INDUSTRIAL ENGINEERING APPLICATIONS IN TRANSPORTATION:TRANSPORTING GOODS

TRANSPORTING GOODS

Cost, Time, and Quality Optimization in Transportation

In the transportation of goods, we often develop models that minimize total cost while maintaining acceptable levels of service. This cost may be a combination of the cost of operating a vehicle (fuel, maintenance, depreciation, etc.), the labor cost of the operator of the vehicle (driver, pilot, etc.), and possibly other fixed costs. In the transportation of small packages or less-than-truckload shipments, sorting cost is also considered, representing the cost of sorting and handling the packages to con- solidate shipments. Cost may be considered indirectly, as in the case of minimizing empty vehicles or maximizing the load factor.

The time needed for transporting goods is another very important factor. Instead of minimizing general cost, transportation time may be minimized directly when time constraints are very tight or when the computation of cost is dominated by the labor cost. Whether the transportation time is

minimized directly or not, the model may include time constraints (e.g., time windows, total time of a driver’s route) or the time element may be incorporated in the input data of the model (e.g., included in the underlying network).

There is a tradeoff between transportation cost and transportation time. In the last decades, shorter product cycles in manufacturing and notions like just-in-time inventory have resulted in an increasing demand for smaller transportation times at higher prices and higher transportation cost. The small- package transportation industry has responded with many ‘‘premium’’ services that guarantee short transportation times. Even when the transportation time is not guaranteed, it represents a primary element of quality of service along with other considerations such as minimization of lost and dam- aged goods, which are often handled indirectly or are external to the cost optimization models.

Integrating Customer Needs in Transportation Planning

Although we often try to minimize cost in transportation planning, what we really want to achieve is maximization of profit (revenue minus cost). Cost minimization assumes that demand is external to the model and unaffected by the solution obtained. Demand remains at assumed levels only if the obtained solution satisfies customer requirements and customer needs are integrated into the trans- portation planning process.

Excluding price and transportation time, many customer needs are not easily incorporated into a mathematical model. They may be included in the input data (e.g., different types of service offered) or considered when alternative solutions obtained by optimization are evaluated. Flexibility, a good set of transportation options, and good communication are of primary importance to customers, permitting them to effectively plan their own operations, pickups, and deliveries.

Forecasting in Transportation Planning

The development of effective transportation plans is highly dependent on our ability to forecast demand. Demand, in the case of the transportation of goods, refers to the expected number of ship- ments, the number of packages associated with a shipment, and the frequency with which such shipments occur. When developing transportation routes and driver schedules, demand levels are used as input and therefore need to be forecasted. Changes in demand can occur randomly or can follow seasonal patterns. In either case, if an accurate forecast is not produced, the transportation planning effort will yield less accurate results. These results have implications in the design of facilities (e.g., capacity), the acquisition of assets (e.g., delivery vehicles), and in the development of staffing plans (e.g., labor requirements). It is important to note that several factors affect demand in the transpor- tation industry. Business cycles, business models, economic growth, the performance of the shipper’s business, competition, advertising, sales, quality, cost, and reputation all have a direct impact on the demand for transportation services.

When developing a forecast, the planner must address some basic questions:

1. Does a relationship exist between the past and the future?

2. What will the forecast be used for?

3. What system is the forecast going to be applied to?

4. What is the size of the problem being addressed?

5. What are the units of measure?

6. Is the forecast for short-range, long-range, or medium-range planning purposes?

Once these questions have been addressed, the steps shown in Figure 3 guide the planner towards the development of a forecasting model that can be used in generating the forecast to support the transportation planning process.

Depending on the type of the transportation problem being solved (e.g., local pickup and delivery operations, large-scale network planning), the user may select different forecasting techniques and planning horizons. For long-range network planning in which the planner is determining the location of future distribution centers, long-range forecasts are required. Sales studies, demographic changes, and economic forecasts aid in the development of such forecasts. When developing aggregate plans in order to determine future staffing needs and potential facility expansion requirements, the planner develops medium-range forecasts covering planning horizons that range from one or two quarters to a year. Known techniques such as time series analysis and regression are usually applied to historical demand information to develop the forecast. Finally, in order to develop weekly and daily schedules and routes to satisfy demand in local pickup and delivery operations, the planner may develop daily, weekly, or monthly forecasts (short-range forecasts). Because demand fluctuations can have a direct impact on the effectiveness of pickup and delivery routes, the use of up-to-date information from shippers is critical. Techniques such as exponential smoothing and trend extrapolation facilitate this type of forecasting.

Forecasting is often considered an art. Because the inputs to any forecasting model are mostly historical and based on experience, the accuracy of such forecasts is based on the model selected and the frequency with which it is updated. Forecasting models are usually grouped into two cate- gories: subjective and objective.

Subjective forecasting models are based on judgement and experience. Several techniques exist that attempt to make use of the ‘‘expert’’ ’s knowledge to develop a forecast. These techniques include the Delphi method, jury of executives, and the use of sales force intelligence to develop the forecast.

Objective forecasting models are also known as quantitative models. The selection of a quantitative model is dependent on the pattern to be projected and the problem being addressed. There are two types of quantitative forecasting models: time series and explanatory models. In time series models, time is the independent variable and past relationships between time and demand are used to estimate what the demand will be in the future. Explanatory models, on the other hand, use other independent variables instead of or in addition to time. The variables used in the forecasting model are those that have shown a consistent relationship with demand.

When evaluating the impact of seasonal variability on the forecast, the planner has several indi- cators that can be used to refine the forecast. Leading and lagging economic indicators of the general business cycles can aid the forecaster in the refinement of plans. The Department of Commerce creates an index of well-known economic indicators. This index can be effectively applied to medium- and long-range forecasts to anticipate demand in the transportation of goods.

Several correlation coefficients can be calculated to determine how closely the forecasts correlate with actual demand. The sample mean forecast error (e.g., the square root of the sum of the squared forecast errors), which provides an approximation of the average forecast error of the forecasting model, can also be used. Choosing the best forecast technique requires an understanding of the particular forecasting problem and the characteristics of the available data. Ultimately, the planner’s ability to use past and current information to forecast stops, delivery volume, and other key inputs to the transportation planning process will determine the quality and accuracy of the transportation plans developed. With changes in technology and with the increased availability and sharing of information between companies, we expect substantial improvements in the way forecasting is done in the transportation industry.