FIRESPACE · LON 51.5074°N · 0.0877°W
● LEO Low Earth orbit
ALT 400 km VEL 7.66 km/s
LINK NOMINAL MMXXVI · PLATE I
← FIRESPACE / LEDGER
Case study · № II

One data model for a business that had been running on seven.

// Technical sheet 2025
CLIENT A Latin American health marketplace
SECTOR ~1.2M patients · Buenos Aires
DURATION 6 months
TEAM 4 engineers
DISCIPLINES Architecture · Analytics · Automation

The finance team was spending the first eleven business days of every month reconciling what their own systems claimed had happened. Different teams had, over the years, shipped their own shorthand for what an "appointment" meant — and the product had quietly grown seven versions.

The shape of the problem

The CRM counted booked intents. The clinic software counted consumed slots. The invoicing tool counted settled fees. None of these numbers were wrong. They were just answering different questions — and the business was treating them as if they were the same answer.

"We had stopped trusting our own dashboards. Meetings began with ten minutes of asking which definition we were using."

— CFO, Client

The work

A semantic layer, carefully: one source of truth, three views built on top, and a migration plan measured in months rather than weekends. No system was switched off before its replacement had run in shadow mode for at least four weeks.

Placeholder — architecture diagram
PLATE · Semantic layer · before / after

Where it landed

METRIC
BEFORE
AFTER
Data sources 7 of record 1 of record · 3 views
Monthly close 11 days < 1 day
Dashboards in use 23 (≈ 6 trusted) 11 (all trusted)
Reports written weekly 0 4