THE AMR INSIDER
Perspectives on Optimization by Janet Suleski
A number of business groups are intimately involved in the creation and execution of supply chain optimization (SCO) technology. Each group, however, has a different perspective on the following issues:
To get their perspectives on supply chain optimization, AMR interviewed two representatives from each of the following communities:
AMR asked this prestigious group to respond to eight questions concerning supply chain optimization. Their responses are summarized below.
WHAT IS YOUR DEFINITION OF SUPPLY CHAIN OPTIMIZATION? (BACK TO TOP)
The consultants' definition of SCO focused on the following factors:
Interestingly, both consultants felt the definition of best "mix" should be left to the manufacturers. One vendor's definition, however, stated that SCO is a tool meant to increase the client's return on assets, suggesting that the best solutions are those that maximize that particular metric. The professors agreed that SCO covers both operational and strategic decision-making. One professor proposed another category, one that cuts across the operational and strategic levels: decision-making about the location and quantity of inventories.
The manufacturers had opposing viewpoints on what constitutes the scope of SCO. One indicated that SCO is the application of formal optimization methods (meaning linear or integer programming, excluding heuristics) to a model, which covers a reasonably comprehensive scope of a company's internal and external supply chain. The other manufacturer stated that SCO is the ability to look across only the internal supply chain of a company and determine the optimal way to produce and distribute products to meet demand.
WHEN IS SUPPLY CHAIN OPTIMIZATION MOST USEFUL? (BACK TO TOP)
There was a consensus from all parties, except the vendors, that SCO is most useful in situations where a company or a product has a complex supply base, a complex manufacturing process, a complex distribution system, and volatile demand. Essentially, whenever there is uncertainty in the behavior of supply chain operations or in market demand, SCO could benefit a company. It was felt that SCO would not be as useful--it may even be unnecessary--in a make-to-order environment with few supply points and a simple supply chain. The vendors, however, felt optimization was useful to all companies, regardless of the makeup of the supply chain.
One of the manufacturers described two conditions necessary for the implementation of useful SCO initiatives:
The manufacturer felt that if a company could not meet these two conditions, any effort at supply chain optimization would be unsuccessful.
FOR WHAT TYPES OF DECISIONS IS SUPPLY CHAIN OPTIMIZATION MOST HELPFUL? (BACK TO TOP)
One of the professors described SCO software as a tool to help people in various functional areas of a company broaden their focus to encompass the entire supply chain, from raw materials to the final customer. With this "supply chain perspective," the functional teams would have a common language to make better decisions about optimizing the flow of products through the supply chain.
In contrast, the consultants and manufacturers focused on the specific decisions for which they would use SCO:
The vendors emphasized that SCO could help to identify supply chain problems such as bottlenecks or unreliable suppliers. The SCO process directs planners to supply chain operation problem areas. While the software could suggest improvements, decisions about changes should remain in the planners' hands.
IN WHAT INDUSTRIES IS SUPPLY CHAIN OPTIMIZATION MOST USEFUL? (BACK TO TOP)
The participants put forth a number of possible answers to this question. One consultant and one vendor replied that there are three general types of industries where SCO is useful:
Each type of industry benefits from SCO, but each should optimize different things. For distribution-intensive industries, transportation optimization and inventory deployment were regarded as the two areas where SCO would be most effective. In the asset-intensive industries, optimizing throughput times, product mix, and setups were identified as important. In material-intensive industries, decisions about what to produce, where to produce it, and who to source materials from in order to generate high margins were noted.
In general, one consultant felt that process industries would find SCO more useful than discrete industries. In contrast, a vendor pointed out that this is a myth, born out of the early adoption of SCO by process manufacturers. The vendor also pointed out that techniques for tackling some of the more complex, discrete industry optimization problems have only recently become available. Both process and discrete manufacturers will find SCO useful.
Turning the question around, the other vendor felt that some companies would not benefit from SCO. These companies include those with level demand, predictable supply, and predictable competitive environments. The vendor also noted, however, that he didn't believe that such stable environments exist in any industry today.
WHAT SOLUTION METHODS ARE MOST USEFUL? (BACK TO TOP)
The perspectives on solution methods can be summed up with a quote from one of the vendors: "Different problems require different solutions." There are some general categories, however, where certain solution methods are predominant. The categories fell along the lines of the decision-making hierarchy:
The consensus among the participants was that at the strategic level, optimization most frequently employs mathematical programming, especially linear and integer programming. At the tactical level, combinations of linear programming, heuristics, simulation, genetic algorithms, and other methods are most frequently used. At the operational level, heuristics is most common.
The participants, however, also agreed that the method employed to solve any particular problem was much less important than understanding and properly framing the business problem. When the problem is understood, the granularity of data needed to solve it and the ease of acquiring and using this data play a role in determining which method will be the most useful.
HOW GOOD OR CLEAN MUST THE DATA BE TO PRODUCE USEFUL RESULTS? (BACK TO TOP)
This question provoked a range of answers from "very clean" to "use whatever data you have." In part, the answers varied because participants had various views on what constitutes a useful result. At one extreme, a consultant stated that data must be very clean because many optimization projects fail because of bad data. In failed projects, implementation could be successful, but the results coming out of the system are not useful. One professor's view was that the data that the industry deems most important must be very clean. Data relating to less critical parts of the business must be good, but less precision is required.
In contrast, the vendors felt that companies should not "exhaustively comb through data files to get 98% purity." By using whatever data is available, the company may be able to make better, if not completely optimal decisions. In addition, the process of optimization helps companies pinpoint what data needs to be improved first. The vendors felt that planners would detect and correct a faulty plan quicker and easier using optimization software. The gist of the vendor perspective is worry about your data, but don't make that an excuse for not optimizing. They also felt that "data gets better the more you exercise it."
The manufacturers we interviewed felt that the data put into an optimization system does not need to be "squeaky-clean." It should be 90% clean and perhaps, more importantly, representative of the problem the user is trying to solve. One manufacturer stated that problems are generally not caused by dirty data. They are more often caused by "data that measures something slightly different from what you really need in your model." For example, plugging the cost of port-to-port transportation into a model requiring the cost of transporting a good from the plant's dock to the retailer's door will result in poorly optimized transportation costs.
AMR's research has shown the accuracy and timeliness of data is important to both conventional materials requirements planning (MRP) and SCO. Many manufacturers reported struggling to cleanse data in preparation for SCO.
WHEN CAN ONE USE AGGREGATED DATA AND WHEN IS DETAILED DATA IMPORTANT? (BACK TO TOP)
Most survey participants believe that aggregated data is appropriate for optimizing at the strategic level and detailed data is appropriate for the operational (day-to-day, assembly-line) level. The vendors, however, emphasized that it depends on the planning problem. The manufacturers believe that there is an art to finding the right level of aggregation. One consultant felt that detailed data is important for key business drivers (e.g., transportation data in a distribution-intensive industry), but that other data could be aggregated. A vendor expressed the feeling, based on his experiences, that neophytes in the optimization business often use detailed information where aggregated data is appropriate. He also believes that companies should spend more time both before and during implementation to determine the right level of aggregation. Companies should also be prepared to revise levels of aggregation as the supply chain modeling process advances.
WHAT TRENDS DO YOU SEE IN THE USE OF SUPPLY CHAIN OPTIMIZATION? (BACK TO TOP)
The different communities see a variety of trends shaping the use of SCO over the next few years. The consultants reported the following trends:
The professors reported the following trends:
Vendors see the following trends:
Manufacturers see the following trends:
THE FUTURE OF OPTIMIZATION (BACK TO TOP)
Regarding the creation and execution of SCO technology, several consistent themes emerged from our conversations with the different constituencies:
While the participants in our survey had varying opinions on the aspects of supply chain optimization, one thing is clear: As a business practice, supply chain optimization is here to stay.
Copyright © 1998 AMR Research, Inc. All rights reserved. Published in the USA. Contents may not be reproduced in whole or in part without the written permission of the publisher. All information is the proprietary product of AMR Research. AMR Research retains ownership of the copyright to its materials.