Dynamic predictive visibility: advanced AI increases accuracy of estimated vessel arrival times

Predicting the future has always intrigued mankind. Thankfully, AI has proven to be far more efficient than the average crystal ball. The logistics industry moved fast to reap the benefits of advanced AI’s ability to provide future insights by launching Dynamic Predictive Visibility (DPV).

DPV gathers relevant information from a range of sources, analyses filtered data and provides a comprehensive outlook for container arrival times. In practice, this means that it consolidates container status updates, schedule changes and delays from multiple sources and standardizes them across all carriers. Given the sheer volume of global trade and number of factors that can impact shipments, we need AI for this task, as there’s too much data for humans to analyse and communicate manually.

The drive to capitalize on DPV began more than a decade ago. Since then, DPV has evolved, thanks to the proliferation of algorithms working in the background. This increased processing capacity has helped to improve the accuracy of DPV’s predictions from just 30 percent in its very early stages to over 70 percent today.

"With AI's evolution, the voyage from 30% to 70% accuracy in estimated vessel arrival times showcases the power of innovation. DPV's real-time dashboards provide a comprehensive outlook, enabling proactive decision-making, efficient inventory management, and avoiding delays and charges. At CEVA Logistics, DPV is more than a tool; it's our commitment to over 1.3 million annual shipments. As we navigate the seas of improvement, collaboration between DPV developers and logistics experts will ensure a highly responsive system, heralding a future where 90% accuracy is not a distant shore but the next port of call." - Joshua C. Bowen, Global Head of Ocean Freight, CEVA Logistics.

 

The advantages of DVP include:

• Visibility on status changes, delays and warnings, which leads to efficient inventory management and logistics planning
• Proactive decision making enablement for shippers and distributors
• Ability to plan and mitigate demurrage and detention charges
• Avoidance of fees from late and lost shipments
• User-friendly real-time dashboards for enhanced visibility
• Comparison of performance over time and across carriers
• Predictive learning analytics, which help improve performance and decision-making over time
• Carrier-neutral data
• Benchmarked information

 

Real-time, accessible updates

Prior to DPV, buyers depended heavily on carriers to inform them of the status of a vessel’s journey and estimated arrival time in order to arrange for inventory management and ground transportation, as well as avoiding demurrage charges. But only the captain and crew would know about important changes or issues with a vessel’s status.

Delays in communication often resulted in buyers being ignorant of any issues with a shipment, causing financial and logistical problems. Now, thanks to DPV, we can receive important updates about a shipment’s status automatically, including updates about:

• Vessels in distress (fire, collision, etc.)
• Arrival times
• Port strikes
• Inclement weather

One of the key benefits of DPV is the way it delivers the information in a succinct dashboard, which will update continuously to reflect any changes that could impact a shipment’s arrival time. Dashboards can also show statistics and compare DPV’s estimated time of arrival (ETA) against the original estimated time of arrival and/or contractual requirements, as well as generating a variety of reports and data analysis.

The real-time dashboard shows the turnaround times at the port of loading (POL) and port of discharge (POD). The dashboard indicates the average container waiting time in conjunction with historical statistics about turnaround times as per each individual POL. At a POD, the dashboard will provide information on container waiting times and time in delivery until the container returns to the depot.

Returning containers to depots can involve ground transportation from the port to an inland container facility, conducted by the carrier as per predetermined parameters. It’s possible to adjust these parameters to reflect container waiting time in the last inland facility on the dashboard, along with that facility’s historical statistics on turnaround times.

 

Increased data drives improved predictive capacity

DPV incorporates predictive learning, where it can learn from previous occurances and predict them better in the future. For example, if there were a three-day delay due to a category two hurricane in the North Pacific at a certain latitude where a vessel was passing through, DPV would use this estimate to inform other estimated arrivals should the weather forecast predict another hurricane of the same or similar strength, at the same or similar location.

All industry verticals need an accurate estimate of a shipment’s arrival. Timely shipments and inventory have major impacts on time-sensitive events, such as product launches or holiday guarantees. On the other hand, for low-cost, all-season products, shipment delays can translate into storage and warehouse costs. DVP allows shippers and distributors to make informed business decisions.

“At CEVA Logistics, we use Dynamic Predictive Visibility (DPV) to track the status of containers and offer real-time status updates to our customers. With over 1.3 million containers shipped annually at CEVA, this is an extremely important tool to provide the visibility that is necessary for our customers. There is still room for improvement with DPV, but with AI evolving at an unprecedented pace, we are positive that DPV will continue to advance and benefit our customers.” - Bassel El Dabbagh, Regional Managing Director, Levant, CEVA Logistics

DPV must continue to add more algorithms to further enhance its efficiencies. Its current accuracy rate is approximately 70 to 80 percent, so it will accurately predict the arrival information for seven to eight of every ten shipments. Errors are often the result of an overestimated delay, due to the inclusion of contingency delay risks and the gap can be significant. For example, on a recent

shipment, DPV predicted a seven-day delay which actually turned out to be just a two-day delay. Based on the rate of improvement over the past ten years, plus the current accelerated development of AI, one could argue that it should take no more then three years to reach almost 90 percent accuracy.

Without doubt, the logistics industry will continue to heavily rely on DPV and its dependence on carrier information will be marginal. Through this amplified usage, DPV is maturing and evolving. It is no longer a case of only including additional algorithms. Instead, we need to encourage collaboration between DPV developers and logistics experts, in search of a highly responsive DPV system that will most fully benefit the industry.