When Covid-19 hit Brazil, hospitals had no centralized way to manage bed availability and resource allocation. Most were tracking beds, equipment, and patient flow through spreadsheets — updated manually, often hours behind reality. The "Todos Pela Saúde" initiative, backed by Itaú, needed a backoffice portal that could replace that patchwork with a single source of truth.
The constraint was time. The pandemic wasn't going to wait for a polished product. We had to ship something functional fast, then iterate.
Understanding the problem
We started with interviews — doctors and nurses on the front line. What we heard wasn't what we expected. The biggest pain point wasn't lack of data; it was data scattered across too many places. A nurse checking bed availability might need to cross-reference three different spreadsheets, a WhatsApp group, and a phone call to the floor supervisor. By the time the information was assembled, it was already outdated.
The other insight: different roles needed the same data presented differently. Hospital administrators needed aggregate views — total beds, occupancy rates, trend lines. Floor nurses needed real-time, unit-level status: which specific beds were available, which patients were pending transfer. A single dashboard wouldn't serve both.
Design decisions
The portal was structured around three principles:
We designed, tested, and iterated in compressed cycles. Prototypes went to hospital staff for feedback within days, not weeks. Several assumptions we had about navigation and priority were wrong — and only surfaced because we tested early with people under real operational pressure.
What we delivered
The portal replaced spreadsheet-based tracking across participating hospitals. Resource allocation decisions that previously required manual data gathering could now happen from a single screen. The project shipped under the timeline pressure of the pandemic and continued to be refined based on real usage data.
The constraint that defined this project: designing for users who are exhausted, under extreme pressure, and have zero tolerance for a tool that slows them down. Every interface decision was filtered through that reality.