Skip to content
Everyone on Site

All notes  /  In practice

How This Goes Wrong

The recurring failures on construction sites, each findable in a morning, each with a specific correction.

In practice · Analysis

Site workforce systems fail in recognisable ways, and almost none of them is the hardware.

For teams turning how this goes wrong into a repeatable record, read the guide provides a useful operational reference for time, attendance and reporting workflows. It should be assessed alongside Chartered Institute of Building so the software choice remains grounded in site practice and applicable guidance.

Exits not recorded

Symptom: the register says two hundred at four in the afternoon and the site has forty.

Correction: equal lanes both ways, or a sign-out tied to something people want. And measure the unmatched-entry rate.

The register cannot be produced at the assembly point

Symptom: a drill where somebody runs back to the office for a printout.

Correction: an offline view, tested, plus an automatically refreshed printed backup.

Competence checked once

Symptom: cards expiring on site with nobody noticing.

Correction: expiry held against the person and enforced at every access, with a thirty-day forward list.

Temporary access becoming permanent

Symptom: a rising count of people admitted under exception.

Correction: a written rule with an expiry, and a weekly report of who is on it.

The chain not recorded

Symptom: somebody on site whose engaging company nobody can name.

Correction: engaging company as a required field, and a flow-down clause that is actually passed on.

Induction that does not expire

Symptom: people briefed on a site that no longer exists.

Correction: re-induction at phase changes, and a refresher after a long absence.

Plant authorisation merged with access

Symptom: either a gate that cannot cope or machines operated by whoever is available.

Correction: two registers, deliberately separate, consolidated weekly.

Records lost at project end

Symptom: a question in year three that nobody can answer.

Correction: an archive decided at project start with a named owner and a handover task.

Nobody owns the register

Symptom: all of the above, arriving together.

Correction: a named person with time, which is the intervention that prevents the others.

The common thread

Every one is a process decision rather than a technology problem.

Which is why replacing the system fixes none of them, and why the correction in each case costs a decision rather than a purchase.

And most are found by one morning of looking: compare the register against a head count, ask for the evacuation list, pull the expiry report, and count the temporary accesses.

The order to fix them in

The evacuation view first, because it is the only failure with consequences that are not commercial.

Then the exits, which is what makes the evacuation view wrong.

Then competence expiry, which is the enforcement that actually refuses somebody.

Then the chain recording, then the archive, then the ownership — although naming the owner first would produce all the others in sequence, which is the argument for doing it first regardless.

Finding them in one morning

Stand at the gate at finishing time and watch whether people sign out.

Ask for the evacuation list, timed.

Pull the expiry report and count what is already past.

Count the temporary accesses granted this month.

Pick three people at random and ask which company engaged them.

Five checks, one morning, and they will locate every failure in this note — which is a better use of a morning than any procurement exercise.

What none of them requires

A new supplier.

Additional hardware.

Or a larger budget, in most cases.

Which is worth saying because the response to a visible failure in this area is usually a proposal to replace the system, and replacing it carries every process decision across unchanged.

Two more worth naming

Multiple gates with partial coverage. A site with three entrances and readers on two has a register that is wrong by however many people use the third, and nobody knows the number because the third gate produces no data at all.

Correction: cover every entrance or close the uncovered ones, and say which was chosen.

The vehicle gate. People arrive in vans and only the driver presents a card.

Correction: a reader for passengers, or an accepted and stated undercount — but not an assumed capability.