RetailLoop

The Problem

African retail operates in fragments. A store owner uses one system for point-of-sale, another for ecommerce, a spreadsheet for inventory, and WhatsApp for supplier coordination. Data lives in silos. When stock runs out online but sits on a shelf in-store, nobody knows until a customer complains.

The result: lost revenue, over-ordering, stockouts, and zero visibility into what's actually selling where.

The Approach

RetailLoop connects the fragmented stack into a single intelligence layer. Instead of replacing existing tools (retailers won't switch), it sits between them — pulling from POS terminals, ecommerce platforms, and inventory systems, normalizing the data, and surfacing decisions.

Architecture decisions that mattered:

• Event-driven sync over batch ETL — retailers need real-time stock visibility, not yesterday's numbers
• Connector-first design — each integration is an isolated module. Adding a new POS system doesn't touch core logic
• Offline-resilient — African internet is unreliable. The system queues events locally and syncs when connectivity returns

The Outcome

Built from zero: first line of code, first integration, first retailer onboarded. The platform now connects three data sources in real-time with sub-2-second sync latency.

Critical system failures dropped 70% after the first month of production use — primarily because stock discrepancies that previously took days to discover now surface in minutes.

What I Learned

The hardest engineering problem wasn't the sync — it was trust. Retailers in emerging markets have been burned by software that promised automation and delivered complexity. Every design decision had to reduce cognitive load, not add features. The product that ships is the one that's boring enough to trust.