Supply Chain Optimization: Just the Facts

In the area of supply chain planning, advanced planning and scheduling (APS) vendors have put together solutions that help planners make better decisions. For planners, APS quickly analyzes the implications of alternative decisions. By performing what-if simulation analyses, APS systems provide information about whether plans are reasonable or if they, for example, exceed resource constraints or result in inadequate performance.

Recently, in a fashion similar to the AI efforts, there has been a trend to embed sophisticated optimization logic into APS suites to improve decisions of supply chain planners. If used successfully, this type of optimization promises to drastically improve a company's supply chain performance in a variety of areas:

• Reduce supply chain costs
• Improve product margins
• Lower inventories
• Increase manufacturing throughput
• Better return on assets

This potential for improvement is generating a great deal of interest in supply chain optimization. Many established and startup APS vendors are now using the concept as a selling point, and in some cases, as a key marketing differentiator. While optimization methods have been around since post World War II--with the advent of operations research and management sciences--there has been only a marginal interest in applying these concepts to supply chain planning. Will companies embrace these newer versions of optimization technology? The answer is yes. What company would not want to optimize its supply chain? These solutions, however, are expensive. Do all buyers really understand how to apply these solutions and what they are getting for their money? Is optimization worth the cost? Does it really work? AMR addresses these questions by focusing on supply chain optimization technology from leading APS providers. Our paper covers the following topics:

• The growing optimization technology market and reasons for its growth
• The supply chain planning optimization framework
• The importance of data, models, and solvers in optimization
• The holistic optimization approaches vendors use within applications
• Optimization technology design issues
• Guidelines for the use of optimization in supply chain planning

The Market for Optimization Is Growing
Today's market dynamics have made supply chains extremely complex and planning more difficult. The following true story is a case in point:

ABC Inc., a small, 100-year-old agricultural seed company called in a consultant to work on a production scheduling problem. The consultant talked to upper management and found that over the last couple of years the company was having difficulty meeting increasingly diverse customer needs. The president, who came from a Fortune 500 company, seemed to understand the problem best. One night, he noted his production planning people were working later than usual and he asked them why. They said that they had been working to develop a plan that would meet marketing's forecasts, but they were not able to do it, despite working on it for a couple of weeks! Over the last few years, the business had become very competitive and the company's product line had expanded to several hundred items, making planning much more difficult. The president explained that, while these planners had over 20 years of experience, the complexity of the environment had exceeded their ability to do the production plan on paper and spreadsheets using the guidelines and rules-of-thumb that they had developed over the years. The president stated he wanted the consultant to develop software to help them schedule better. He was familiar with this type of software from his experience at the Fortune 500 company. The consultant said, "Of course, you will want the software to give the planners the optimal lowest cost solution." The president stated: "This would be extremely desirable, but just make sure it gives them a production plan that meets our marketing forecasts, as well as our production and distribution needs. Right now, it is of paramount importance that we generate realistic plans that satisfy our customer demand."

Manufacturers Are Showing a Greater Interest in Optimization
The situation at ABC, Inc., described above, has been happening for many years in all sizes and types of companies throughout the manufacturing industry. Customer demand and competition have made supply chain planning and scheduling more challenging and complex. A number of major trends have contributed to this increasing complexity:

• Customer demand for shorter cycle times and specialized packaging/delivery requirements
• Mass customization of products
• Product line and stock keeping unit (SKU) proliferation
• Globalization of operations - including sourcing, production, sales and marketing
• Greater outsourcing of manufacturing operations
• Increased use of third party logistics (3PL) providers
• Implementation of co-managed inventory programs with both suppliers and customers, such as vendor managed inventory (VMI) and continuous replenishment programs (CRP)
• Implementation of agile manufacturing initiatives
• Implementation of supply chain integration concepts
• Company mergers, acquisitions, and consolidations

These trends are contributing to an explosion in the number of entities that have to be planned for, driven by increases in the number of the following elements:

• Items
• Production and distribution facilities
• Functions
• Customers and suppliers

For many years manufacturers have been moving toward improved use of technology to support complex, diverse planning processes. Some, such as ABC, Inc., are doing it largely to maintain control of their operations in order to meet customer demand. Having already achieved control, many manufacturers are using APS technology to increase the productivity of planning processes and to lower supply chain costs.

Generally, companies are looking for planning solutions that consider major supply constraints, which leads them to constraint-based optimization. Supply chain planning optimization techniques and solutions attempt to accomplish the following tasks:

• Determine a feasible plan that meets all demand needs and supply limitations
• Optimize the plan in relation to corporate goals such as low cost and profitability

While a feasible, realistic plan is of paramount importance, an optimized plan is better. It is the need for realistic, optimized plans that is driving many manufacturers away from classic materials requirements planning (MRP)-based planning solutions, which do not consider supply constraints (especially material constraints) and frequently generate an unrealistic supply plan.

Vendors Are Embedding Optimization in Their Planning Applications
Consistent with this corporate trend toward greater need for supply chain planning technology, the APS market is increasing dramatically. While this has happened over the last decade, only within the last two to three years has optimization been widely incorporated into APS suites. Examples include the following events:

• Slightly over one year ago, Manugistics (Rockville, MD) embedded various optimization solution methods into its integrated supply chain planning suite, especially its Supply Chain Navigator product.
• In 1997, i2 Technologies (Irving, TX) extended its optimization capabilities by purchasing the CSC Operations Planning Group (Austin, TX), which developed customized optimization solutions for the consumer packaged goods (CPG) market. i2 Technologies also purchased Optimax Systems, a pioneer in the use of genetic algorithms to optimize the scheduling of assembly lines. Lastly it introduced its Global Decision Support Architecture to lay the framework for enhanced optimization functionality across diverse environments.
• Recently, Logility (Atlanta, GA) announced plans to embed optimization software technology from INSIGHT, Inc. (Manassas, VA), a provider of supply chain optimization software for over 20 years.
• In 1997, SynQuest (Atlanta, GA), a production scheduling solution provider, acquired Bender Management Associates (Arlington, VA), which has a long history of customized supply chain optimization solutions. This purchase added to the optimization functionality acquired in 1996 from Log'In, a developer of simulation and optimization software.
• InterTrans Logistics Solutions (Toronto, ON), an i2 Technologies company and a provider of transportation management software, acquired Strategic Decisions, Inc., the developer of Supply Chain Strategist, a supply chain network design tool, to help its 3PL customers optimize their distribution networks.
• ILOG, Inc. (Mountain View, CA), a supplier of supply chain optimization software components to APS vendors, purchased CPLEX Optimization, Inc. (Incline Village, NV), a supplier of linear and mixed integer programming tools.

Enterprise resource planning (ERP) vendors have also noted the dramatic growth in the supply chain planning market and some have announced plans to add optimization functionality:

• PeopleSoft (Pleasanton, CA) leveraged its purchase of the Red Pepper APS optimization functionality to create what it calls Enterprise Resource Optimization (ERO).
• SAP (Waldorf, Germany) announced an initiative to develop supply chain planning functionality, which it calls Supply Chain Optimization, Planning and Execution (SCOPE). A key component of SCOPE will be Advanced Planner and Optimizer (APO), which will use optimization techniques.
• Baan (EDE, The Netherlands) acquired Berclain (Toronto, Ontario), a production planning and scheduling vendor, and has since developed Baan SYNC, which will include constraint-based planning and scheduling functionality.
• J.D. Edwards (Denver, CO) announced plans to embed optimization technology from ILOG into its future APS solution to be offered as part of its OneWorld ERP system.

Renewed Interest in a Mature Market
Despite the recent flurry created by the APS and ERP providers, it should be noted that supply chain planning optimization technology solutions are not new. There has been a market for optimization solutions for over 30 years. The market has slowly evolved from toolkit-based products to a packaged application market. Early adopters of optimization technology tended to be quantitative analysts, usually with degrees in operations research, who worked in the corporate world. Many worked in process industries such as Chemical, Paper, and Steel. These early adopters used general-purpose optimization tools (e.g., linear programming packages) purchased from software vendors to develop custom planning tools that typically ran in a batch mode. Early optimization tool vendors include the following companies:

• IBM (Armonk, NY), which provided the Optimization Solutions and Library (OSL) suite of optimization tools
• Lindo Systems, Inc. (Chicago, IL), which marketed a set of optimization tools, including one of the first spreadsheet- based user interfaces
• Ketron Management Sciences (Arlington, VA), which, in addition to operations research consulting services, sold optimization software

Few if any of the customized planning solutions developed by these early adopters dealt with large portions of a company's total supply chain. They usually focus on one important aspect of it. Some of the early applications dealt with specialized optimization problems:

• Blending--involves determining the best mix of raw materials required to form a product with specific characteristics. Blending optimization is important in some businesses where raw materials with different compositions can be obtained from multiple sources. For example, various crude oil supplies are blended to produce certain types of gasoline or oil products with specific viscosity grades. There is no fixed bill-of- material or recipe for a blended product. The planning problem is how to determine the optimal mix of constrained raw materials that are needed to produce specified amounts of finished product, often at a target or lowest price.
• Trim or Cutting Stock--involves the cutting, according to size--of semi-finished products to form a finished product. Examples of this type of planning occur in the Paper and Steel industries. Typically, large rolls of semi-finished products need to be cut to order. The planning problem is how to develop a cutting plan that satisfies customer orders while minimizing the amount of waste or "trim." Network Flow or Transportation-- involves shipping products from a number of origins to a number of destinations. This was one of the earliest approaches for supply chain design or tactical planning, used to determine inbound or outbound shipments. The objective is to develop a shipping plan that minimizes transportation costs between origins and destinations.

As this market progressed, a few early supply chain planning vendors started to sell general-purpose optimization applications. These applications made it easier for corporate users to develop supply chain planning solutions on their own or working with the vendor's consultants. Two such vendors are Chesapeake Decision Sciences (New Providence, NJ) and Numetrix (Toronto, ON). As general-purpose optimization applications, these types of solutions allowed users not only to model specialized planning problems like trim, blending, and network flow, but they also allowed users to model more general planning problems, such as combining blending with production scheduling.

Despite some early success in the use of optimization, the market was relatively stagnant until recently. Advances in powerful computer technology have helped to accelerate the growth of the APS market. The technology has also allowed APS vendors to embed optimization into their solutions more seamlessly and transparently. This has made it easier for users to model their planning environment, even those users not trained in optimization techniques.

Today there are many popular APS solutions with embedded optimization. Despite this, the optimization aspect of the market carries a certain amount of mystery. Optimization is difficult to understand because of the jargon used by practitioners on both the user and vendor side. To many non-practitioners this is a very confusing, but seemingly intriguing and important area. Our short "Primer on Optimization" should clear up some of the confusion. It details the concepts and language behind optimization techniques and methods, and provides the context for the remainder of this paper.

Supply Chain Planning Optimization Framework
A supply chain planning environment can be described in terms of the supply chain's structure and the level of planning being supported. In general, the following holds true:

• A supply chain's structure is described using a network model.
• The planning processes are divided into three hierarchical-based levels: strategic, tactical, and operational.

Figure 2.	Network Representation of Manufacturer's Supply Chain

Planning Processes Have Three Hierarchical-Based Levels
Planning processes are typically subdivided into multiple hierarchical-based planning levels. Each level has a planning cycle that its processes follow. Currently, supply chain planning is usually done using three hierarchical planning levels:

• Strategic or high-level planning is typically done yearly or on an ad hoc basis.
• Tactical or mid-level planning is typically done quarterly or monthly.
• Operational or low-level planning--largely involving scheduling, rescheduling, and execution--is typically done weekly, daily, or by shift.

Figure 3.	The Scope of APS processes in relation to the three supply chain planning levels

Figure 3 depicts the scope of APS processes in relation to these planning levels. A description of each supply chain planning level and how optimization is used within it follows.

Strategic Level Planning--Supply Chain Network Design
To support supply chain design, optimization determines the location, size, and the number of plants, distribution centers, and suppliers. This level of planning includes sourcing and deployment plans for each plant, each distribution center, and each customer. It also considers the flow of goods through the supply chain network. Generally, supply chain network design is done infrequently (i.e., every few years) as companies do not need to add new plants or distribution centers on a routine basis.

Conceptually, supply chain network design can be thought of as determining the nodes and arcs within a supply chain network. Multistage production processes are usually handled by creating nodes for each major step in the process (to handle cases such as feeder plants). The concept of time is usually not a consideration in supply chain network design. Month-to-month or week-to-week changes--for example in demand--are typically inconsequential to supply chain network design decisions.

Supply chain network design products include the following:

• Supply Chain Strategist from InterTrans/i2 Technologies
• Supply Chain Designer from CAPS Logistics (Atlanta, GA)
• SAILS and the Global Supply Chain Model from Insight
• Network Designer from SynQuest

Tactical Level Planning--Supply Planning
Supply chain planning at a tactical level is called supply planning and involves optimizing the flow of goods throughout a given supply chain configuration over a time horizon. Similar to supply chain design, supply planning develops sourcing, production, deployment, and distribution plans. But there are some major differences:

• The supply chain network configuration is already in place with supply entities such as suppliers, plants, distribution centers, and transportation lanes.
• Supply plans are time-dependent--using a "time buckets" concept. (Generally, supply planning is done monthly or weekly.)
• Supply plans may consider aggregate views of multistage production processes by incorporating partial levels of a plant's routings and a product's bill-of- materials.
• Setup and changeover times may also be considered, but not the sequencing of orders through a manufacturing facility.

Some manufacturers and software vendors subdivide the supply planning process into two or more planning processes and hierarchies. For example, a first level of planning might involve the development of material and product flows within the supply chain network. This establishes optimized sourcing plans, which determine the suppliers for each plant and the plants/distribution centers that will replenish the distribution centers and customer stocking locations. This planning is typically done using aggregated product groups over longer periods of time (quarters or months). The second, lower level tactical planning might involve more detailed, optimized manufacturing master plans and distribution center inventory replenishment plans for individual SKUs over shorter periods of time (months or weeks).

Operational Level Planning--Production Scheduling
At an operational level, supply chain planning can be viewed as supply scheduling. For a manufacturer, supply scheduling is essentially production scheduling done on a plant-by-plant basis. Production scheduling develops a minute-to-minute or hour-to-hour schedule for all of a plant's resource needs, including labor, equipment, and materials. Generally, production scheduling is done frequently, potentially several times a day to account for changes to orders, machine failures, material shortages, and other plant disruptions. The process usually considers the lowest level of detail on plant routings, product bills-of-material, and changeover and setup times.

Solvers Are Key Elements Of Optimization
Models and data are two very important elements of supply chain optimization. A plan's usefulness greatly depends on the quality of both. If a planning process is based on a model that inadequately represents reality or if the data used in a model is wrong, the solutions developed by the optimization will not be meaningful or executable. The two elements go hand-in-hand for the following reasons:

• Data is always needed to populate a model.
• A model should not be built needing data that is not available.

Another important element of an optimization process is the solver, which solves for an optimized solution. The planning model determines the best solver. (See the sidebar "Optimization Solvers"). Together, data, models, and solvers represent key elements of optimization-based planning.

Good Quality Models Are Needed
A supply chain planning optimization process requires "good quality" models. For good optimization, the level of detail in the model must be appropriate for the planning level. For example, in deciding how much to make in a plant this month, the model does not have to consider the exact sequencing of orders (actual or forecast) placed on the plant. On the other hand, to decide on the plant schedule for a specific day, the model should consider the order sequence.

While models and data are extremely important to the usefulness of a plan developed using optimization, realistic, executable plans can sometimes be developed without detailed models and data. A common guideline in supply chain planning optimization is that detailed models are not needed for strategic planning but are needed for operational and tactical planning. For strategic and higher levels of tactical planning, a model could be based on aggregate data such as product groupings or regional demand. It is important for an operational model to be based on detailed data such as SKUs and customer orders.

Solvers Depend on Models Used
In addition to models and data, solvers are important since they generate the optimized solutions. Yet one of the most confusing aspects of optimization technology is the various methods or mathematical algorithms deployed in these solvers. Many of these methods were developed to solve problems with a specific model or mathematical structure, while others were developed to improve the computational speed of the solver. This led to esoteric names for these solvers and added to their "rocket-science" positioning. In the area of supply chain optimization, most methods can be grouped as one of the following types:

• Mathematical programming (largely linear and mixed integer programming)
• Heuristics (including scheduling methods like the Theory of Constraints or Simulated Annealing)
• Genetic algorithms
• Exhaustive enumeration

Generally, mathematical programming methods are used in solvers for strategic and higher levels of tactical planning. These methods generally work only for solving linear- and some integer-based models, commonly used in strategic levels of planning. Tactical and operational models are usually not linear and are much too complex to solve using mathematical programming methods. For this reason, heuristic methods are generally used in tactical and operational planning level solvers.

Genetic algorithms are used primarily in operational planning to consider a large number of possible solutions. The Theory of Constraints, a heuristic method based on work by Eli Goldratt, is another solver method commonly used in operational planning. Vendors that use solvers based on the Theory of Constraints include the following:

• i2 Technologies
• STG
• Thru-Put Technologies

While not a formal optimization technique, exhaustive enumeration is predicated on using the computer to find a solution by looking at all possible alternative plans. This method proves useful in simple supply chain situations. Otherwise, this method is computationally intensive and slow to generate a solution. Distinction Software (Atlanta, GA), uses this optimization method for its manufacturing planning solution. Since the company focuses on mid-tier and smaller manufacturers, the exhaustive enumeration approach is feasible.

Vendors Are Moving Toward More Optimization
Generally, optimizing each piece of a plan in isolation from other plans does not guarantee that optimization is achieved for the total planning process. For example, developing an optimal production plan in isolation from distribution does not guarantee that the total production/distribution plan is optimal. Vendors are addressing this concern in the following ways:

• Moving toward synchronized concur rent planning from synchronized sequential planning
• Providing functionality that helps to synchronize planning levels
• Moving toward real-time planning and execution

Figure 4.	Synchronized Sequential Planning

Each of these vendor trends attempts to provide users with a more holistic approach toward optimization. They each address planning processes that may be somewhat subdivided but that support optimized plans across the entire supply chain. Each trend is discussed below.

A Move Toward Synchronized Concurrent Planning
Figure 4 is a graphical representation of a synchronized sequential supply chain planning process. In this process, demand planning is done first. The resulting forecasts are used as input to the distribution planning process (including planning for inventory, transportation, and warehousing). The distribution plans developed by this process are then used as input to the manufacturing planning process. Lastly, the output of this process is used as input for the procurement planning process. The planning process is synchronized, since any change in "downstream" plans is automatically reflected in the "upstream" planning processes. From a product perspective, Manugistics' supply chain planning suite was the first major solution providing this type of synchronized sequential planning approach.

The synchronized sequential approach works well in integrating supply chain planning and in moving it to a consumer demand "pull" environment. It has, however, two problems from an optimization perspective:

• Each planning component's resource constraints are typically not considered, nor is there any attempt to optimize an objective.
• Separate optimization of each planning process rarely produces an optimized plan in the context of the whole supply chain.

Figure 5.	Synchronized Concurrent Planning

Some optimization processes solve these two problems through "synchronized concurrent" planning (See Figure 5). In synchronized concurrent planning, the demand, distribution, manufacturing, and procurement plans are jointly or simultaneously developed. All constraints along the supply chain and optimizing objectives, such as cost or profitability, are considered within the planning process. Many vendor applications have moved or are moving toward this planning approach. For example, Manugistics' Supply Chain Navigator provides synchronized concurrent planning.

Planning Levels Are Being Synchronized To Ensure Optimality
In most optimization frameworks, optimal solutions generated at higher levels in a planning hierarchy are used as starting points or constraints on the optimization taking place at lower levels. In a planning environment where optimization is a goal, plans at different levels may need to be synchronized to ensure that the supply chain is continually operating in an optimized fashion.

Figure 6.	The difference between architectures that do and do not have synchronization functionality

Figure 6 depicts the difference between architectures that do and those that do not have synchronization functionality. Disjointed plans and data levels do not ensure that optimization is achieved. To ensure synchronization and consistent optimization across hierarchical planning levels, vendors use several approaches:

• The use of telescoping planning horizons
• The use of a common data structure for all planning levels
• The monitoring and control of the degree of synchronization

In a telescoping planning horizon, "time buckets" vary over time (typically increasing):

• The first few weeks may be planned in continuous time, like minutes or hours.
• The next few months might be planned in weeks.
• The last periods of the planning horizon might be represented in months or quarters.

This approach helps ensure synchronization along the time dimension for the different planning levels. For example, Thru-Put Technologies' Resonance product allows the user to define a telescoping planning horizon for viewing results. Other vendors with telescoping planning horizons include Chesapeake and Numetrix.

Vendors also address the synchronization issue by using a common data structure for all planning levels. This common data structure ensures that aggregated data is always derived from and synchronized with detailed data--possibly down to the lowest level of base data (e.g., ERP transaction data). If needed, these solutions also allow a planner to build models using detailed data. For example, if a lower-level work task or portion of a plant's routing is always a major bottleneck, a user can include it in higher level plans to ensure that the solution is optimal and feasible. For example, Paragon Management Systems uses a common data structure approach within its applications; Ortems has a data structure with aggregated data derived from detailed data.

Some vendors address the synchronization issue by incorporating functionality to monitor and control the degree of synchronization among planning levels. As a planner works with the system, if a lower-level plan gets significantly out of synch with higher-level plans, the higher-level plans are regenerated. Paragon Management Systems has designed its APS product suite to ensure synchronization among planning levels in this way. Numetrix's Collaborative Enterprise Network product takes a slightly different approach. The application monitors the degree of synchronization among plans and sends a planner an alert message when they are out of synch.

Real-Time Planning and Execution Is a Goal
As computers get more powerful, companies are shrinking planning cycles and achieving real-time planning and execution. In turn, this moves them toward achieving real-time optimization, as well as reducing the need for hierarchical planning processes with long cycle and lead times. A reduction in these planning cycles would lead to the following improvements:

• Reduced supply chain inventories
• More responsive supply chain operations with improved customer service

Up to now, reducing planning cycles with the use of applications has been limited by the speed at which an optimized plan can be generated. While some manufacturers have moved from planning in a batch mode, many have not. Applications are getting closer to generating optimized plans in real-time.

To do this, some vendors, such as i2 Technologies and Ortems, use memory-resident planning applications loaded into and executed in main memory. This speeds up the application substantially. Other vendors, such as Manugistics, have designed applications to use both memory and hard disk space. Generally, both approaches achieve the goal of faster plan generation. Neither approach works better in all planning environments. For example, memory-resident applications are faster as long as the planning problem can fit into memory, but they quickly degrade in performance on large problems that also require hard disk space.

Application Architectures Differ Substantially
A variety of design issues must be considered to develop the architecture or framework of a supply chain planning optimization application:

• The balance between modeling flexibility and built-in functionality
• The ability for the user to maintain control of optimized solutions
• The use of third party optimization components
• Special functionality for handling multiple objectives
• Functionality for supporting special optimization problems

Vendors have addressed these issues in a variety of ways, many times depending on a specific vendor's history and market focus.

Applications Need To Balance Flexibility and Built-In Functionality
Optimization depends heavily on the ability of the application to model real world issues. This can be done with built-in modeling functionality, such as predefined models for setting safety stocks. Rather than built-in functionality, an application may provide flexibility with general-purpose modeling capability that allows users to create models--such as defining safety stocks models. Applications need to provide users with a balance between built-in capability and flexibility.

Some vendors provide general-purpose modeling applications that allow users the flexibility to optimize across a broad range of decisions. These products allow users to tailor the optimization to their own environment, letting them optimize decisions uniquely important to them. MIMI Planning and Scheduling is a general-purpose modeling application from Chesapeake Decision Sciences, now owned by Aspen Technology. Other companies providing these types of applications include CAPS Logistics, i2 Technologies, and Numetrix.

To provide some level of built-in functionality, many of the general-purpose modeling application vendors are developing or offering templates. These are semi-custom applications, created using their general-purpose modeling applications. Templates reduce implementation configuration efforts. For example, i2 Technologies is building templates that are specific to industries such as Semiconductor, High Tech, and Metals. In addition, CAPS Logistics now markets Supply Chain Designer and Supply Chain Coordinator, strategic and tactical supply chain planning applications built with its toolkit.

Other vendors offer built-in capabilities rather than general-purpose modeling application functionality. While these applications limit the ability to optimize unique decision environments, they minimize the implementation configuration efforts for specific supply chain problems. For example, Manugistics has recently added optimization capability to its product suite tailored for consumer products and distribution-intensive companies. While this provides less flexibility to optimize, it reduces configuration efforts. Logility is developing similar optimization capability.

Users Need To Maintain Control of the Solutions
Generally, optimization solutions must be understood and controlled by a user. Optimized solutions are frequently not reasonable or executable and need to be adjusted by the planner. Remember, these are decision-support, not decision-making systems! In fact, one APS vendor claims that around 40% of software development efforts are spent solely on enabling user control functionality.

The major purpose of user control is the inability to enter all details into the optimization model. For example, an optimized solution may suggest that a customer's due date be pushed out because it is critical in achieving a low-cost solution. The manufacturer, however, may have already pushed out the due date for the customer's last three orders. This would be an unacceptable solution!

To ensure that optimization applications provide reasonable, executable solutions, most vendors provide graphical user interfaces (GUIs) to facilitate manipulating data and modifying solutions. This functionality includes the following:

• Graphical drag-and-drop planning boards
• User-defined constraints and rules
• Control of the optimization procedure

Many graphical planning boards allow users to change a variable (e.g., a date or order) and immediately see the impact on the objectives and assess new constraint violations. Vendors also give users control over the solution by allowing them to incorporate unique constraints or rules into the model. For example, a planning application might allow a user to specify that customer due dates can be relaxed by one or two days, or the user may approve the maximum level of overtime. While this is a useful functionality, some users are tempted to abuse it by changing the objective. This is conceptually an invalid approach to an optimization problem. Solutions that are not reasonable are best changed by altering the constraints.

Some vendors allow users to control the progress and performance of the solver method. This is usually done by allowing the planner to set a time limit on the solver and then pause. This lets the planner evaluate the solution so far. If it is good enough, the user can finalize the plan. If it is not, the user can have the solver search longer for a better plan.

Many Vendors Use a Mix of Third Party and Internally Developed Optimization Components
Embedded within vendor solvers are components, submodules that perform the underlying optimization logic. Some vendors use optimization components from third party vendors to shorten development time. This approach also allows the vendor to focus ongoing development resources on the planning application while the third party takes on the responsibility for improving the optimization component.

The leading optimization component vendor is ILOG. The company offers three optimization components including market-leading LP/MIP acquired from CPLEX. The CPLEX solution is embedded in many applications including those from the following vendors:

• Manugistics
• i2 Technologies
• CAPS Logistics
• SynQuest

ILOG's other optimization components are general-purpose planning and scheduling components called ILOG Planner and ILOG Scheduler. Fygir has embedded the ILOG Planner within its FIT planning and GRIP scheduling products.

Sunset Software Technology (San Marino, CA) is another component vendor. Adapta Solutions (Hawthorne, NY), has embedded this company's LP/MIPS functionality into its planning suite.

Many application vendors exclusively offer optimization components that are internally developed, as they believe that they have a core competency in optimization, especially for planning environments in their target markets. For example, Ortems uses proprietary heuristics based on "graph theory" in its SRP product, which focuses on production scheduling and the lower levels of tactical planning. Similar to Ortems, Thru-Put uses its own proprietary methods based on the Theory of Constraints. The company believes that these methods provide greater optimization benefit for the complex manufacturing environments it targets (such as Automotive and Industrial Product manufacturers).

Most application vendors use both internally developed and third party optimization components. These vendors usually have a broad suite of supply chain planning products. For example, i2 Technologies offers a full suite of optimization products that use internally developed components based on proprietary heuristics (including those based on Constraint Anchored Optimization and Simulated Annealing methods), genetic algorithms, and third-party LP/MIP components.

Some Vendors Provide Multi-Objective Functionality
Some vendors allow planners to define multiple objectives. For example, both customer service and costs can be optimized together--despite the fact that two or more objectives cannot be optimized at the same time. Maximizing customer service is not likely to minimize costs; quite the opposite usually holds true. To accommodate planners that want to strike a balance among competing objectives, some vendors provide solvers that find an optimized solution using the objectives in priority sequence. i2 Technologies offers this type of solver using a "hierarchical" LP/MIP algorithm.

Another way in which applications allow planners to use multiple objectives is through the use of weights. This is accomplished by creating a single objective that is a combination of the different objectives, weighted by a number. For example, a customer service objective might carry a weight of 90 (out of a possible 100), while a cost objective might carry a weight of 70. The solvers in these applications optimize a single, composite weighted objective. ProMira, recently acquired by Manugistics, provides this functionality, incorporating a graphical slidebar interface to make it easy for a planner to set the weights.

Some Applications Handle Special Optimization Problems
Some special planning environments involving trim and blending optimizations cannot be accommodated by many supply chain planning applications. Some vendors have incorporated such capability to be integrated into their supply chain planning suites.

For example, integration of trim optimization in the paper industry can lead to substantial benefits by reducing waste paper from cutting operations. Chesapeake and IBM both have special trim optimization functionality to support paper manufacturers as part of a total supply chain planning optimization application.

Blending optimization is used in the fruit beverage and oil industries, where raw materials are limited and prices are volatile. Materials must be efficiently blended to create products that meet specifications. Fygir, Chesapeake, and STG have products that can incorporate blending optimization into the supply chain planning application.

Optimization Usage Guidelines
While the use of optimization in supply chain planning is extremely appealing, the technology may not be for everyone. Rather than by employee headcount reduction alone, these solutions are justified by such supply chain benefits as the following:

• Inventory reduction
• Improved customer service
• Increased asset utilization
• Improved corporate profitability

Though there are no hard-and-fast rules for manufacturers deciding whether to purchase optimization technology, AMR offers the following guidelines:

• Optimization is generally beneficial in complex manufacturing environments where many interrelated decisions need to be made. These include environments with many resource constraints and large numbers of products, plants, suppliers, and distribution centers. Planners in these environments need computer support to make optimized decisions. In contrast, planners may not need optimization support in simple, mature environments, where methods based on experience may already yield nearly optimal decisions.
• In strategic and higher-level tactical planning, the pursuit of optimized solutions is typically more important than it is for low-level tactical and operational planning. In the former, the feasible set for decisions is much larger, meaning there are more opportunities to make poor decisions. Also, these decisions have greater revenue and cost implications.
• The answer to "Where is the most pain in my supply chain?" will be important in deciding what portion to optimize. In supply-constrained industries that experience material shortages, optimizing the use of these materials in the manufacturing process is important. In make-to-order environments, especially in discrete manufacturing, optimized production schedules are crucial. For distribution-intensive environments, planning must focus on optimizing manufacturing and distribution operations simultaneously.
• Optimization is more useful in mature, relatively non-volatile manufacturing industries where product demand and manufacturing processes are more predictable. In these planning environments, realistic models can be constructed to support all levels of planning. In volatile manufacturing environments, optimization will be less useful for strategic and tactical planning. In these environments, planning focuses on supply chain readiness and responsiveness rather than on operational efficiency. Optimization will be more useful for operational planning, when the level of uncertainty is substantially reduced (e.g., when many customer orders are already placed).

CONCLUSION
The guidelines above will hopefully reduce some of the mystery surrounding optimization, and they should help buyers during their technology selection and implementation processes. The research for this paper has led to the following conclusions:

• While the term optimization is confusing to buyers, the APS market is getting a lot of attention. This is primarily driven by the increasing complexity of manufacturers' supply chains. This complexity is caused both by the trend toward globalization and the myriad products, materials, facilities, trading partners, and trading relationships that need to be planned. In many companies, planners are becoming overwhelmed by the complexity, and they need better computer support in decision-making.
• The ability to model a manufacturer's environment adequately is a critical success factor in being able to use computer-generated optimized decisions. The availability of data to input into one's supply chain models is closely related to this factor. Thus, modeling capabilities and data requirements are the key factors to consider in implementing optimization solutions.
• One of the most confusing aspects of the optimization market is the variety of solver methods marketed with names like mathematical programming, heuristics, Theory of Constraints, Simulated Annealing, and genetic algorithm. Generally, vendors employ solvers depending upon the structure of the optimization model. By and large, vendors are diligent (some even fervent!) in using appropriate methods within their solvers-- whether the methods are proprietary, purchased from another vendor, or based on known methods. This makes solver type a secondary consideration in choosing among different optimization- based applications.
• Vendors design their optimization applications in various ways. This makes comparing them difficult and requires taking a hard look at each vendor's solution.

To summarize, optimization technology offers great promise for improving a manufacturer's supply chain performance. But understanding it will take a great deal of work and great effort on the part of users. AMR expects, however, that its use in supply chain planning will continue to grow, since the benefits should far outweigh the efforts.

What is an Optimization Problem?

Generally, optimization problems seek a solution where decisions need to be made in a constrained or limited resource environment. Most supply chain optimization problems require matching demand and supply when one, the other, or both may be limited. By and large, the most important limited resource is the time needed to procure, make, or deliver something. Since the rate of procurement, production, distribution, and transportation resources is limited, demand cannot be instantaneously satisfied. It always takes some amount of time to satisfy demand, and this may not be quick enough unless supply is developed well in advance of demand. In addition to time, other resources, such as warehouse storage space or a truck’s capacity, may be constrained in meeting demand. An optimization problem comprises four major components:

Decision Variables are within the planner’s span of control:

• When and how much of a raw material to order from a supplier
• When to manufacture an order
• When and how much of the product to ship to a customer or distribution center

• A supplier’s capacity to produce raw materials or components
• A production line that can only run for a specified number of hours per day and a worker that must only work so much overtime
• A customer’s or distribution center’s capacity to handle and process receipts

Constraints in an optimization problem are either hard or soft. Hard constraints, such as the number of working hours in a shift or the maximum capacity of a truck, must be adhered to or satisfied. Soft constraints can be relaxed or violated. Examples of soft constraints include customer due dates or warehouse space limitations. Customer due dates can be changed or a product may be squeezed into a warehouse temporarily, making constraints less stringent. Most optimization problem formulations designate cost penalties if a soft constraint is not met. The penalties allow constraints to be weighted by importance. For example, missing a customer due date is a more important concern than cluttering a warehouse aisle.

Objectives

Objectives maximize, minimize, or satisfy something, such as the following:

• Maximizing profits or margins
• Minimizing supply chain costs or cycle times
• Maximizing customer service
• Minimizing lateness
• Maximizing production throughput
• Satisfying all customer demand

Those unfamiliar with optimization are often confused about the difference between a constraint and an objective. Fueling this confusion, some factors can be formulated as either an objective or a constraint. For example, in most problems customer due dates are hard constraints. On the other hand, in resource constrained environments, while meeting customer due dates is important, it may not be possible. Therefore, the objective to meet customer due dates is expressed by maximizing customer service.

Models

Models describe the relationships among decisions, constraints, and objectives. These are often expressed in the form of mathematical equations. This is probably the most important but least understood part of an optimization problem. Generally, the model must represent the "real world" to the degree needed to capture the essence of the problem. It must represent the important aspects of the supply chain in order to provide a useful solution. For example, strategic planning typically uses aggregate models, which do not include every factor. On the other hand, operational planning uses models that include almost all factors and require detailed data.

• Feasible Solution - satisfies all the constraints of the problem.
• Optimum Solution - the best feasible solution that achieves the objective of the optimization problem. Although some problems may yield more than one feasible solution, there is usually only one optimum.
• Optimized Solution - a solution that partially achieves the objective of the optimization problem. It is not the optimum or best solution, but it is a satisfying or reasonable one. This is usually one of the best feasible solutions. However, for optimization problems that have no feasible solutions, it may be one of the best infeasible solutions. For example, in a resource-constrained environment, it may be a solution that is infeasible because it does not meet all customer due dates, but it may minimize operating costs.

Figure 1.	Current State

Figure 1 represents an optimization problem with a generalized set of objectives to maximize. It depicts the different types of solutions that might be developed by a solver. While an optimum solution is intuitively appealing in most cases, it is unattainable in very complex problems. Unless a problem has a very specific structure (such as a linear programming problem), an optimized rather than an optimum solution is the best that can be generated. In some cases a solution may be a local optimized one (see Figure 1).

About The Authors
Jim Shepherd
Vice President, Research
AMR Research, Inc.

Jim Shepherd is the vice president of research, leading coverage on topics including Enterprise Resource Planning (ERP), Manufacturing Execution Systems (MES) and Supply Chain Management (SCM). Jim is also an Editorial Board Member of Achieving Supply Chain Excellence through Technology.

Jim has over 20 years of experience in manufacturing operations and software industry management. Prior to joining AMR Research, Jim was the director of product strategy at Avalon Software, a client/server business systems developer for Fortune 1000 manufacturing companies. His previous experience includes three years as vice president of business development and product strategy at Enterprise Planning Systems, and eleven years with ASK Computer Systems, most recently as director of advanced manufacturing systems. His background also includes several years as vice president of manufacturing at Conrac, a multinational manufacturer of electronics and telecommunications equipment.

Jim speaks frequently at industry trade conferences, user group meetings, and executive conferences. He is widely quoted in leading business and industry trade publications.

Jim has an undergraduate degree from Wesleyan University.
You may contact Jim at jshepherd@amrresearch.com.

Larry Lapide, Ph.D.
Vice President & Service Director, Supply Chain Strategies
AMR Research, Inc.

As Vice President and Service Director, Supply Chain Planning, Larry Lapide is responsible for business practices and decision-support technology required to achieve efficient supply-chain operations. Larry has over twenty-five years of business experience including over twelve years in the consulting industry and ten years in the high-tech industry. He is a leading authority in the application of business practices and decision-support technology enabling clients to achieve efficient supply chain operations. His primary areas of expertise are in demand planning and forecasting, inventory and distribution planning, and production planning and scheduling.

Larry joined AMR Research from Benchmarking Partners, where he was director, value chain strategy. Prior to that, he was an associate partner in Accenture's Logistic Strategy Practice, where he served manufacturing and retailing clients across a diverse set of industry segments, including High Tech, Telecommunications, and Consumer and Industrial Products. Larry was also responsible for managing Andersen's logistic training programs and for tracking logistics technology.

Furthermore, Larry has authored and co-authored various papers, and has presented to a variety of professional groups. For over ten years, he was a part-time lecturer at the University of Massachusetts where he taught statistics, forecasting, and operations management.

Larry holds a B.S. and master's degree in electrical engineering from The Cooper Union and M.I.T. respectively, and holds a Ph.D. in Operations Research from the University of Pennsylvania's Wharton School of Business.

You may contact Larry at llapide@amrresearch.com.