We build Raslak Hub, software for construction and service companies, and our own construction company in Helsinki is where we test it. One thing we wanted for our own crew was simple: stop making people remember to clock in. This post is about how we built clock-in on arrival in our time tracking app, Rax Time, and about the decisions that mattered more than the code.

Why manual timesheets drift

When hours are written down at the end of the week, they are estimates. People who are unsure round down rather than risk overstating, so the error tends to go one way: hours the company never bills. A better form does not fix that. Not needing the form does.

Geofences around job sites

Each project in Hub can have a site location and a radius. The phone registers geofences for nearby sites, clocks the worker in when they arrive and clocks them out when they leave. iOS limits how many regions one app can monitor, so the app watches the 20 closest sites.

The site location comes from three sources, in order of trust:

  1. A location the supervisor has set on the project.
  2. The project's street address.
  3. The average position of earlier clock-ins on that project.

The third source matters more than we expected. Plenty of jobs have no clean address, and after a few days of ordinary clock-ins the system knows where the site really is.

Loose limits on purpose

Our first instinct was a tight radius. That was the wrong call. GPS inside buildings, in basements and between blocks of flats is unreliable, and a tight geofence produces false clock-outs. False clock-outs are the fastest way to get people to switch a feature off.

So the defaults are generous:

  • The default radius is 600 metres, and the smallest allowed radius is 150 metres.
  • A visit shorter than three minutes is ignored, so driving past a site does nothing.
  • If a worker leaves and comes back to the same site within 90 minutes, the same timecard continues and the gap is recorded as a break.
  • A gap shorter than 15 minutes is not recorded as a break at all. Fetching a drill from the van is not lunch.

Lunch was the case that settled it. If automatic clock-out ends the working day when someone walks to a café, the worker has to repair their hours by hand, and the automation has created work instead of removing it.

Every automatic entry can be edited

An automatic clock-in is a well-founded guess, not a fact. Entries created this way are marked as automatic, the worker can correct them afterwards, and a supervisor approves hours in one view before they reach payroll or an invoice. A correct timesheet with one manual fix beats an automated one that nobody trusts.

Location only when the worker turns it on

Location is used only if the worker switches automatic clock-in on. Without it, clocking in is still two taps on the phone. We treat location as a convenience the worker opts into, not as a way to watch people.

Clock in first, choose the project later

Not every hour starts on a known site. Workers can clock in without choosing a project and assign the hours afterwards, in bulk if needed. Capturing the hours is what counts. Sorting them can wait until the end of the day.

What we would tell anyone building this

  • Design for the worst GPS you will meet, not the average.
  • Make every automatic action reversible by the person it affects.
  • Judge the feature by how few manual corrections people make, not by how many events it fires.

Rax Time is on the App Store for iPhone, and automatic clock-in is part of Raslak Hub, our field service app for construction and service companies. The time tracking page shows how hours move on to approval and invoicing.