A 180-store apparel retailer suffered frequent oversold online orders because store point-of-sale systems, warehouse software, and e-commerce synced only once a night via fragile FTP CSV files. SNWN built an event-driven integration layer using Apache Kafka and idempotent REST APIs connecting SAP, Salesforce, Shopify Plus, and store scanners in real time. Stock accuracy climbed to 99.6%, oversold order refunds dropped by 76%, and same-day store pickup grew 2.3x.
Why Overnight Batch Files Broke Same-Day Store Pickup
Shoppers expected to check store availability on their phone, buy online at lunch, and pick up their bag on the way home. Because store registers updated central e-commerce stock only at 2 AM, items sold at 11 AM in-store still showed as available online all afternoon.
Key Operational Bottlenecks Before the Project
- 11% of buy-online-pickup-in-store (BOPIS) orders had to be canceled because the last item sold on the shop floor hours earlier.
- Nightly FTP sync scripts failed twice a month during holiday sales, freezing warehouse pick lists until mid-morning.
- Customer service reps had to log into four separate screens to check an order, loyalty balance, or return status.
How We Connected Stores, Warehouses, and Online Checkout in Real Time
Instead of ripping out SAP or store registers, we placed a clean, fault-tolerant event hub in the middle so every system shares live stock updates within seconds.
Untangled Point-to-Point Scripts Into Clean Canonical Data Schemas
We mapped 26 brittle point-to-point scripts across SAP ERP, warehouse WMS, Shopify Plus, and Salesforce CRM, defining one clean JSON format for SKUs, stock reservations, orders, and returns.
Deployed Managed Kafka Streams for Sub-Second Stock Updates
Every barcode scan at a cash register, warehouse dock receipt, or online cart checkout now publishes an immutable inventory event to Kafka, updating available-to-promise counts across all channels in under two seconds.
Built Store-Offline Buffering and Dead-Letter Auto-Retry Logic
Retail mall internet drops happen. Store POS gateways now queue transactions locally if broadband flickers and replay them safely with idempotency keys the moment connectivity returns - preventing duplicate charges or lost stock counts.
Unified Customer Loyalty and Order History Inside One CRM View
We exposed a clean GraphQL gateway for store associates and support reps, pulling online purchases, in-store receipts, and loyalty points into one fast Salesforce screen.
Fewer Canceled Orders and Smooth Black Friday Peak Traffic
Store teams stopped apologizing for missing pickup orders, and the retailer sailed through Black Friday at four times normal checkout volume with zero sync delays.
Cross-Channel Stock Accuracy
Live inventory parity across 180 retail stores, 3 warehouses, and e-commerce.
Drop in Oversold Cancellations
Saved hundreds of thousands in refunded orders and apology discount vouchers.
End-to-End POS to Web Sync
Replaced 24-hour overnight FTP batch windows with continuous event streaming.
Day-to-Day Business Impact
- Same-day click-and-collect orders grew 2.3x within six months of launch.
- Support agents cut average customer lookup time from 4 minutes to 35 seconds.
- Processed 42,000 peak-day orders over Black Friday weekend with zero dropped webhooks.
Tools, Platforms, and Technologies Used
Every component in this deployment was selected for long-term stability, clear observability, and straightforward handover so internal engineers never get locked into black-box scripts.
ERP, CRM & API Integration
See how our specialists design, migrate, and manage workloads like this every day.
Frequently Asked Questions About This Project
Did the retailer have to replace its legacy SAP ERP or store POS hardware?
No. SNWN built lightweight adapter microservices around the existing SAP and POS interfaces, modernizing data sync speed without a costly multi-year ERP replacement.
How do you prevent two shoppers from buying the last item at the exact same second?
Atomic Redis stock-reservation locks hold the SKU during checkout while Kafka reconciles confirmed ledger events, preventing double-allocation during flash sales.
What happens if an external API like Salesforce or a carrier webhook slows down?
Kafka decouples producers from slow downstream APIs. Events stay safely queued with exponential backoff and dead-letter alerting so checkout never slows down for shoppers.
