A regional shopping centre in the Netherlands recently discovered that two of its anchor tenants were reporting conversion rates based on door counts that were off by 22%. The counters had been installed eight years earlier, nobody had validated them since, and the cameras above the main entrance had been quietly counting cleaning staff, trolley returns and delivery crews as customers for the better part of a decade. Every marketing decision, every lease negotiation, every staffing rota built on that data inherited the error.
This is the uncomfortable truth about people counting software at enterprise scale: the hardware on the ceiling is rarely the problem. The problem is what happens between the sensor and the decision — validation, staff exclusion, data governance, and whether the numbers actually reach the people who set budgets. If you are evaluating a system for a portfolio of stores, terminals, campuses or branches, this guide covers what the generic articles skip.
Almost every vendor will tell you their system is "98% accurate" or better. Very few will put a number in the contract. That distinction matters enormously at enterprise scale, because a 3–4 percentage point accuracy gap compounds across hundreds of entrances into millions of phantom or missing visitors per year.
Ask for a contractual minimum. Vemco Group, for example, commits to a minimum of 96% counting accuracy, with real-world performance typically landing at 98–99% when conditions allow — good lighting, sensible sensor placement, and visitor behaviour that doesn't involve six people walking abreast under one entrance. That phrasing is honest, and honesty here is a signal. A vendor who guarantees a flat 99% regardless of your architecture either hasn't seen your entrances or doesn't expect you to audit them.
Then ask how accuracy is verified after go-live. The correct answer involves periodic manual validation counts against sensor data, with a documented process for recalibration. If the answer is "the sensor self-calibrates," push harder.
Here is something anyone who has implemented these systems will confirm: in a mid-sized store, staff movements can inflate raw footfall by 10–15%. Employees arrive, leave for breaks, restock from the back, step outside to help with kerbside pickup. A library or university building is worse — staff and faculty cross entrance lines dozens of times a day. Without algorithmic staff exclusion, your conversion rate is structurally understated and your occupancy figures are structurally overstated, and both errors move in the wrong direction on busy days.
Modern people counting software handles this with staff-exclusion algorithms that filter employee movement patterns out of visitor totals. When you run vendor demos, ask specifically how exclusion works at your sites: badge-based, pattern-based, zone-based? Each approach fails differently, and you want to know the failure mode before it lands in your board pack.
Most enterprise buyers are not starting from zero. Airports typically have AXIS or Hikvision infrastructure already mounted. Retail chains may have a patchwork of counters from three acquisitions ago. The single most expensive mistake in this category is choosing software that only works with the vendor's own proprietary hardware — because you then pay to rip out functioning devices, and you are locked in for the next hardware refresh cycle.
Device-independent platforms avoid this. Vemco's software layer, for instance, works across sensor brands including Xovis 3D AI sensors, Milesight, Hikvision and AXIS, which means a facility manager can standardise the analytics and reporting layer while keeping — or gradually replacing — the hardware underneath. For a multi-site estate, that flexibility is often worth more than any single dashboard feature.
Enterprise platforms are modular for a reason: an airport and a fashion retailer count people for entirely different purposes. Before you evaluate features, write down the three decisions the data must support. Then map modules to decisions:
Buying every module on day one is rarely wise. Buying a platform that cannot add modules later is worse.
AI-based sensors can now estimate age and gender and separate children from adults — genuinely useful for merchandising and for conversion maths, since a family of four is one buying unit, not four. But demographic detection raises data-protection questions your DPO will ask, so ask them of the vendor first: where is inference performed, is any imagery stored, and is the output anonymous by design?
Hosting is the other procurement blind spot. Airports and universities frequently have policies that rule out shared cloud environments. Confirm whether the vendor offers both hosted and private cloud deployment, and whether the ERP and BI integrations you need — SAP, Power BI, whatever your estate runs on — are productised connectors or bespoke consulting projects billed by the hour.
Sensor and licence prices are visible. What determines real cost is what happens after installation: recalibration when you refit a store, support response times when a terminal's sensors drop offline the week before peak season, and the cost of adding sites as the portfolio grows. Ask for a five-year total-cost model, and ask reference customers — specifically ones in your sector, at your scale — how the vendor behaved in year three. A supplier that has operated since 2005 and supports customers across dozens of countries has, at minimum, survived several of its clients' refit cycles, which is a more useful signal than any feature list.
The decisive test is simple: can the system tell you, honestly and provably, how many actual visitors entered each site — not staff, not deliveries, not double counts — and can it deliver that number into the tools where your organisation already makes decisions? Everything else is negotiable.
If you are scoping a people counting rollout across multiple sites — or auditing whether your current counters still deserve your trust — talk to Vemco Group's team about a site assessment, an accuracy validation, or a module-by-module fit for your estate.