Microsoft is one of the largest tech companies in the world, with a sales revenue of over 90 billion US dollars. More and more of its sales revenue now come from selling physical devices, and this exposes the company to an accompanying level of complexity and uncertainty. The company has over 30,000 products with a large variation in life cycles, more than 600 suppliers, 13 contract manufacturers, and 52 distribution centers distributing products into 191 countries.
In 2015, Goldratt Research Labs was contracted by Microsoft to design, validate, pilot and implement an end-to-end Supply Chain Management solution based on Theory of Constraints (TOC) best practices. This formed part of Microsoft’s “One Devices Supply Chain” (1DSC) solution designed to achieve:
- One set of Planning, Execution and Improvement rules for managing all products and order fulfillment strategies within Microsoft’s supply chain.
- One system (SAP) in which to implement these rules.
A critical part of the project was for Goldratt Research Labs to develop a self-configurable Supply Chain simulation model using AnyLogic to validate the likely operational and financial improvements that would be achieved from the TOC based solution design before going live.
Dr. Alan Barnard and his Goldratt Research Labs (GRL) simulation team lead by Dr. Andrey Malykhanov are very passionate about helping organizations answer two simple but important questions: "How much better can you do?" and "How best to do it?". They work with leading organizations from around the globe, in both private and public sectors, among them: BHP Billiton, Cargill, TATA Steel, ABB, Daiwa House, Utah Gov., and Larsen & Toubro.
GRL uses robust research methods and advanced technologies like Simulation modeling, Artificial Intelligence and Expert Systems to identify how much better organizations can do if they switched from “local optima” to the “global optima” rules of Theory of Constraints (TOC). GRL is also a spearhead for new TOC based research and development, to develop new knowledge and applications of the ever-growing TOC body of knowledge.
The project started in September 2015. The Theory of Constraints principles and best practices were used by GRL to design a world-class supply chain for Microsoft. During the pilot stage, the Theory of Constraints based SCM rules were tested in an AnyLogic simulation model to validate that it will result in better operational and financial performance (e.g. less shortages and surpluses resulting in more revenue, higher profitability with lower inventory). By mid-December 2015, the first set of TOC derived rules for Build-to-Order (BTO) went live within Microsoft’s SAP system. In 2016, additional rules for Build-to-Availability (BTA), “Assemble or Customize-to-Order” (ATO) and “Build-to-X” (BTX) were tested in implemented. The whole project was finished in less than 9 months.
A Strategy-and-Tactic tree (S&T Tree) was developed to show exactly what rules were required to achieve the desired performance improvement and objectives of the 1DSC initiative. As per figure below, six rules needed to be changed:
- Order fulfillment strategy: should it be make-to-order (BTO), assemble-to-order (ATO), make-to-availability (BTA), or build-to-X(where X is launch target quantity) (BTX)
- ATO/BTO buffering: how to quote reliable lead times to BTO/ATO customers.
- BTA/BTX buffering: how to ensure we have sufficient stock of parts and finished goods.
- Release control: how to control the release of BTO/ATO, BTX and BTA orders.
- Shared priority: how to achieve a Single Priority System throughout supply chain.
- Improve flow: how to systematically improve flow.
With the S&T Tree, the Why, What and How of each of these rules were defined to help all stakeholders understand the assumptions behind the rules and to find the simplest way to implement these within standard SAP.
Designing TOC-based SCM solution
The AnyLogic simulation model that was developed to test and verify the operational and financial performance of the new rules.
Supply Chain and Product Data is extracted from the SAP system into an Excel file to fully model and configure the Microsoft global supply chain. Additionally, data of actual customer demand and actual daily on-hand-stock over the modeling period were extracted to allow the model to compare simulated (using the new TOC-based rules) vs. actual past performance.
Randomness was introduced in the model by the Supply chain planning team around such things as variability in production cycle times and distribution supply lead time. Also, information on random events such as planned and unplanned maintenance as well as the impact of positive and negative demand forecast error can be integrated to stress-test the Microsoft supply chain.
The AnyLogic custom model was designed to have a simple user interface to allow stakeholders to manage the scenarios and select the TOC rules to use, the simulation time period, product-channels to include or exclude when running the model without any knowledge of AnyLogic. The output of the simulation is visible directly from the AnyLogic model and can also be exported to Excel. Should results not meet expectations, extensive logs are created to help with diagnostics.
The model can be run in three different ways:
- Single run experiment
- Sensitivity analysis to test how sensitive certain parameters are to the outcomes.
- Scenario comparison
Project basic architecture
The AnyLogic model setup user interface includes network flow and world map to visualize the supply chain configuration for which data was imported, the product categories as well as Replenishment types. Users can view cumulative plots of actual vs. forecast demand vs. production capacity to get a sense of possible issues and users have the ability to use both actual demand forecast and customer orders or to generate forecasts from actual orders or orders from forecasts with a predefined positive or negative forecast error to test how the supply chain will cope with demand that is significantly more or less than expected.
During the running of the simulation model, users can view the operational and financial performance and zoom into specific parts, such as a contract manufacturer to see if backlogs are developing and compare simulated vs. actual stock-on-hand for any specific DC and any specific product to show the benefit of using the TOC rules.
Users can “zoom” into a specific DC and a specific product in that DC to see how the TOC-based rules for managing inventory that dynamically resizes the Target Stock Levels based on “too much red” and “too much green” buffer zone penetration performed against past actual daily stock-on-hand.
Microsoft’s CTO of Global Supply Chain, Robert Meshew, confirmed the incredible results they have achieved since implementing the new TOC-based end-to-end Supply Chain Solution into their SAP system. “The outcome has been nothing short of remarkable. In that time, we've seen our service levels rise to our customers by over 5%. At the same time, we've seen our inventory levels drop by quarter billion dollars across the board which has led to reduced markdowns and reduced excess and obsolescence of over a hundred million dollars”.
Using the capabilities of AnyLogic, offered a fast, low cost and low risk way to validate and stress-test the Theory of Constraints based end-to-end Supply Chain solution design.
The next stage of this project will be to work closely with the Microsoft team lead by Manohar Madhira integrate the model into the company's Sales and Operations planning process as well as to enhance its functionality to be used for “What’s ifs” related to Global Product Launch to support better faster analytics and management decisions.
Project presentation by Dr. Alan Barnard, Goldratt Research Labs