Replacing paper with proof: four years of a field operations platform

The real problem was proof

Nothing in that process confirmed a job had actually been done. A completed form said a crew had been to a site. It did not say anyone had been there.

That matters more than it sounds. When a client called to ask why their site had been skipped, the honest answer was that nobody could tell. The paperwork said it was done. There was no way to check, and no way to show the client anything better than the same paperwork.

Every disputed visit cost a phone call, a supervisor's afternoon, and some trust.

What we built

Locspect is a field operations platform: a mobile app for crews and supervisors, and a web application for the office.

The mechanism that solved the proof problem is deliberately boring. A crew member starts a job in the app and ends it when they leave. The app records the coordinates at both ends. No forms to fill in, no photos to take, nothing extra to remember.

That single change turned "the paperwork says it was done" into "the job started here at 08:12 and ended here at 09:47." Disputes stopped being arguments and started being lookups.

The office side replaced the spreadsheets. Sites, work orders and schedules live in one place, and the data arrives as the work happens rather than days later through a stack of paper.

Simple on purpose

The crews using this are not office workers. They are cleaning sites, often in a hurry, often on their own phone, sometimes in a second language. A system they need training for is a system they work around.

So the app does very little. Start the job, end the job. Most people are using it correctly within minutes, and nobody sits through onboarding.

That constraint shaped everything. Every feature proposed since has had to answer the same question: does this add a step for someone standing in a stairwell at 6 a.m.? Most of the good ideas ended up on the office side of the system instead, where someone has a desk and a full screen.

Under three months from the first line of code to crews using it every day. Not a pilot. Real work, on real sites.

What it looks like four years later

About 1,000 locations. Around 500 people using it daily. Still in production, still being developed.

What it grew into, in the order the work demanded it:

• Deadline alerts. The database checks which inspections are approaching their deadline and raises alerts before anyone has to notice manually.
• Push notifications for jobs that have been running unusually long, sent hourly, so a supervisor hears about a problem while the crew is still on site.
• Photos, added later. Supervisors document conditions when something needs a record. Phones produce HEIC files that most systems cannot read, so a separate service converts every upload automatically. Nobody in the field has to think about file formats.
• An operator view for managing multiple companies from one place as the platform grew past a single client.

What keeps it running

Four years of daily use by 500 people is a different problem from launching something. What that actually requires:

• Alerting on the things that hurt, not everything. Errors, slow requests, database load, failed scheduled jobs.
• An audit trail. Every change to a company, a site or a user link is recorded with the old values and who made the change. This was added after a site was deleted directly in the database and nothing recorded who did it. Now something does.
• Infrastructure as code, so the environment is reproducible rather than something assembled by hand and remembered.
• A continuity document. Plain instructions for what to do if I am unavailable: where everything lives, how to get into it, what breaks first. It is tested, not just written.

The data stays in Canada. The whole system runs in Google Cloud's Montréal region.

What to take from this

• The obvious problem is rarely the expensive one. Re-typing paper into spreadsheets was slow. Not being able to prove work was done cost more, and nobody had costed it.
• The fix should be passive. Asking crews to do more admin to prove they did their job fails. Recording coordinates on a button they already press works.
• If it needs training, it needs simplifying. Field software competes with paper, and paper needs no onboarding. Two buttons beat a better-designed form.
• Three months to daily use is possible, when the first version does one thing properly instead of everything adequately. Photos, alerts and notifications all came later, once people were relying on it.
• Ask anyone you hire what happens after. Four years in, the interesting parts of this system are the audit trail, the alerts and the continuity plan. None of them were in the original scope. All of them are why it still works.