tenant-level traffic to defend rent reviews and turnover leases, which means counts per unit entrance, not just the mall doors.
  • An airport or a university library wants live occupancy for capacity and cleaning schedules, which pushes the requirement toward occupancy management software with real-time thresholds rather than daily reports.
  • A facility manager wants zone-level dwell to justify a floor-plate reconfiguration, which needs directional counting, not a simple in/out gate.

If a requirement in your tender cannot be traced back to one of those decisions, cut it. Bloated specifications favour bidders who are good at answering questionnaires, not the ones who are good at counting people.

Write the accuracy clause so it can be enforced

Almost every bid will claim 98% or 99% accuracy. Almost none will commit to it in the contract. The difference between a marketing figure and a contractual figure is the single most valuable line in your tender.

Ask for a contractual minimum accuracy, defined against a manual count method you specify, measured at entrances you choose, over a period you set. Vemco Group, for example, commits to a contractual minimum of 96%, with typical performance of 98 to 99% where lighting, layout and visitor behaviour allow. That distinction matters: the 96% is what you can hold a vendor to, the 98 to 99% is what you should expect on a well-installed site. A vendor who will only put the typical figure in a brochure and not the minimum in a contract is telling you something.

Specify the audit too. State that you may perform manual validation counts at any entrance, at a time of your choosing, with a defined remedy if the site falls below the threshold. Without the remedy, the clause is decoration.

Separate the software from the sensor

This is where the shopping centre in the opening paragraph went wrong. If the people counting software only works with one manufacturer's devices, every future hardware decision is made for you, and the vendor knows it when renewal pricing comes around.

Require that the platform be device-independent and sensor-agnostic, and ask bidders to list which sensor types they currently ingest: stereo video, thermal, time-of-flight, Wi-Fi probe, existing CCTV analytics. A platform that can take counts from several hardware families lets you fit thermal sensors at a glass-roofed atrium where video struggles, cheaper units at a back-of-house staff door, and keep whatever the previous supplier installed at sites you acquire. Vemco Group's platform is built that way precisely because estates rarely have uniform hardware, and a merger or a refurbishment should not force a rip-and-replace.

Also ask how the platform is hosted. Hosted cloud is simplest; a private cloud option matters for airports, universities and public libraries with data-residency rules. Get both options priced.

A practitioner's note on where accuracy is actually lost

Installers will tell you that most counting errors have nothing to do with the sensor. They come from a ceiling height that was measured off the drawings rather than on site, a promotional gondola pushed under the counting line three weeks after commissioning, or a staff entrance behind the tills that nobody flagged as an entrance. Wide sliding doors in a mall lobby where groups walk shoulder to shoulder are a genuine challenge; a single 90cm shop door is not.

So in your tender, ask each bidder to identify your three hardest entrances from the drawings you supply, explain how they would handle them, and confirm whether the accuracy commitment applies there too or only on average across the estate. An estate-wide average lets a vendor hide two badly performing flagship entrances behind forty easy ones.

Integration is a requirement, not a roadmap item

Counts on their own tell you how busy you were. Counts joined to transactions tell you conversion; joined to rota data they tell you staffing efficiency; joined to lease data they tell you whether a tenant's turnover claim is plausible. Ask specifically how the retail analytics platform connects to your POS, BI, ERP and CRM systems, and whether that is a documented API and connector layer or a professional-services project billed by the day.

Vemco Group handles this through VemFusion, which is the layer that pushes footfall into the systems you already report from. Whatever the vendor calls theirs, insist on seeing it working against a system you actually use, not a slide.

Scoring the bids: a weighting that resists the cheapest sensor

A vendor evaluation that gives price 40% of the score will pick the hardware discounter every time. A weighting that has held up well across estates of different sizes looks roughly like this:

  • Contractual accuracy and audit terms: 25%
  • Sensor independence and hosting options: 20%
  • Integration depth demonstrated live: 20%
  • Five-year total cost including licences, support, sensor replacement and data export at exit: 25%
  • Scalability from your pilot sites to the full estate without re-platforming: 10%

Note that price is scored on five-year cost, not unit price, and that "data export at exit" is inside it. A people counting system that charges you to take your own historical counts when you leave is a lock-in fee by another name.

What to request rather than assume

Certifications, service-level figures, support hours and reference sites all vary by market and by contract. Do not accept them from a website, including ours. Put them in the tender as documents to be supplied: security and privacy certifications relevant to your jurisdiction, the SLA with measurable response times, the support model for a sensor failure at a site 400km from head office, and three reference customers in your sector you may contact directly. A vendor with 55,000+ installations across 98+ countries, as Vemco Group has, should be able to produce a reference that looks like your estate; if any bidder cannot, that is a finding.

Finally, run a paid pilot at two or three sites before the full award, with the accuracy audit method you intend to use in the contract. Twelve weeks of real data at your hardest entrance is worth more than any evaluation matrix.

If you are drafting a footfall tender and want a second opinion on the accuracy clause, the sensor-independence wording or the scoring weights before it goes out, contact Vemco Group and we will review the requirements against what we have seen work in comparable estates.