Fluidity Group 2026

The situation

A construction site spanning two city blocks needed to know exactly who was on it at any moment — not just how many, but whether each person was authorised to be there and whether their certifications were still valid. Every assumption a conventional access-control system makes was unavailable.

A sign-in sheet that is wrong by ten in the morning

Large sites carry a duty to know who is on them — for evacuation, for security, and increasingly for compliance. But headcount alone is not the question. The questions are who, are they authorised to be here, and are their certifications still current.

A clipboard in the site cabin answers none of those reliably. It is a record of intent, taken once, and it decays from the moment it is written.

Everything a normal design assumes was missing

Most of the site was outdoors — no ceilings to mount to, no building to shelter equipment.
The entry and exit points move. The gate that is right in month two is under a slab by month six. Anything bolted down permanently is wrong within a quarter.
No wired communications. No structured cabling to reach, and nothing worth trenching for on a site that will be reconfigured repeatedly.
Almost no wired power. A handful of entry gates had a supply for the turnstile itself. Everything else — every proximity station out across the site — had none. That constraint decided the technology.

Why this technology

With no power and no network, the energy budget selected the technology — not the positioning specification.

UWB would have given precision the use case did not need, at an infrastructure and power density a site with no mains supply could not sustain.
GNSS looks like the obvious outdoor answer and is the wrong one here. The requirement was proximity and authority at a gate, not coordinates in a field — and GNSS cannot answer "did this person pass this turnstile with a valid certification". Higher tag cost and a charging regime, for a question it does not address.
RFID portals would have worked at a fixed gate. The gates were not fixed.
Wi-Fi was not an option. There was no network to use.
BLE fit exactly: low-power receivers, inexpensive tags for a transient workforce, proximity semantics that map directly onto gates and turnstiles, and an energy budget small enough to run on solar.

What we built

A system designed to be moved

Minew BLE ID tags carried in a pocket, or helmet tags fixed to the hard hat — whichever suited the trade. Off-the-shelf devices from a partner we know at volume, because 1,200 workers is a logistics exercise, not a hardware experiment.
Hands-free proximity points at entry gates. A worker walks through and the site knows, with nothing to present and nothing to remember.
Close-proximity "tap points" on turnstiles where access authority has to be positively managed — entitlement and certification checked at the moment of entry, rather than reconciled after the fact.
Fifteen self-contained proximity stations, engineered by Fluidity. No product on the market did this job, so we built one: off-the-shelf electronics, our own firmware, and a housing designed for pole mounting, forktruck relocation and two years of weather. Five covered the controlled gates; ten went out across the site, each with its own LTE backhaul, solar panel and battery.
An off-the-shelf platform, customised where it fell short. We do not build software where good software exists — we built the certification integrations and the site-specific logic on top of a proven platform, and spent the engineering budget where nothing off the shelf would do.
Solar as the power source, not a top-up. The stations ran continuously on solar — the battery was a reserve, sized to carry a station through thirty days without usable sun, which in practice was rarely called on. The design goal was simple: no station ever needs a visit because of power.
Portable in practice, not just on the datasheet. Over the life of the build the stations were relocated around twenty times — lifted by forktruck, moved in minutes, and re-established in the platform in seconds.
Self-locating infrastructure. Each station carries its own GPS, so even before anyone updates the platform, the system knows roughly where its own equipment is standing. A location system that has to be told where it is has missed its own point.
Reused beyond the project. The first site ran for two years; when the build completed, the same equipment moved on to the next sites, and the engagement continues today. The client bought infrastructure, not a fit-out.

The turnstile knows whether your ticket is still valid

Safety, health and competency certifications for construction workers live in third-party systems, and they expire. Traditionally nobody finds out until an audit — or until something goes wrong and the paperwork is examined afterwards.

Fluidity built the integrations so those certifications are checked at the turnstile, in the second the worker taps. If a certification has lapsed:

Hard refusal. The turnstile does not open. Not a warning, not a flag for later — the person does not get on site.
Their contractor's manager is notified immediately — who was refused, which certification lapsed, and exactly when they presented.
The problem lands with the person who can fix it. Not the gate operative, and not a compliance team three weeks later.

That last part matters more than it sounds. Access control that only says no creates an argument at the gate. Access control that says no and tells the right person why creates a correction — and an auditable record that it happened.

A roll call against who is actually here

Because ingress and egress were positively controlled, the site always held a live answer to the question every evacuation starts with: who is on site right now? Muster then ran against that list — not against a sign-in sheet from seven in the morning.

A proximity station at the muster point. Workers assemble, their tags are heard, and the count executes automatically against the live on-site list. Whoever is missing is named, not estimated.
A muster point in a pocket. If a station was offline — or the assembly point had to move — a smartphone app scanned for every tag around it and located them against the phone's own GPS. Any supervisor's phone could become the muster station.

The second layer is the one worth noticing. Emergencies are precisely when infrastructure is most likely to be compromised — so the fallback cannot depend on the infrastructure. A phone, a scan and a GPS fix is a muster system that survives the fire that triggered it.

The infrastructure argument, proven

Access control justified the project on its own. But once self-powered proximity stations were distributed across the site, the same infrastructure delivered something the client had not paid for separately.

Zonal awareness of where people were — and, more valuable day to day, tags on 5,000+ items of plant, tooling and construction material. For every asset, the site could answer three questions it never could before:

Did it actually arrive? Delivery confirmed by presence, not by a signature on a docket for a crate nobody opened.
Where was it put? In the yard, on the floor, up the structure — found in seconds instead of searched for by radio.
Did it move out of hours? An asset that travels at two in the morning is a question someone should be asked. Now it gets asked.

The marginal cost of all three was a tag.

This is exactly the case we make on every project: build the foundation properly and the next four use cases cost almost nothing.

The hard part

Hands-free entry is exactly what a workforce wants at seven in the morning with hundreds of people arriving inside twenty minutes. It is also an open invitation to walk through behind someone else — and a headcount that can be beaten by following a colleague closely is not a headcount at all.

On a site where the entire point is knowing exactly who is present, this had to be solved properly rather than accepted. It took three layers:

A positive tap. At controlled turnstiles the worker taps their BLE card — a deliberate act tied to one identity, not a passive detection.
A receiver on the far side. If more than one tag crossed on a single authorisation, the system knew — because it was reading the far side of the turnstile as well as the front.
Mechanical enforcement. The turnstile locks after a single passage. One authorisation, one person, every time.

Two of those are physical and one is RF, and that combination is the point. On a real site the answer to a human behaviour problem is rarely software alone.

What transfers

Design for the site you will have in month twelve, not the one you are standing on today. Ours moved twenty times and never needed a redesign — because moving was the design.
Power is the constraint, not positioning. On a construction site the energy budget picks the technology, and everything else follows from it.
Access control and asset tracking are the same infrastructure. Justify it once, use it four times.
Route the exception to whoever can resolve it. A refusal at the gate is only useful if the person responsible for that worker hears about it in the same minute.
Assume people will tailgate. Hands-free entry and an accurate headcount are in tension. Resolve it mechanically as well as by RF, or accept that the number is approximate.
A transient workforce is a logistics problem. Issuing, recovering and reassigning tags has to take under two minutes at the gate, or it will not happen at all.