StockIQ

Demand forecasting system using SARIMAX — converts raw inventory and order data into reorder recommendations and multi-warehouse decisions.

The Problem

D2C businesses face persistent uncertainty in demand forecasting, multi-warehouse inventory allocation, and reorder decisions. Without a systematic approach, teams rely on intuition and spreadsheets, leading to stockouts, overstocking, and missed revenue opportunities.

Tech Stack

Python

Pandas

NumPy

Statsmodels

FastAPI

Streamlit

Key Features

Probabilistic demand forecasting (P10/P50/P90) using SARIMAX time-series models with configurable horizons
Multi-warehouse inventory optimization with stable allocation shares and configurable risk floors
MOQ-aware reorder logic accounting for minimum order quantities, lead times, and safety stock targets
COD and RTO intelligence providing explainable cash-on-delivery eligibility decisions based on historical patterns

Challenges & Solutions

Probabilistic Forecasting

Problem

Raw historical order data is noisy and seasonal — point forecasts alone are insufficient for inventory decisions with asymmetric costs.

Solution

Implemented SARIMAX models that output P10/P50/P90 probability distributions, allowing stakeholders to make risk-informed decisions based on their tolerance for stockouts vs. overstocking.

Multi-Warehouse Allocation

Problem

Distributing inventory across warehouses while minimizing risk and maximizing fill rates is a constrained optimization problem with competing objectives.

Solution

Built an allocation engine that computes stable shares using historical demand patterns per SKU, with configurable risk floors to prevent any single warehouse from being understocked.

Constraint-Aware Reordering

Problem

Real-world constraints — minimum order quantities, variable lead times, and safety stock targets — interact non-linearly and are often violated by naive reorder points.

Solution

Developed an iterative reorder algorithm that respects all constraints simultaneously, computing the exact quantity and timing for each SKU-warehouse combination.

COD Decision Intelligence

Problem

Cash-on-delivery orders carry return-to-origin (RTO) risk, but blanket restrictions hurt revenue. A transparent, explainable decision was needed.

Solution

Built a rule-based decision system with transparent feature contributions — each COD eligibility decision includes an audit trail showing which factors drove the outcome.

Outcome

StockIQ converts raw operational data into deterministic, explainable business actions: exactly how much to reorder, where to place inventory, and whether to allow COD — reducing guesswork and improving inventory turnover across warehouses.