# Demand Forecasting and Inventory Optimiser

Hierarchical SKU-and-store forecasting with automatic model selection. Procurement teams get a 12-month forward view that survives the long tail.

## Key facts

- **Domain:** RETAIL & FMCG
- **Status:** Delivered across multiple FMCG engagements
- **Use when:** Retailers or FMCG groups with thousands of SKUs need to retire gut-feel procurement.
- **Reference engagement:** A regional FMCG retail group

## The problem

Demand patterns vary wildly across thousands of SKUs. Seasonal products, promotional spikes, new launches, and regional variation defeat a single forecasting approach.

## The approach

Hierarchical forecasting with a LightGBM and Prophet ensemble. Each SKU-store combination trains independently. The framework picks the better-performing model on rolling backtests.

## The architecture

Python pipelines orchestrated by Apache Airflow into Snowflake. Model selection, backtest, and forecast generation run on the same monthly cadence. A Tableau exception dashboard surfaces SKUs whose actuals diverge from the procurement plan.

## The outcome

A defensible, auditable forecast for every SKU in the range, with model selection rationale traceable per item. Buyers stop forecasting; they react to exceptions.

## ATI shape

- Predictive intelligence — LightGBM + Prophet ensemble per SKU-store
- Agentic execution — Exception alerter that escalates only divergent SKUs
- Secure data — Pipelines live inside the customer warehouse
- Actionable outcomes — Monthly procurement decisions backed by per-item forecasts

## Tech stack

- Python
- LightGBM
- Facebook Prophet
- Apache Airflow
- Snowflake
- Tableau

## Impact

- Per-SKU — Model selection across the catalogue
- 12 months — Forward forecast horizon
- Exception-driven — Buyers focus only on divergent SKUs

## Related

- [All accelerators](https://levent.ai/accelerators/)
- [Next: Recommendation Engine](https://levent.ai/accelerators/recommendation-engine/)

Metrics on this page are estimated and expected improvements describing the design intent of the accelerator. Real-client delivered metrics stay in private decks; see https://levent.ai/ai-content-policy/ for the abstraction policy.

---

**Canonical URL:** https://levent.ai/accelerators/demand-forecasting/