The enhanced algorithm uses more than 150 data points associated with a single load — including shipper, carrier, lane, rest patterns, load, traffic and weather — and learns continuously from millions of monthly shipments to give shippers and carriers more precise arrival times.
“Ultimately what we need to apply any AI or data science is enough data,” says Priya Rajagopalan, chief product officer of FourKites. “This is not something we could have done four years ago. Every year we successfully add terabytes of data to the platform, enabling us to do highly advanced predictive models on ETA and dwell time.”
An early adopter of FourKites’ new feature is a top food and grocery wholesaler with more than 30,000 weekly deliveries. The customer achieved 91 percent accuracy in predicting arrival times within a one-hour window, according to the announcement, for improving communications with its own customers to coordinate delivery operations and staffing efficiency.
FourKites says the new feature can be useful for carriers as well. Some carriers already use ForKites as a customer service web portal to provide their shippers with shipment status information.
Additionally, FourKites is working on a new recommendation engine for shipper and carriers that will specify the time of day that loads must depart by in order to reach their destinations on time. The company is planning to release this engine, which uses historic and real-time traffic conditions, in the second quarter, she says.
The accuracy of DynamicETA is much more precise than the accuracy rates of traditional shipment visibility tools. These tools are roughly 45 percent accurate for day-of arrivals, FourKites says, and are based on top-down estimates using limited data – typically the standard drive time of the projected route, adjusted by remaining hours of service.
