Case study · 2024 · Nordklinikk
Scheduling for four thousand clinicians who all have a good reason
Replacing eleven regional rota systems with one, without pretending that the eleven local practices were wrong.
Overview
Eleven health regions each ran their own rota system. Nine had been bought, two had been written in-house, and every one of them encoded a different definition of a fair week.
I led design on the consolidation: one scheduling product for four thousand clinicians, plus the migration path that moved each region onto it without a hard cutover.
Problem
A rota is not a calendar. It is a negotiation between safe staffing, contracted hours, individual preference and whatever happened last month — and it is renegotiated every time one person calls in sick.
The brief called this a data migration. Eleven definitions of a fair week made it a policy question that nobody had the authority to settle.
Research
Six weeks across four regions, forty-one rota builds observed end to end, and shadowing the three schedulers everyone else called when their own system said no.
The finding that reframed the project: every region ran a shadow spreadsheet beside the official system. The spreadsheet held the exceptions, and the exceptions were the actual rota.
Process and explorations
We explored three positions: one national rule set, per-region rule sets over a shared core, and a rules engine each region configured for itself.
The configurable engine demoed well and failed the first workshop — it asked each region to write down rules they had never written down. We shipped the shared core with regional overrides, and made every override visible to every region.
Final design
Schedulers see conflicts before they see a grid. The system proposes a rota, states which constraint it broke and why, and leaves the decision with the person who will have to defend it.
Clinicians get one screen: what they are working, what they asked for, and what it would cost a colleague to change it.
Prototype
We rebuilt one region's live November rota as a clickable prototype and asked its three schedulers to break it. They did, in eleven minutes, on a rule none of us had written down.
Impact
Measured across the first four regions, six months after cutover.
- 11 → 1
- Regional systems consolidated
- −63%
- Time to build a monthly rota
- 4,000
- Clinicians on one system
Reflection
The shadow spreadsheets were not a workaround to be removed. They were eleven regions documenting their real policy for free, and we spent the first quarter trying to delete them.
The thing I keep: a rota tool that cannot explain its own refusal is a spreadsheet with worse permissions.