Why field rollouts
stall in week three.

Ask a contractor why the last system did not stick and you will usually be told the lads would not use it. That is a description, not an explanation. Field software is almost never rejected in week one, when it is new and somebody senior is watching. It dies in week three, quietly, when the novelty has gone and the old way is still available.

The shape of the failure

The pattern is consistent enough to set your watch by.

Week one goes well. There has been a briefing. Somebody from the office is on site. The supervisor is making a point of using it. Compliance looks close to total, and everybody involved reports back that it has landed.

Week two holds, mostly. The visible support has gone but the memory of it has not. A few people miss a day. Somebody in the office fixes it by hand rather than making a fuss during a fragile period, which is a kind and entirely reasonable decision that will turn out to be the problem.

Week three is where it goes. There is weather, or a delivery, or an inspection, and the day gets away from everybody. Half the gang does not clock in on the app. The office fills the gap from the old paper sheets, because the payroll deadline does not move for anybody. And at that moment the system stops being how hours are recorded and becomes one of two ways hours can be recorded, one of which requires effort.

Nothing is ever decided. The old route simply stayed open, and water goes downhill.

Why it is week three and not week six

Three ordinary forces converge at about the fifteenth working day.

• Attention has moved on. Whoever sponsored the rollout has a business to run. The daily prompt that carried weeks one and two is gone, and nothing has replaced it.
• The first exception has been absorbed. Someone's phone died, or a new starter had no induction, and the office worked around it. That workaround is now precedent, and precedent spreads faster than policy.
• The first payday has landed. If hours recorded the new way and hours recorded the old way both produced correct pay, the system has just told everybody, truthfully, that using it is optional.

That last one is the strongest signal in a rollout, and it is almost always sent by accident.

The five specific failure modes

1. The paper route stayed open. This is the big one and it swallows the others. If the old form still exists, the old form wins on the bad day, and every day on a construction site is somebody's bad day.
2. It is a stranger’s app. An app store install is a real barrier on its own: an account, storage, a personal device used for work. It gets worse when the icon belongs to a vendor nobody in the yard has heard of, because then it is a stranger’s app on a personal phone, and a password on top of that. Every one of those is a legitimate reason to not do it today, and every operative who does not do it today becomes an exception the office covers.
3. It requires a password. Passwords are fine in an office. On a gate at seven in the morning in the rain, a forgotten password is a person who does not clock in, and after the second time they stop trying.
4. The supervisor is not visibly using it. Site behaviour follows the supervisor, not the memo. If they are still taking their own notes on paper, the gang reads that correctly as the real system.
5. It asks for more than it gives back. A system that only extracts data from site returns nothing to site. If the operative never sees a benefit, and the supervisor never sees a benefit, then every interaction is a tax.

What the firms that got it to stick did

Not one of these is technical.

• They named a date and closed the paper route on it. Not "we are moving to the app". "From the 1st, the paper sheet is not accepted, and if the app fails, ring the office and we will record it with a manual flag." That sentence is the whole rollout. It gives an escape hatch that is visibly worse than compliance and does not quietly replace it.
• They put the scan point where people already stop. The cabin door, the gate, the welfare unit. Not where the work is, where the pause already exists. How many scan points to run, and what each scan proves, is set out in QR codes on construction sites.
• They made the supervisor the first user, not the last. A fortnight before the gang, so that on day one the person the gang watches is fluent rather than learning alongside them.
• They let the first payday do the teaching. Hours came from the system, and where they did not, the correction was visible and slightly inconvenient rather than silent. Nobody was punished. Everybody understood.
• They gave something back in week one. Usually the simplest thing: the operative can see their own hours, and the supervisor stops spending Friday deciphering handwriting. A system that removes a job somebody hated gets defended by that person.

Adoption is not a training problem. It is a question of which route is open on the worst day of the month.

The parts that are not adoption problems

It is worth separating the failures that discipline can fix from the ones it cannot, because treating a real constraint as a motivation problem is how a rollout loses the site team's goodwill for good.

Signal is the obvious one. Where coverage is genuinely poor, no amount of expectation-setting changes the physics, and treating it as a motivation problem makes the office look like it has never been on a site. The answers are practical rather than clever: put the scan point where the gang passes anyway, or have the office enter those hours on the operative's behalf with a manual flag and the same audit trail behind them.

Language is another. If a meaningful part of the gang does not read English comfortably, an interface is not the fix and a briefing in the right language is. And shared phones, common on some sites, need deciding deliberately rather than discovered when two people's hours land on one device.

How to tell in week two whether it is working

Do not measure logins. Measure the exception rate: how many records the office had to create or correct by hand, as a share of the total. It is the only number that predicts week three, because every hand correction is the old route being used with extra steps.

If that share is flat or falling, the rollout is fine and you can stop watching. If it is rising, something specific is wrong and it will have a name, usually one of the five above. Find out which before the next payday, because after two clean paydays the habit sets, and after two mixed ones so does the other one.