Recently, I was asked to setup a forecasting system for a trucking company. Typically, trucking companies get along with common sense, gut feeling and a dose of good luck. This is not always enough. They often suffer from low loading degrees, high charter costs, waiting times, late deliveries, and complaining customers. Forecasting the “erratic market fluctuations” can help them perform better and makes sense because these fluctuations are almost always caused by known factors. So an important part of the fluctuations are predictable!
In order to make useful demand predictions we need to understand the nature of the fluctuations. Depending on that, we can select the most suitable forecasting method. Let us first study some charts.
This chart shows random fluctuations. These are a fact of life and can be forecasted because the fluctuations remain within a certain bandwidth. That’s a good start, proceed to the next one.
In this graph we see a pattern that repeats itself over time. It is important to include this pattern into the prediction. The third chart shows us a rising trend line. We know that the global need for transport is rising steadily, so this comes as no surprise at all.
Transport planners can usually list the most important sources of fluctuations in the demand for transport, for example:
- Daily patterns, usually one or more days of the week have higher or lower demands
- Monthly seasonality, demand fluctuating per month of the year
- Sales promotion by customers
- Demand shift caused by holidays, e.g. around Christmas or Easter
- End-of-quarter and end-of-year peaks, caused by Sales
It is important to identify these factors and study their behaviour and impact, in order to incorporate them into useful predictions. With a clear understanding of the so-called forecast profiles, the most suitable forecasting method can be selected.
The next table illustrates a useful classification of forecasting methods.
This table can be a guide to select the proper forecasting method. That is an important step, but please bear in mind that successful implementation also requires:
- Acceptance by the planners, the forecasting system should fit their decision-making processes
- Availability of historical data, preferably covering more than 3 years
- Automatic detection of changes, because the influence of different factors can change over time (trends may break)
- Continuous analysis of prediction errors and adjustment of the forecasting parameters
- Measurable and traceable results
With a well designed demand forecast system, planners can make better decisions. They will be able to achieve a better match between demand and required resources like trucks, drivers and depot employees. Thus they will improve service levels, truckload utilization, charter costs and they will increase the efficiency of transhipment activities.