Modules running today.
The Granary modules currently running across the 11-store Mumbai pilot. Engine layer (ML Forecasting, Cortex Planning) has been live since 13 - Nov - 2025 with daily updates. The Command Centre is the unified operator surface that consolidates these — five operator screens running on real data today (Home / Category Overview / Range Review / Delist Detail / Requests).
| Module | Status | Anchor outcome |
|---|---|---|
| ML Forecasting Engine | Liveon Databricks | 12K SKUs × 4K stores · 48M rows · MAPE 41% (from 55%+ baseline · research) |
| Cortex Planning Workbench | Livedaily updates | End-to-end Assortment + Inventory Planning · 11-store pilot |
| Assortment Intelligence | Live | Range Review · Classification matrix CX/CY/CZ · Margin-vs-Volume quadrants · multi-select delist with reason classification |
| MBQ Automation | Live | Article-level stock targets per store · replaced state-level manual approach |
| Top 250 SKU availability checklist | Live | Auto-derived top SKUs · shared with store ops for shelf validation |
| Rules Engine | Live | Operator-tunable · audit-grade |
What operators see.
Six surfaces from the Command Centre. Five tagged Live (real article + sales data; running across the 11-store pilot). One tagged Building (Baseline Forecast screen is in build; the underlying forecasting engine is Live on Databricks).
Filter strip (Zone / Format / State / City / Store / Segment / Category / Period) · Exception cards (Zero Sales · Negative Inventory · High DOH · High Markdown) · DOH Distribution Heatmap with per-bucket inventory cost. The exception-driven homepage that surfaces what needs operator attention first.
CX / CY / CZ revenue bands across A / B / C demand stability tiers · Core Stars · Niche Premium · Rationalize · Traffic Drivers quadrants with No-of-SKUs / %-Margin / %-Volume. The decisioning frame for what's strategic, what's reviewable, what to delist.
Sales Trend YoY · Gross Margin Performance · Region-wise Performance · Overstocked + Understocked stores. Category-manager view with drilldown from format to individual store.
9 reason-codes for every delist decision: High inventory holding cost · Product discontinued · Quality/defect · Supplier reliability · End of seasonal lifecycle · Regulatory/compliance · Strategic portfolio optimization · Negative or low profit margin. Audit trail by design.
Article-week forecast across the 11-store pilot. Read view live in SIT; full forecasting screens (consensus, override, scenario) in build. The forecasting engine on Databricks is Live and feeding this view.
Total Sales · Markdown · Margin · Active SKUs · Availability · ROS · DOH metric tiles · Overall Performance (CY vs LY) · Brand Performance top-50 · Sales Trend Analysis. The morning-briefing view category managers and store-ops open first.
