Could "Big Data" Predict Freight Rates?

What if you had a good idea of what container freight rates for your cargo might be next week? If you knew they would go up, you could book your shipment now; if you knew they would go down, you might be able to wait for a lower price. Christian Ove Sørensen, global head of marketing and sales at software company Portix Logistic Software (PLS), believes that there may be a digital means to make this possible. 

At present, PLS offers shippers and freight forwarders a software suite that organizes all their transportation rates – sea freight, airfreight, LTL, tariffs, rebates, allocation management and quotations. This gives the firm considerable insight into the world of freight, and Sørensen thinks that some of this data could be used for sophisticated forecasting. 

"Can you predict with some kind of accuracy what freight rates will be one to two weeks out?" he asks. "Some customers tell me that this will never fly, that it's not a rational market. But some are interested to start exploring this."

Sørensen wants to approach this problem using predictive analytics and machine learning – an advanced computing technique in which a program "studies" a giant data set to find patterns and create forecasts. In the energy industry, this technique is already in use for predictive maintenance – the practice of analyzing sensor data to forecast equipment failure and optimize interventions. Predictive analytics is also widely in use in the retail business for determining pricing strategies, forecasting demand and figuring out the best way to advertise. 

Sørensen thinks that the freight forwarding sector could benefit too. For forwarders, the value wouldn't only be in the forecast itself – it would also be in its ability to confirm human intuitions about where rates will go. "If a forwarder could tell a shipper that they have a statistically-driven prediction for what they believe rates will do in two weeks, and that the prediction supports their own independent analysis, they could use this to make an additional argument for a sale," Sørensen says. 

To build this predictive model, PLS needs a large amount of past and present pricing data. While the firm already handles plenty of this information every day, it doesn't aggregate its clients' data or use it without permission. "We would need a customer who would let us work with their data, and we would partner with an analytics firm to see if we can make a predictive model based on that information. We already see interest from our existing customer base for this and we hope that this joint approach will enable us to determine if this is a nut that can be cracked," he says. If successful, the project could bring the kind of data-driven forecasting used by tech giants and oil majors to the much more traditional business of booking cargo.