Home

Welcome to the Supply-Chain-Analytics wiki!

This project was undertaken as part of the RWTH Aachen Business School Analytics Project for Barkawi Group, a consultancy firm in the field of Supply Chain Optimization. The project was on Supply Chain analytics to evaluate the effect of re-balancing of inventory between retail outlets by modelling the uncertainties of demand using a stochastic dynamic program to optimize the supply plan.

The project on Demand Forecasting was the first phase of this project. The second phase of the project was to import the forecasted values of demand and build a supply plan around this. The original model comprised of a hub spoke model where the inventories of the retail outlets were satiated by ordering the differential quantities from the central warehouse every week subjected to the Minimum Order Quantity constraints. This was refined upon by introducing re-balancing of inventory between the outlets along with replenishment orders from the central warehouse every week. The supply decisions were made once every week; the lead time for receiving orders from the warehouse was 1 week and the lead time for re-balancing of inventory was zero. This was evaluated by building a deterministic dynamic program to minimize the overall cost. The optimization model at each time step was a linear program that minimized the total cost including the transport cost, inventory holding cost and most importantly the cost owing to loss of sales.

The third phase of the project was to model the demand uncertainties into the above deterministic program by using a stochastic dynamic program which worked with demand scenarios generated around the deterministic demand forecasts as mean values. Different demand scenarios were generated considering the growth of the market and probabilities were assigned to each of the demand scenarios using a triangular distribution. This was then programmed into the stochastic optimization program and was solved to generate a supply plan that gave a minimum average value of cost over the demand scenarios.

A rolling horizon kind of approach was followed to simulate the cost incurred over different weeks by solving the optimization model over a two week time horizon period each time. The results of simulations helped evaluate and quantify the hypothesis proposed by the consultants.