Mistake Count Tracker

The problem

Operational mistakes were remembered, not recorded. Which means they were remembered selectively: the dramatic one from last Tuesday stayed vivid, and the small one that quietly happened eleven times did not. Without a record, the loudest incident sets the agenda instead of the most frequent one.

What I built

A deliberately lightweight log. Recording a mistake takes seconds, because anything that takes longer than seconds doesn’t get recorded during a busy service, and a tracker with gaps in it is worse than useless, since it makes the gaps look like zeroes.

How it works

Each entry is categorised by the kind of mistake rather than attached to a person. Counts accumulate by category, and the useful output is the trend across weeks rather than any individual row.

Technical approach

The design constraint doing the real work here is speed of entry. Every field added to the form is a reason not to fill it in mid-service, so the record stays minimal on purpose: enough to count, not enough to become paperwork.

It counts categories, not people. The moment a log like this becomes a record of who is at fault, it stops being filled in honestly, and an incomplete log produces confidently wrong conclusions. Tracking the type of mistake keeps the incentive pointed at fixing the process.

Current state

In use. Patterns that used to take months to notice (a step that’s systematically error-prone, a time of day where errors cluster) became visible in the trend. That’s the point: the tracker doesn’t fix anything itself, it just makes the thing worth fixing obvious enough to argue about with data.