The constraint came before the requirements
They coordinate around 400 carers visiting roughly 2,600 clients across two provinces. Rostering is genuinely hard: continuity of carer matters clinically, travel time between visits is dead cost, and availability changes daily.
Their incumbent software did the job adequately and cost €112,800 a year. It also couldn't model travel time properly, which their own analysis suggested was burning something like 340 carer-hours a month.
Their COO had scoped a custom build twice and stopped both times at the same wall: the data protection officer would not approve any arrangement where client data, or code touching client data, moved through non-EU infrastructure. Every studio she'd spoken to either couldn't answer the question or answered it with a shrug about their US cloud provider's Frankfurt region.
How we handled it
Region pinning is a first-class option in our build process, not an accommodation. They selected EU, and that selection propagated:
- Build VM and staging environment on EU-hosted infrastructure, Frankfurt
- Model inference routed only to EU-resident endpoints
- Artifact and object storage EU-only, encrypted at rest with keys held separately
- Full infrastructure attestation produced as a build deliverable, listing every service, its jurisdiction, and its data handling
The DPO reviewed that attestation, came back with four questions, and required a second review after we changed one storage provider mid-build. That second review added eight days to the schedule. We'd rather lose eight days than have that conversation after launch.
Delivered
€61,000 fixed, twelve milestones, nine weeks plus the compliance pause.
The rostering engine optimises for carer continuity first and travel second, which was the priority order their clinical lead insisted on and the old vendor had inverted.
Six months after go-live:
- Licensing cost eliminated: €112,800/year
- Travel time per carer down 22%, roughly 290 hours a month returned to client-facing work
- Continuity of carer (same carer across a 4-week window) up from 61% to 79%
- Roster build time for the scheduling team: 9 hours a week down to 2
- Payback on the build: 7 months
The COO: "Our DPO has never signed off on anything on the first pass. That he signed off at all is the part of this I'd have bet against."