Architecture Documentation
Complete technical overview of the Mapp Size Allocation Intelligence platform
System Architecture
Data Flow Pipeline
Mapp Size Allocation Intelligence is a full-stack analytics platform that optimises size-level stock allocation for fashion retailers. It combines a FastAPI backend with DuckDB analytical storage, a Next.js 15 frontend, and a suite of Python data generators that produce realistic synthetic retail data from real Loft (Ann Taylor) catalogue attributes.
Three-Layer Architecture
| Layer | Technology | Purpose |
|---|---|---|
| Data Pipeline | Python generators + CSV + DuckDB | Generate, validate, and store 3,665 products across 62 stores (60 physical + 2 online DC) with 4 seasons of transactions |
| Intelligence Backend | FastAPI + 8 API routers | Analysis engines, decision generation, simulation, prediction, customer analytics, campaigns, and configuration management |
| Presentation | Next.js 15 + React 19 + Recharts | 20 interactive pages with React Query data fetching and real-time chart visualisation |
Key Numbers
| Metric | Value |
|---|---|
| Products | 3,665 (from Loft catalogue) |
| Stores | 62 (60 physical across 15 US states + 2 online DC) |
| Allocations | ~2.4M records |
| Sales | ~11.6M records |
| Stockouts | ~1.8M records |
| Markdowns | ~1.7M records |
| Returns | ~323K records |
| Customers | 4.7M (supply-side assignment from sales) |
| Customer Transactions | 11.6M records |
| Size Systems | 3 (US Numeric, UK Numeric, Letter) |
| Product Attributes | 15+ (fit, texture, occasion, colour, pattern, wardrobe, range, neckline, sleeve length, length, waist, etc.) |
| Store Archetypes | 5 (Urban, Family, Mature, Student, Online) |
Mapp Size Allocation Intelligence v0.9.0 — Architecture Documentation