Everything About People Counting Solutions & Features

sensor health monitoring people counting — Knowing a Sensor Is Down Before the Data Is Wrong | Vemco Group

Written by Admin | Sep 4, 2026, 6:31:55 PM

The most expensive footfall failure is not the sensor that dies. It is the sensor that keeps reporting. A dead device produces an obvious gap — zeros in the dashboard, a ticket raised within the hour. A drifting device produces numbers that look entirely plausible: down 8% versus last week, which the trading team explains away as weather, promotions, or the school holidays. Six weeks later someone runs a manual count and discovers the store's conversion rate has been fiction since a shelving unit was moved under the counting zone.

If you own footfall data quality across an estate, this is the scenario you are actually paid to prevent. And the honest position is that most estates cannot prevent it today, because their monitoring answers the wrong question. "Is the device online?" is not the same question as "Is the data right?"

Why ageing estates fail quietly

Sensor estates degrade in ways that rarely announce themselves. A network switch gets replaced during a refit and the device comes back on a different VLAN, still pingable but no longer pushing counts. A ceiling tile is disturbed and the sensor's angle shifts two degrees. A new till bank creates a dwell zone directly under a count line, inflating entries. Lighting is changed to warmer LEDs and a marginal installation that was performing well now double-counts at dusk.

None of these produce a zero. All of them produce data that passes a casual glance. The reporting looks normal until someone checks — and in most organisations, nobody checks until a stakeholder disputes a number, at which point you are defending your entire dataset rather than fixing one device.

Here is a practitioner detail that separates people who have actually run estates from people who write about them: the single most common cause of silent drift is not hardware failure at all. It is store operations. Seasonal displays, queue barriers, promotional bins and Christmas trees placed inside the counting zone will corrupt data faster than any component failure, and store staff will never report it because they have no idea the ceiling device exists. Your monitoring has to catch what your stores will never tell you.

The Irisys end-of-life problem nobody budgeted for

There is a specific, dated reason this matters more now than it did two years ago. Irisys hardware reached end of life in March 2023, with manufacturer support ending in March 2024. If your estate runs Vector 4D or Gazelle devices — and a large share of European retail estates do — the manufacturer safety net is gone. Firmware issues, health telemetry, diagnostic support: none of that comes from the factory any more.

That does not mean the hardware is finished. These devices are well-built and many estates will run them for years yet. But it does mean ongoing health monitoring now depends entirely on the platform layer sitting above the devices, not on the manufacturer beneath them. If your platform only aggregates counts and does not actively manage device health, you are running end-of-life hardware with no one watching it. That is a risk position, and it belongs in your next budget conversation framed exactly that way.

This is where partner depth matters. Vemco Group, founded in Denmark in 2005, has been an Irisys partner for more than 15 years — long enough to know these devices' failure modes at firmware level, not just at dashboard level. Their Estate Manager is fully integrated for remote access and control of Vector 4D and Gazelle devices, which in practice means reconfiguring a count line or pulling diagnostics from a store in Aberdeen without booking an engineer visit.

What real health monitoring looks like

Strip away the vendor language and health monitoring for a people counting estate needs to answer four questions continuously, not weekly:

  • Is the device reachable? Basic connectivity, but flagged in real time — not discovered when the overnight batch fails.
  • Is data arriving? A reachable device that stopped transmitting counts is the classic silent failure. These are different checks and must be monitored separately.
  • Is the data plausible? Statistical checks against the location's own history: sudden ratio shifts between in and out counts, entry patterns that no longer match trading hours, a Tuesday that suddenly outperforms every Saturday on record.
  • Is the device still configured correctly? Count lines, zones and firmware versions audited centrally, because local changes are where drift begins.

The Vemco platform covers this with full health management, flagging in real time both when a sensor is offline and — the harder problem — when data is inaccurate. It sits within an enterprise solution that runs from device management through to advanced analytics for footfall, queue measurement and re-ID tracking, which matters for a practical reason: when the same platform manages the device and consumes its data, anomaly detection can act on device telemetry, not just on the numbers after they have gone wrong.

Accuracy is a maintained state, not a purchase

Be sceptical of any vendor quoting a single flat accuracy figure with no conditions attached. The honest framing is the one Vemco uses contractually: a minimum of 96% counting accuracy, typically achieving 98–99% where lighting, store layout and visitor behaviour allow. That gap between minimum and typical is not hedging — it is an admission that accuracy is an operational property. A device validated at 99% on installation day does not stay there by itself. It stays there because someone notices when conditions change.

For the analytics owner, this reframes the budget case. You are not buying monitoring software; you are buying the difference between an estate whose accuracy erodes invisibly and one whose accuracy is defended daily. Every downstream metric — conversion, staff-to-traffic ratios, marketing attribution, lease negotiations backed by traffic data — inherits the quality of the count. A 5% silent error in footfall does not stay 5%; it compounds through every ratio built on top of it.

Three actions worth taking this quarter

  • Audit your alerting logic. Confirm whether you would be alerted to a device that is online but sending implausible data. Most estates find they would not.
  • Map your Irisys exposure. Count how many end-of-life devices you run and document who is responsible for their health now that the manufacturer is not.
  • Measure time-to-detection. Pull your last three sensor incidents and calculate how long each ran before anyone noticed. That number, multiplied across your estate, is your data quality risk in concrete terms.

The estates that get this right are not the ones with the newest hardware. They are the ones where a drifting sensor becomes a ticket within hours instead of a dispute within months.

If you are running Vector 4D or Gazelle devices without manufacturer support, or you cannot currently say with confidence which of your sensors are counting accurately today, talk to Vemco Group about a health assessment of your sensor estate — a direct conversation about your device inventory, your detection gaps, and what real-time health monitoring would look like across your locations.