E-Commerce & Marketplaces

Use a combination of relational (PostgreSQL) for structured data like pricing and inventory, and NoSQL (MongoDB) for flexible product attributes. Implement PIM (Product Information Management) for data governance. Use Elasticsearch for search with denormalized product data. Cache hot products in Redis. Implement product data versioning for audit. Use CDN for images. Design for multi-tenant if marketplace. Partition data by category for scale.

Implement pessimistic locking at database level for inventory updates. Use Redis for real-time inventory counters with Lua scripts for atomic operations. Implement circuit breakers to handle traffic spikes. Use queuing (SQS/Kafka) to serialize inventory updates. Reserve inventory on add-to-cart with timeout-based release. Implement rate limiting per user. Use eventual consistency for analytics but strong consistency for purchases. Consider overselling buffer for popular items.

Combine collaborative filtering (users who bought X also bought Y) with content-based filtering (similar product attributes). Use user behavior signals - views, cart adds, purchases, search queries. Implement real-time recommendations using feature stores. Use ML models (matrix factorization, deep learning) trained on historical data. A/B test recommendation algorithms. Implement hybrid approach with fallback rules. Consider cold-start problem for new users/products. Personalize based on session context.

Implement Saga pattern for distributed transactions across services (inventory, payment, order). Use compensating transactions for rollback. Event sourcing for order state changes. Idempotency keys to handle duplicate requests. Two-phase commit for critical paths. Dead letter queues for failed events. Implement order state machine with clear transitions. Use distributed tracing for debugging. Monitor for orphaned orders and implement reconciliation jobs.

Track order events (delivered, returned, cancelled) in event store. Calculate settlement per seller factoring in commission, shipping subsidy, penalties. Implement T+N settlement cycle (typically T+7). Handle refunds and returns affecting settlement. Deduct TDS at source. Generate detailed settlement reports. Implement reconciliation with logistics and payment data. Handle disputes and adjustments. Comply with GST invoicing requirements. Use batch processing for large-scale settlements.

Minimize checkout steps, implement one-page checkout. Prefetch user data (addresses, payment methods). Lazy load non-critical elements. Implement address autocomplete. Cache payment tokens for returning users. A/B test checkout flows. Implement cart persistence across devices. Optimize for mobile-first. Reduce payment failures with fallback methods. Implement abandoned cart recovery with notifications. Use analytics to identify drop-off points.