Amazon Seller Tools: A Technical Deep Dive

We've built multiple Amazon seller tools - from Chrome extensions that overlay live data on product pages to full analytics dashboards tracking sales velocity across thousands of ASINs. Here's how we do it.

The Amazon SP-API Landscape

Amazon's Selling Partner API (SP-API) replaced MWS in 2023, and it's both more powerful and more annoying to work with. Key challenges:

• Rate limiting - different endpoints have different throttle rates
• Token management - LWA (Login with Amazon) OAuth flow with refresh tokens
• Data latency - some reports take hours to generate
• Regional endpoints - NA, EU, and FE have separate API hosts

Our SP-API Integration Pattern

We built a reusable integration layer that handles:

1. Automatic token refresh - background job refreshes tokens before expiry

2. Smart rate limiting - request queue that respects per-endpoint throttle rates

3. Retry with exponential backoff - handles transient failures gracefully

4. Multi-marketplace support - single seller account spanning US, CA, MX, UK, etc.

Real-Time Data Pipeline

The core of any Amazon tool is its data pipeline. Here's our production architecture:

Ingestion Layer

• Cron jobs trigger data pulls every 15 minutes for active metrics (sales, inventory)
• Bull queues manage parallel API calls across hundreds of seller accounts
• Redis caches intermediate results and deduplicates requests

Processing Layer

• Node.js workers parse raw API responses and normalize data
• PostgreSQL + TimescaleDB stores time-series data with automatic partitioning
• Materialized views pre-compute expensive aggregations (daily sales, weekly trends)

Presentation Layer

• React + D3.js renders interactive charts and dashboards
• WebSocket connections push real-time updates to connected clients
• Export service generates CSV/Excel reports on demand

Key Features We've Built

Sales Velocity Tracking

Real-time revenue, units sold, and profit margin calculations. The tricky part? Amazon's fee structure changes frequently, and FBA fees depend on product dimensions, weight, and category.

Keyword Rank Monitoring

We track keyword positions hourly across 50+ keywords per product. This generates massive amounts of data - a seller with 100 products tracking 50 keywords each produces 120,000 data points per day.

Competitor Price Monitoring

Automated tracking of competitor prices, reviews, and BSR (Best Seller Rank) using a combination of SP-API data and Keepa API.

Inventory Forecasting

ML-powered demand prediction that factors in:

• Historical sales velocity
• Seasonal trends
• Lead time from supplier
• Current inventory levels
• Upcoming promotions

Performance at Scale

Our largest deployment processes 2M+ data points daily across 500+ seller accounts:

Lessons Learned

1. Cache aggressively - Amazon API calls are expensive (rate limits, not cost). Cache everything that doesn't change frequently.

2. Design for eventual consistency - Amazon's data is eventually consistent. Your UI should handle this gracefully.

3. Build for multi-marketplace from day one - retrofitting multi-marketplace support is painful.

Building an Amazon tool? Let's talk - we've shipped multiple successful Amazon analytics platforms.