Ken Adamo | Chief of Analytics, DAT
IN PRICING, THERE’S A “ZONE OF INDIFFERENCE,” which is the range of price points where the buyer will not change a purchase decision. A similar concept applies to forecasts.
Predictions don’t need to be perfect. You probably wouldn’t lose faith in a meteorologist if they’re off by a couple of degrees or if a rainstorm starts an hour earlier than predicted. But you still need to be able to plan for tomorrow.
In order to do that, predictive models have to produce accurate, reliable results in a consistent pattern that supports real-world solutions, allowing businesses to improve profitability and mitigate business risk.
DAT iQ analysts began research and back-testing of predictive models in late 2018, refining the approach for 18 months before launching the Ratecast freight rate prediction model. The testing revealed four requirements for any freight forecasting model to produce accurate results.
To achieve any degree of accuracy, a forecasting model requires a large database of historical rates recorded over a period of years, with minimal gaps in that timeline.
In order to avoid forecasts becoming biased, many sources are required to keep individual companies from influencing the predictions.
3.Supply and demand metrics
Since individual lanes, regions and markets respond differently to seasonal and economic influences, a forecasting model requires data that measures supply and demand directly. Metrics such as tender rejection rate only serve as a proxy for supply and demand.
Freight markets are varied and often volatile, so freight forecasting requires applied algorithms that account for variability over time.
These four requirements have been key to the success of the Ratecast predictive model, available in DAT RateView. Across more than 1.5 million daily pre-dictions, Ratecast has been over 95 percent accurate.
This next generation of freight analytics will be instrumental in nav-igating the uncertainty ahead.
To learn more about predictive models from DAT iQ, visit DAT.com/Ratecast