The Returns Visibility Problem
In e-commerce, returns are inevitable. But at scale, managing returns becomes incredibly complex. When I joined the returns team at Walmart Global Tech, associates were juggling multiple systems to track a single return through its lifecycle. We needed a unified dashboard that brought together data from multiple sources into a single, real-time view.
Architecture Decisions
GraphQL as the Data Layer
I chose GraphQL for several reasons:
1. Data Aggregation
GraphQL's ability to stitch together data from multiple sources was perfect for our use case. A single query could fetch order details, return status, shipping info, and refund status.
2. Frontend Flexibility
With GraphQL, the frontend team could iterate quickly without backend changes. Need a new field? Just add it to the query.
3. Performance Through Batching
Using DataLoader, we batched and cached database calls, reducing load on downstream services by 70%.
Tech Stack
- Backend: Node.js with Apollo Server
- Frontend: React with Apollo Client
- Data Sources: REST APIs wrapped as GraphQL data sources
- Real-time Updates: GraphQL Subscriptions over WebSockets
- Caching: Redis for response caching
Building the Timeline View
The timeline view was the most challenging and most valuable feature. Associates needed to see the complete journey of a return at a glance.
Real-Time Updates with Subscriptions
The game-changer was adding real-time updates. When a return's status changed, all viewing associates saw the update instantly using GraphQL subscriptions.
Impact Metrics
The dashboard's impact was measurable and immediate:
- 60% reduction in average handle time per return
- Increased visibility - all stakeholders could track returns in real-time
- 30% faster refund processing
- Adopted by 200+ associates across customer service teams
- Standardized tracking led to better reporting and analytics
Developer Experience
Building with GraphQL made the entire development process smoother:
- Strong typing caught bugs at compile time
- GraphQL Playground made API exploration easy
- Automated documentation from schema reduced onboarding time
- Code generation from schema to TypeScript types eliminated boilerplate
This project reinforced my belief that the best internal tools are ones that get out of the way and let people do their jobs effectively. By focusing on real-time data, standardization, and excellent performance, we built something that associates actually wanted to use every day.