Poster

Client Context

Starbucks, an international coffee roaster and coffeehouse chain, has a supply chain network of five regional distribution centers (RDCs). These RDCs supply finished non-refrigerated goods to company-operated stores and licensed stores. As it is a manually-operated facility with an estimated 6% expected demand growth rate in the upcoming fiscal year, the Nashville RDC was identified as a valuable area of improvement.

Project Objective

During demand spikes, the Nashville RDC relies on overtime labor and contract workers to ensure service levels are met, leading to an increase in the overall planning and fulfillment costs. Between October 2018 and June 2019, these extra labor costs made up 22% of their total labor costs. Additionally, Starbucks estimates that they spend 1,350 hours annually reactively reworking demand allocations and order schedules, as there is not a standard measure of throughput capacity across their RDCs. Thus, RDCs are given more demand than they can fulfill or spend unnecessary costs for extra labor.

We identified two primary opportunities for this project. First, we aimed to increase throughput capacity in two ways. We wanted to reallocate slots based on SKU demand to reduce the risk of a delayed pick, as 22 case-picking hours are lost daily due to delayed picks caused by insufficient inventory at pickfronts. We then wanted to relocate SKUs along a new pick path to better utilize cross-aisles and reduce picking time, as 82% of picking time is spent traveling. Our second opportunity was to improve operational planning due to the demand allocation reworks explained above.

Design Strategy

To increase throughput capacity and mitigate risk of delayed picks, an optimization model was developed to reallocate the number of slots assigned to each SKU. Then, the number of slots assigned to each SKU was compared to the number of pallets of the SKU picked each day to ensure that enough inventory was available to meet demand. Then, the SKU locations were reassessed based on priorities such as density and velocity. The new locations were validated using a Python-based simulation to estimate travel time improvements due to the new pick path. To improve operational planning, a capacity estimation model was designed using a regression-based method that utilizes the attributes of forecasted demand to estimate the maximum daily demand a RDC can handle. The regression model was then applied to 2 weeks’ worth of demand data to compare the estimated capacity to the demand the RDC processed.

Deliverables

For each of the three recommendations, the Nashville RDC management team will receive a business case, estimated savings, and standard operating procedures for the best case and alternative strategies. The models for the strategies relating to the first objective are packaged into an Excel tool and Python script. These models should be run on a monthly basis to take into account both historical and forecasted demand information. The management team will also receive an Excel dashboard with a built in regression model to use when determining their throughput capacity and planning labor.

Project Information

Spring 2020
Starbucks

Student Team

Lauren Boddiford, Jared Borders, Mihir Dhoot, Ilesh Jain, Wendy Ng, Savannah Quinn, Olivia Ulrich, Llewellyn Weeks

Faculty Advisor

Faculty Evaluator