Offline-First Architecture: Why It's Non-Negotiable for West African Deployment

In 2022, we deployed a health information system across 120 public health facilities in Ghana. Before we wrote a single line of code, we spent three days at facilities in the Brong-Ahafo region with our laptop, a mobile hotspot, and a test environment. What we found shaped every technical decision that followed.

The connectivity situation across many facilities was not intermittent — it was absent. Facilities in semi-urban areas might have 2G connectivity for two to four hours a day, depending on network load and weather. Facilities in rural catchment areas sometimes had no connectivity for days at a time. If our system required a live connection to function, it would fail the clinical staff who needed it most.

Offline-first is an architectural principle, not a feature. It means the application works completely and correctly without any network connection. When connectivity is available, it synchronizes — intelligently, efficiently, and without corrupting data. This is harder than it sounds, and most developers underestimate it.

The core technical challenge is conflict resolution. If a nurse in Accra and a nurse in Kumasi both update the same patient record while offline, and both subsequently sync, whose version wins? The naive answer — 'last write wins' — is medically dangerous. Our answer was a merge strategy that preserved all edits as distinct events, surfacing conflicts to a designated resolver rather than silently dropping data.

We chose IndexedDB as our local data store, with a custom sync engine that serialized operations as an append-only event log rather than state snapshots. This design meant that even a partial sync — if the connection dropped mid-transfer — left the database in a consistent state that could be cleanly resumed rather than rolled back.

Data compression was critical. A full patient record might contain scans, lab results, and clinical notes. Over a 2G connection with 50KB/s bandwidth, naive syncing would take hours. We implemented delta syncing — only transmitting what had changed since the last successful sync — and compressed payloads using gzip. Average sync payloads dropped by 85%.

The result was a system that worked everywhere, every time. Clinical staff adopted it quickly because it never failed them due to connectivity issues. Within six months of deployment, 96% of clinical interactions were documented digitally — a dramatic improvement from the paper-based baseline.

Our hard-won advice: if you're building for deployment in connectivity-constrained environments, start with offline-first as your primary design constraint, not an afterthought. Everything else — data models, sync protocols, UI patterns — flows from that foundational decision.