Optimizing an Apparel Company’s Supply · Optimizing an Apparel Company’s Supply Chain by...
Transcript of Optimizing an Apparel Company’s Supply · Optimizing an Apparel Company’s Supply Chain by...
Optimizing an Apparel Company’s Supply Chain by Combining Agent-Based Modeling
with Geographic Information Systems
Beth Tyrie
Manager, Data Science
11/4/2015
• 4 Separate Business Units
• SKU and retailer intensive company
Problem Statement
• Will it benefit the company in terms of cost and supply chain optimization to add a new distribution center (DC)
on the East coast or West coast or will it benefit the company to redistribute products to a pre-existing DC?
Simulation Modeling Importance
• Solves real-world problems without “real-world” experiment costs
• Enables abstraction
• Simplifies complex systems by parameterizing only relevant details
• Simulations can be conducted faster than real time
• Allows the testing of multiple variations of the experiment
Systems Engineering Fundamentals.Defense Acquisition University Press, 2001
Modeling Software: AnyLogic
“The only simulation tool that supports Discrete Event, Agent Based, and System Dynamics Simulation”
Supply Chain Agent-Based Modeling
• Supply chain participant examples:
• Producers (Cotton farmers)
• Processors (Yarn mills)
• Companies ( FOTL distribution centers)
• Wholesalers
• Retailers
• Each have their own goals and rules and can naturally be represented as agents
Data Collection
Data Required
• DC Customer Shipment Data
• High Demand Customer locations
• Total shipments, total units
• Shipment Types and Rates
• Truckload, LTL, Rail
• Distance from DC to Customers
• Overhead Cost Estimates
• Cost to build new DC
• Fixed and variable costs per products
AnyLogic
DC to Customer Shipment
Data
Shipment Types and
Rates
Distance
DC->
Customers
Overhead Cost
Estimates
Collaborative Data Collection
• Data Warehouse
• Logistics Planning and Analysis
• Transportation Analysis
• Process Engineering
• Simulations MIS dept
• IT- In Transit
• Business Solutions Manufacturing Systems
http://pardington10.wikis.birmingham.k12.mi.us/Collaboration+Techniques
Data Exploration: Outlier Detection
GIS (Geographic Information Systems)
• Geographically referenced data (i.e. customer location data) can be spatially visualized and analyzed
• Spatial relationships, patterns, and trends are revealed that are not readily apparent in spreadsheets leading to:• Cost reduction• Identification of
opportunities• Streamlined operation
http://www.esri.com/
GIS Network Analysis of Original DC to Customers
Note: Mock Data
Chosen Location: McCook, NE
GIS Network Analysis to Determine Optimal Location for DC based on Weighted Distances
Note: Mock Data
Model Calculations• Model Probabilities:
• Percentage of Shipments per Customer:
• Demand per Shipment in Units per Customer:
• Benefits:
• Model does not depend on single orders
• Model can be flexible in terms of total units and shipments
• Model is equipped to be predictive
custNumShipments
totalShipments
(custNumShipments/ totalCustUnits)
totalShippedUnits
Model Assumptions
• Latitude/Longitude calculation by Haversine Formula used by NOAA and NASA
• Cannot take into account road distances; however, 325 mile error was reasonable in terms of total cost
• Currently, model does not take into account small package or consolidated shipments
• Approximately 85% of shipments modeled
Model Agents• Distribution Centers
• Parameters: location, units, overHeadCost, startUpCost
• Customers
• Parameters: location, demandRate, totalShipments, distance, freightRate, shipmentType
• Trucks
• Parameters: location, units, owner, destination
• Trains
• Parameters: location, units, owner, destination
Demo
Simulation Results Excel File
Code Behind Simulation: Action Charts
Java Code Behind Simulation: Events
Java Code Behind Simulation: FunctionsCalculate DC with Shortest Distance to Customer Create New DC
Manufacturers to DCs Model• Manufacturers
• Distribution Centers
• Parameters: location, shipPercent, distance
• Loading Ports
• Parameters: location, annual volume
• Ports of Discharge
• Parameters: location, distance
• Vessels
• Parameters: location, owner, destination, dc, oceanRate
• Trucks
• Parameters: location, owner, destination
Conclusions
• Simulations used as exploratory research tool for the business to investigate feasibility of recommendations
• Data-driven insights from GIS and AnyLogic paired with business knowledge creates a full-fledged approach to develop informed supply chain decisions
Thank you!