Thesis/Capstone
Publication Date
Authored by
Topic(s) Covered:
  • Product Development
  • Forecasting
Abstract

This project explores how we can develop a Demand Signal Repository (DSR) to better predict demand. We devised a methodology, collected and cleansed the data, and built more than 240 models before we arrived at models that could improve forecast accuracy. The models explained 85%+ of weekly demand per customer per region and 60%+ on a daily basis compared to time-series models that remained below 50% on a monthly basis. We applied the model for predicting bottled water sales, but a similar approach can be taken with other products.

Authors: Ahmed Rashad and Santiago Spraggon
Advisor: Dr. Shardul Phadnis

Attachment(s)
Presentation
Executive Summary
Abstract