TJ

Tushar Jalan — Sales & Inventory Optimization

BDM Capstone Project: Agarwalla Masala

A data-driven strategy to solve real-world inventory challenges for a family-run supermarket, using one year of ledger data to create a production-ready optimization plan.

222
Unique SKUs Analyzed
27
"Dead Stock" SKUs Found
3-5%
Projected Profit Increase
15-Day
Strict Storage Constraint

Project Summary & Methodology

This project tackles the critical operational challenges of Agarwalla Masala, a residential supermarket. The core goal is to transition the business from reactive decision-making to a proactive, data-driven strategy to enhance profitability and ensure sustainable growth despite significant physical and logistical constraints.

The Problem

  • Severe Space Constraint: The store's physical storage can only hold about 15 days' worth of inventory, creating a constant risk of stockouts.
  • Seasonal Stockouts: Predictable demand spikes of 40-60% during festivals like Diwali and Holi frequently lead to stockouts of high-demand items due to long supplier lead times.
  • Poor Data Quality: The raw data suffered from significant inconsistencies, most notably the "62 Brands Problem," which fragmented transaction histories across ambiguous names.
  • Inefficient Capital: A lack of granular analysis meant capital was tied up in underperforming products, while fast-movers were understocked.

The Methodology

  • Ledger-Only Approach: Used the Stock_Ledger_Summary.csv as a single source of truth to ensure 100% analytical consistency after initial multi-file reconciliation attempts failed.
  • Automated Master View: Developed a Python script to construct a comprehensive view of all 222 SKUs, calculating metrics like revenue, gross margin, ROI, and Days of Inventory.
  • ABC-XYZ Segmentation: Classified products by revenue contribution (ABC) and sales volume volatility (XYZ) to create a multi-dimensional framework for strategic inventory decisions.
  • Seasonal Uplift Model: Built a dynamic planning model that quantifies historical demand spikes to generate a forward-looking inventory plan that respects the 15-day stock cover constraint.

Key Findings & Visualizations

The analysis confirmed a strong Pareto distribution in sales, identified significant inefficiencies in inventory health, and quantified the extreme seasonality of demand.

Finding 1: The Pareto Principle

The top 20% of products (Class A & B) generate 90% of the total revenue, confirming a heavy reliance on a small core of items.

Finding 2: Inventory Health

27 SKUs were identified as "Dead Stock," tying up ₹20,930.72 in capital while generating zero sales.

Finding 3: Quantified Seasonal Demand

Niche categories like "Puja Items" exhibit extreme volatility, with sales spiking over 3 times their January levels during festivals, while core FMCG sales remain relatively stable.

Project Journey — Challenges & Solutions

1. Tackling the "62 Brands" Data Mess

The initial dataset was plagued with inconsistent entries, with over 280 raw "brand" names for just 222 products due to misspellings and data entry errors. Solution: I developed a rule-based Python script to systematically clean and consolidate 62 ambiguous entries. This meticulous process was crucial for creating an accurate and analyzable master dataset.

2. Pivoting to a "Ledger-Only" Approach

My initial plan to merge multiple files failed due to "insurmountable data integrity failures" that produced incorrect financial metrics. Solution: I made the strategic decision to pivot to a "Ledger-Only" methodology. This ensured 100% consistency by deriving all key metrics—sales, cost, and stock flow—from a single, reliable source of truth.

3. Forecasting Within a 15-Day Constraint

Standard forecasting models were impractical because they don't account for physical storage limits. Solution: I built a custom inventory planning model that calculates a monthly Uplift_Factor for demand and translates it into a concrete Target_Inventory level that respects the store's strict 15-day stock cover, making the recommendations realistic and actionable.

4. Bridging Data Science and Business Needs

The store owner is not a data analyst and needed simple, actionable tools, not complex code. Solution: The final deliverables included not just a report, but a user-friendly Forward Inventory Plan.xlsx file and a recommendation for a simple daily staff checklist to improve future data quality, ensuring the project's impact would last.