Back to Blog
ArchitectureReact NativeAmazon SP-APIScaling

Building ScoutIQ: Scaling to 10 Million Product Scans per Week

Lavneet Sharma··8 min read

When I joined ScoutIQ in 2019, it was already a popular tool among Amazon book sellers. But the architecture was showing its age. Sellers in the field needed instant results - they're standing in thrift stores scanning hundreds of books, and every second of latency costs money.

The Challenge

Amazon sellers who do retail arbitrage need to make split-second decisions. They scan a book's barcode and need to know immediately: Is this profitable? What will it sell for? How fast will it sell? Our platform needed to answer all of these questions in under a second, even with spotty cellular reception in rural thrift stores.

The eScore Algorithm

One of my proudest technical achievements at ScoutIQ was engineering the eScore system - a proprietary algorithm that calculates real-time product profitability and sales velocity. The eScore analyzes 6 months of sales frequency data to predict demand, giving sellers a single number they can trust.

eScore = f(salesRank, salesFrequency, competitionLevel, priceStability)

Rather than just showing the current Amazon sales rank (which fluctuates wildly), eScore provides a normalized, predictive score. A book with an eScore of 8+ is almost certainly going to sell within 30 days.

Offline-First Architecture

The biggest technical challenge was building an offline database containing 9M+ items that could deliver sub-second scan results. Sellers often work in areas with poor connectivity - basements of library sales, rural Goodwill stores, estate sales in the middle of nowhere.

We built a custom SQLite-based solution that syncs incrementally:

  • Initial download: ~2GB compressed database
  • Daily deltas: Only changed records sync overnight
  • Query optimization: Custom indexes for barcode lookups achieving <100ms response times

Smart Triggers

We also built smart triggers - an automated accept/reject system that factors in:

  • Current Amazon selling price
  • FBA fees (referral, fulfillment, storage)
  • Seller's cost basis
  • Minimum acceptable profit margin
  • Sales velocity threshold

Sellers could configure their triggers once and then just scan - the app would beep green for "buy" and red for "skip," turning the scanning process into a highly efficient operation.

The Acquisition

In 2022, ScoutIQ was acquired by Threecolts Inc. (UK) in a multi-million dollar deal. I continued for 1.5 years post-acquisition, ensuring seamless integration into their Seller 365 suite alongside other tools like RepriceIQ, PalletIQ, and NuPrice.

Key Takeaways

  1. Offline-first isn't optional for field applications
  2. Proprietary algorithms create defensible moats
  3. Sub-second latency directly impacts user revenue and retention
  4. Incremental sync beats full database downloads every time
  5. Smart defaults (triggers) turn complex decisions into simple yes/no signals

The ScoutIQ years were defining for me - it was where I learned to build systems that directly impact people's livelihoods. When your platform processes 10 million scans a week, every optimization matters, every bug costs someone money, and every feature either accelerates or blocks real-world commerce.