Optimization: Achieving Maximum ROI within the Supply Chain

Identified by analysts like Bermudez as the most exciting and promising segment within Supply Chain Management (SCM), this technology has moved into the spotlight because software and hardware breakthroughs have converged during a period when businesses are challenged to be more efficient, more effective, and globally competitive. Constantly leapfrogging processor performance along with record-level computer price cuts have produced the cheap horsepower to make sophisticated, power-hungry applications more commonplace. Meanwhile, rapid advances on the software end such as algorithmic breakthroughs and modern, object-based languages are making high-end capabilities affordable and accessible to a whole new audience.

Advanced Algorithms Now Available to Everyone
At the heart of today's supply chain optimization technology are complex algorithms that can examine millions of variables and solve increasingly complex problems in ever-shortening time frames, enabling solutions in a matter of hours rather than days. The knowledge and expertise to design such algorithms are in very short supply, however, residing within a small fraction of the world's software developers. It is generally not possible for all supply chain firms to be able to employ this valuable, scarce resource. Two factors are helping to overcome this limitation. The first factor is object technology, which has empowered the second factor: software components. Object-oriented programming is fundamentally different from traditional procedural programming, with data becoming "self-describing" and thus able to be independent from other data within an application. This capability leads to reusable code, which can save programmers significant development time.

Object technology helped create the software components industry. With the most widely known example being Microsoft's Foundation Classes, software components mainly handle routine functions such as database access. Having reusable hunks of code to address even low-level functions like this saves development time and is welcomed by programmers. However, "there are reusable software components and there are reusable software components," noted analyst Jagi Shahani from International Data Corp.² A few software component vendors that are staffed with hard-to-find algorithm experts are able to offer components that are much higher in the value chain, assisting with an application's problem domain and giving developers access to complex functionality. Such components "deliver savings that far exceed the cost of proprietary development," wrote Shahani. "They are scalable and highly customizable, helping in the rapid creation of applications to meet critical business needs."

Within the SCM industry, the recent availability of extremely sophisticated optimization components now puts the most advanced technology within reach of both SCM application providers and companies designing in-house SCM solutions. According to Benchmarking Partners Inc., a firm that closely tracks developments in SCM, "The market will increasingly license or invest in industrial-strength algorithms, just as they license database technology as a standard operating procedure.³ One of the major SCM trends of the late '90s, the firm noted, is the movement of large ERP suppliers into the growing SCM market. The biggest challenge for top ERP firms like SAP and J.D. Edwards was how to incorporate best-of-breed optimization into their planned SCM products, since designing such technology usually requires more than five years of development work and extensive field experience. Many of these large companies have been able to achieve the time-to-market needed today by licensing intelligent optimization components from front-ranked component vendors, which cuts their development costs and gives their customers highly functional products in much less time.

The CIO of a Fortune 100 food conglomerate advised that "beyond-best-of-breed" ERP solutions are those using the best software components from the leading vendors. In fact, an increasing number of the most successful ERP suppliers are notable not for their skill at designing cutting-edge algorithms and optimization components, but for their strength in seamlessly integrating the best components out there, adding value outside the realm of optimization. Indeed, the optimization component suppliers being sought out by top ERP and SCM application vendors are those that deliver the most sophisticated algorithms and the cleanest APIs within an intelligently constructed optimization framework.

Where Optimization Fits Within SCM ­ and ERP
Industry experts believe Advanced Planning and Scheduling (APS) is the next evolution of manufacturing systems. Performing the all-important planning function, APS is a key part of an SCM application. Over the last three years, most APS products have been embedded with sophisticated optimization logic, making APS almost synonymous with supply chain optimization. APS is considered to be one of the hottest segments in the entire realm of enterprise applications, primarily because of what its optimization technology makes possible. Using its optimization "engines," an APS system synchronizes the customer order workflow with the material, manufacturing, and distribution activity required to deliver the order.

Fueled by its optimization component, APS has quickly leaped to the forefront not just within SCM, but within the entire huge ERP market. For more than 30 years, ERP solutions had at their heart a Materials Resource Planning (MRP) system. This has long been in need of improvement due to the increasing complexity of the business environment. Experts expect optimized APS to have a profound impact on the ERP market itself. In fact, AMR Research recently predicted that optimized APS would drive a $1.4 billion (1998 estimate) market for applications to extend the $15 billion ERP market.

Within the SCM arena, optimization handles both operational and strategic decision-making. Users and vendors of supply chain optimization believe that it is most useful in situations where a company or a product has:

• A complex supply base
• A complex manufacturing process
• A complex distribution system
• Volatile demand

In essence, supply chain optimization is the best solution whenever there is uncertainty in the behavior of supply chain operations or in market demand. But many users report that optimization provides value regardless of the nature of the supply chain.

Where is Optimization Most Helpful?
Supply chain optimization can help those in various functional areas within a company take a broader, yet more effective approach to the whole supply chain, from raw materials to end users. However, the typical decisions most impacted by supply chain optimization are:

• Sourcing
• Make versus buy
• Production allocation
• Supply chain design
• Sales and operations

Operations is a prime area within which to deploy an optimization solution, since the technology helps identify and resolve problems like bottlenecks or unreliable suppliers. The software can suggest improvements and thereby support human planners faced with making difficult decisions. Any manufacturing company in which improving the supply chain can add significant value could conceivably benefit from optimization. However, certain industries are investing more in optimization because of the nature of their business.

Distribution-intensive industries ­
Food and beverage, consumer packaged goods and the like can particularly benefit. Specific applications include transportation optimization and inventory deployment.

Asset- and capacity- intensive industries ­
Examples include semiconductor and steel manufacturers. Key applications are optimization solutions that address throughput times, product mix, and setups (see profile of Chrysler's high-return paint shop optimization system later in this article)..

Material-intensive industries
Apparel and electronics are notable examples. The focus areas for these manufacturers include decisions about what to produce, where to produce it, and who to source materials from in order to generate high margins.

A popular myth that some proponents still mention is that process industries will find supply chain optimization more useful than discrete industries. Although process manufacturers have been using optimization-based solutions for 20 to 30 years, there are now many optimization solutions that address highly complex problems experienced by discrete manufacturers.

Smart Software Engines Drive Optimization
Supply chain optimization components can enable remarkable levels of ROI because, in essence, the software can think much faster than could a single person. The highly sophisticated algorithms at the heart of the optimization components are tailored to tackle specific problems. Unable to replicate the human thinking process, of course, the system is nevertheless expert at quickly analyzing the implications of alternative decisions, reviewing massive amounts of data instantly. Modern optimization technology has evolved as a result of advances in mathematics and computer science research that have taken place over the past several decades.

Figure 1.	Map of the main optimization problems

Other quickly evolving elements in supply chain optimization are the mathematical programming techniques used to design the advanced algorithms at its heart. Linear programming (LP) and other closely related techniques have been extensively used since the middle of the century for optimization. The early applications for LP that yielded the biggest benefits involved creating schedules for massive capital investments such as airline flight schedules, oil refinery blending schedules, and train or bus schedules. However, within today's ever-increasing competitive markets, additional requirements have been added. While LP remains an effective technique to solve a variety of planning problems, a newer technique called constraint programming (CP) was developed for problems in which the availability of resources is constrained and real-time scheduling decisions are required. Today, many APS systems are constraint-based, that is, they consider the real-time availability of each resource before determining manufacturing and shipment dates. Optimization is now performed by specialized software systems that are specific to various types of problems. In general, long-term planning problems are solved through LP and short-term scheduling problems are addressed by CP.

Figure 2.	Application types and optimization techniques

Problems Addressed by Optimization
To understand the fundamentals of optimization, a decision maker must first understand the attributes of the business questions that need answers. The following strategic business issues must be addressed in the design of a supply chain.

• How many manufacturing plants are required? Where should they be located? Which products should they produce?
• How many distribution centers are required? Where should they be located? Which products should they stock and at what inventory levels? Which customer should the distribution center serve?
• Which suppliers should we select?

Figure 3.	Example of Supply Chain Design

Each of these questions is interconnected. The answer to one affects the solution to another (as shown in Figure 3).

Historically, answering these questions involved a series of "what-if" scenarios. Experienced analysts could produce a good solution after substantial analysis. New optimization techniques improve on this approach to find the best, or optimal, solution.

Closed-form mathematical problems can be utilized to represent numerous business questions, such as those listed above. This leads to the belief that analytical techniques can be used to find the optimal solution. However, since the underlying mathematical problems involve an enormous number of variables and constraints, they cannot be solved manually.

Specialized "solvers" have been required and developed to address these questions. The solvers are presented in the form of an optimization model. In order to create an optimization model, the attributes of each question must be described in terms of variables, constraints, descriptive data, and objective functions.

Figure 4.	Supply chain demand optimization attributes

In the case of the supply chain design problem presented in the section 'Problems Addressed by Optimization,' the table in Figure 4 outlines the inputs of the optimization model.

In addition to the information above, the optimization model requires an objective function. The objective function of the business problem represents the primary goal to be achieved. Possible objective functions include minimizing cost, maximizing customer service levels, or minimizing cycle time. In a commodity market, minimizing cost may be the main objective. Alternatively, for companies trying to get to the market in the short rush of a fad, reducing cycle time may be the key.

Once the variables, constraints, descriptive data, and objective functions are identified, the optimization model is mapped to the appropriate optimization solution technique.

Figure 5.	Chrysler's Vehicle Schedule Problem

The Competitive Advantage
The benefits of utilizing optimization techniques on a supply chain can be substantial, as illustrated in Figure 5.

APS systems must offer a broad range of optimizer solution techniques to solve a company's wide variety of supply chain issues. Additionally, an APS system that is marketed to a variety of industries needs a rich set of optimization techniques to ensure that the appropriate techniques are available for each industry. APS vendors run advanced optimization techniques in the background, but their competitive advantage ultimately comes from business issues they address, such as the footprint of supply chain issues they cover, the industry-specific solutions they have developed, and the ease of implementation of their product.

Optimization solvers themselves do not provide a competitive advantage. Instead, they enable the technology that allows an APS system to address numerous complex business issues. To solve the range of supply chain optimization problems that an APS system must address, a variety of optimization solvers are required. Not having access to the appropriate solver for a business problem places an APS provider at a competitive disadvantage.

APS vendors can obtain the latest optimization solvers and tools from companies that specialize in developing solvers. For most APS companies, developing solvers is not an effective utilization of internal resources. In addition, optimization solvers are the core competency of the specialty companies, who stay on the leading edge of solver technology. Internally developed solvers generally lag behind the market.

Optimization Tool Suites
If an APS vendor outsources its solvers to cover a broad set of supply chain issues, its technical staff may well end up dealing with a variety of solvers, optimization languages, data structures, and solver vendor support staffs. An optimization suite from a single vendor provides an extensive set of optimization alternatives while only requiring one data interface and one supplier relationship. One-stop shopping with an optimization supplier improves the effectiveness of an APS vendor's technical staff.

An Example of Supply Chain Optimization ­ The Chrysler Corporation
The Chrysler Corporation is the fourth largest manufacturer of vehicles in the world. Chrysler knows that automotive manufacturing is among the most complex of manufacturing environments. Two of the most important challenges an automotive manufacturing company, such as Chrysler, faces are those of:

• Scheduling vehicle production.
• Providing its enormous supplier base with timely and accurate orders for material and components.

In the case of Chrysler, even though tracking these two areas is difficult, success would enable the manufacturing operation to provide a competitive advantage over the competition.

Figure 6.	Results from Chrysler's Scheduling System

Chrysler's vehicle scheduling problem is highly complex, as illustrated in Figure 5.

Chrysler's legacy system was slow and cumbersome, and did not provide adequate optimization. To overcome this, Chrysler considered purchasing a general-purpose APS system, but the solutions available were not specific enough to meet Chrysler's requirements. Since vehicle scheduling is a planning operation of strategic importance for the company, Chrysler needed a point solution to address this issue. However, Chrysler did not want to dedicate internal resources to develop leading-edge optimization technology.

Following extensive research, Chrysler found that the ILOG Optimization Suite provided the optimization capability that it was looking for. Additionally, ILOG's proposal provided a modeling environment that allowed Chrysler to quickly develop their initial solution and made it easy for them to maintain and upgrade their existing systems. The results of Chrysler's new vehicle scheduling system are impressive, as shown in Figure 6.

The overall development cost for the project was less than $500,000 and the initial pilot took just three months to develop. The development and testing were performed on a 133MHz Pentium PC with only 32 megabytes of memory. Chrysler estimated the benefits of the first year of use to be over $7 million, yielding a payback period of less than one fiscal year.

New Developments in Optimization
When applying mathematical programming or constraint programming techniques to solve optimization problems today, users must know how to model. In addition, they have to be able to write software to generate matrices and call a solver, and either use mathematical programming languages or write software in C++, using a solver or using a specialized computer programming language that supports constraint programming. In all cases, computer programming has always been required. Also, a language has never existed that bridges constraint and mathematical programming.

This is all going to change in the very near future as a result of breakthrough products that greatly simplify application development. Available this year will be a new generation of optimization components that bridge the gap between LP and CP while also enabling automatic code generation. This will have a profound impact on the supply chain optimization industry.

CONCLUSION
Supply chain optimization is the area within SCM that can most impact efficiency and deliver maximum ROI. Decision makers charged with taking time and cost out of their company's supply chain should closely examine their operation and look for prime areas in which to deploy this technology. Indeed, using supply chain optimization might well move from a competitive differentiator to a must-have for companies seeking to maximize profits. However, manufacturers should seek out SCM applications that include the most sophisticated, high-return optimization components before implementing a solution.

The many benefits to be reaped by effective supply chain optimization schemes will only increase. In fact, this key area will soon see technical breakthroughs that will greatly broaden the market, bringing this high-end technology to even more users.

About the Author
Dr. Irvin Lustig, a key member of the optimization business unit at ILOG, is responsible for the CPLEX, Planner, AMPL and ILOG OPL Studio products. He was one of the lead developers of CPLEX, including the CPLEX Barrier Solver. Prior to joining ILOG, Dr. Lustig was a professor at Princeton University during which time he was awarded the 1991 Beale-Orchard Hayes Prize for Excellence in Computational Mathematical Programming from the Mathematical Programming Society, and the 1992 Operations Research Society Computer Technical Section Prize. Dr. Lustig received his PhD in Operations Research from Stanford University and a ScB in Applied Mathematics/Computer Science from Brown University. He is the author of over 30 scientific research papers.

Footnotes:

1. "The Report on Supply Chain Management," December, 1998
2. Object Tools Bulletin, "ILOG: High Value Components," February, 1997
3. "Supply Chain Market Review," July, 1998