The Ten Rules were originally presented at the annual conference of the Council of Logistics Management (CLM) last year by Dr. Ratliff, who is also Executive Director of The Logistics Institute (TLI) at the Georgia Institute of Technology. Due to extensive interest in the topic, he recently expanded his presentation into a white paper for the logistics industry.

    "Based on the enthusiastic response we received, we felt we had some specific logistics insights we could share with the industry," said Dr. Ratliff. "Logistics optimization is neither easy nor cheap, but it is currently the biggest opportunity for most companies to significantly reduce costs. This is especially true in optimizing transportation operations, where companies can realize 10-40% cost savings. This white paper will help shed some light on the essential requirements for capturing value with logistics optimization technology."

1. Objectives - Objectives must be quantified and measurable

Objectives are the way that we specify what we want to accomplish. In order to optimize something, you have to decide how you will know it is optimized. Using quantitative objectives, computers determine whether one logistics plan is better than another, and management determines if the optimization process is providing acceptable ROI. For example, a delivery operation might define the objective to be "minimize the sum of the daily fixed cost of assets, the daily cost of fuel and maintenance, and the daily cost of labor." These costs are both quantified and reasonably easy to measure.

2. Models - Models must faithfully represent required logistics processes

Models are the way operational requirements and constraints are translated into something the computer can understand and process. For example, we need models to represent how shipments can be combined into loads for a truck. A very simple model such as the total weight/volume of the shipments will faithfully represent some loading requirements (e.g., bulk liquids). If a total weight/volume model is used for loading new cars onto a car-hauling truck, however, the model breaks down because it is not a rich enough representation of the situation. In this example, describing the capacity of a car-hauling truck as "45,000 pounds worth of vehicles" is an impractical simplification; in reality, the trailer will hold a discrete number of vehicles depending on vehicle type, trailer configuration, and other factors. With a simple weight/volume model in this case, many of the loads that a computer thinks will fit cannot actually be loaded, while better loads are discarded because the computer thinks that they will not fit. If the model does not faithfully represent the loading process, the loads created by an optimization system are likely to be either impractical or inefficient.

3. Data - Data must be accurate, timely, and comprehensive

Data is what drives logistics optimization. If the data is not accurate and/or it is not received in time to include it in the optimization process, the resulting logistics plans will obviously be suspect. For logistics optimization processes that create operational plans for execution, the data must also be comprehensive. For example, having the weight of each shipment is not sufficient if the volume of the truck limits some loads.

4. Integration - Integration must support fully automated data transfer

Integration is important because of the large amount of data that must be considered for logistics optimization. For example, optimizing deliveries from a warehouse to stores each day requires data regarding the orders, customers, trucks, drivers, and roads. Manually entering anything other than very minor amounts of data is both too time consuming and too error prone to support optimization.

5. Delivery - Optimized plans must be delivered in a form that facilitates execution, management, and control

Solutions provided by logistics optimization are not successful unless people in the field can execute the optimized plan and management can be assured that the expected ROI is being achieved. The field requirements are for simple, unambiguous instructions that are easily understood and executed. Management requires more aggregate, centralized information regarding the plans and their performance against key benchmarks over time and across facilities and assets. Web-based interfaces are becoming the medium of choice for both management and execution.

6. Algorithms - Algorithms must intelligently exploit individual problem structures

The biggest differentiators among logistics optimization technologies are the algorithms (the computer-based processing strategies used to find the best logistics plan). An irrefutable fact regarding logistics problems is that each has some special characteristics that must be exploited by the optimization algorithms in order to provide optimum solutions in a reasonable time. Therefore, it is critical that (1) their special problem structures be recognized and understood by the analysts setting up each logistics optimization system; and (2) the optimization algorithms being used have the flexibility to allow them to be "tuned" to take advantage of this special structure. Logistics optimization problems have a huge number of possible solutions (e.g., for 40 less-than-truckload shipments there are 1,000,000,000,000 possible load combinations). Failure to take advantage of special problem structures means either that the algorithm will pick a solution based on some unreliable rule-of-thumb or that the computational time will be extremely long (perhaps infinite).

7. Computing - Computing platforms must have sufficient power to find optimum plans in the available time

Because of the enormous number of possible solutions in any real-world logistics problem, problems of any significant size require significant computing power. Such computing horsepower allows the optimization technology both to find the best logistics plans and to find those plans in a reasonable amount of time. Obviously, for optimization technology to be practical in a day-to-day execution environment, it must produce a logistics plan in minutes or hours (rather than days of computing time). Algorithms that utilize powerful clustered servers and parallel architectures leveraging numerous computers simultaneously can be expected to provide significantly better and faster solutions than algorithms that use single PC-based or workstation-based technology.

8. People - People responsible for the technology must have the domain and technology expertise required to support the models, data, and optimization engines

Optimization technology is "rocket science," and it is unreasonable to expect it to function well over time without at least a few "rocket scientists" to keep it running. These experts must ensure that the data and models are correct and that the technology is working as designed. It is unrealistic to expect a complex set of data, models, and software to be operated and supported without considerable effort from people with the appropriate technical and domain knowledge and experience.

9. Process - Business processes must support optimization and have the ability to continuously improve

Logistics optimization requires a significant ongoing effort. There is invariably going to be change in logistics goals, rules, and processes. Systematic monitoring of data, models, and algorithm performance is required not only to adapt to change, but to initiate change when opportunities arise. Failure to put into place processes to monitor, support, and continuously improve logistics optimization invariably results in optimization technology being either poorly utilized or becoming "shelf-ware."

10. ROI - Return on investment must be provable considering the total cost of technology, people, and operations

Logistics optimization is not free. It requires significant expenditures for technology and people. Proving ROI requires two things:

(1) an honest assessment of the total cost of optimization and

(2) an apples-to-apples comparison of the solutions being produced by optimization technology versus benchmarked alternatives.

On the cost side of the calculation, there is a strong tendency to underestimate the ongoing cost of using logistics optimization technology, especially if a company buys "do-it-yourself" PC-based packaged software that requires a team of trained users and support personnel to run it on an ongoing basis. It is seldom the case that the ongoing annual cost of effectively utilizing logistics optimization technology is less than the initial cost of the technology (e.g., software licenses, implementation fees). If the total cost of a logistics optimization solution decreases after the first year, it is likely that the solution quality will decrease proportionally.

On the benefits side, determining the impact of logistics optimization technology requires:

(1) benchmarking with regard to key performance indicators before implementing the technology,

(2) comparing the results from logistics optimization to the benchmarks, and

(3) performing regular audits of your logistics optimization performance.

Developing an ROI requires having good ways to determine what the baseline is, understanding the cost of the technology and the people involved, measuring improvement, and then continually monitoring performance. Because performance data is rarely available and the monitoring process requires ongoing attention, few companies today know how effective their logistics optimization efforts actually are.

    Dr. Ratliff's white paper on the "10 Rules for Logistics Optimization" can be obtained, free of charge, by logging onto the Velant web site at www.velant.com.