I'm Ben! đź‘‹
Our team was approached by the Port of Montreal, along with other Canadian port authorities, with a problem regarding the unexpected or inaccurate arrival window of vessels at the port terminals.
As part of kicking off the Vessel ETA (or VETA) project, we needed to understand the ins and outs of port and terminal operations to better assess where we could provide software and technology that would be useful for the entire port ecosystem.
Port authority stakeholders were interviewed about their reality and needs in terms of vessel arrival time information and how they could better insure their teams would be available to assist vessels in navigating the Saint-Laurent’s difficult waterways. Terminal operators were interviewed to better understand how vessel arrival could help them staff their offloading teams to meet more efficient turnaround times for ocean liners and improve their reputation, thus increasing throughput at the port terminal in question.
The predictability, but also the evolution of vessel arrivals through time was clearly a resounding need throughout all our user research. Our team had to wrangle a lot of data sources in order for our prediction model to be accurate enough to be more useful than their current methodology.
Something to know is that vessel operators’ business interests don’t really align with the port ecosystems’. On the one hand, vessel operators need to move cargo from one place to another while optimizing their fuel costs as much as possible, meaning lowering speed when possible, and on the other, ports would like ships to fit into a one-day arrival window to insure offloading crews and berths are available.
This also means that weather, shipping routes, tidal movements and currents greatly influence the arrival time of vessels since it is not in their financial interest to accelerate to meet the port’s timetable. The result of these conflicting interests is a suboptimal solution: the ports would contact the last port of call, the port from which the vessel departed last, and ask when the vessel left the terminal so they could estimate a window of arrival, give or take a week. They would also attempt to contact the vessel crew to get a sense of their estimate, but often wouldn’t greatly increase their confidence.
Yeah… the margin of error is that big, hence the reason they we’re looking for technology solutions that could help them narrow down the window of arrival significantly.
From the insights my colleague and I gathered from our user research, our development team, assisted by our solution strategists started reaching out to data providers to complement the historical data that was provided by the terminals and port authorities.
While the development team was working on crafting the custom predictive AI model for vessel arrival times, I got started on drafting ideas of features to include in the software could provide.
Our business team also needed some help on marketing our solution and peek interest from Canada’s maritime and logistics players. We settled that demonstrating the performance of our AI model by running it against the historical data that was provided would be a nice way for others to understand the accuracy of the model through time. We wanted the demonstration to feel like a polished product that could be reused down the road to continue monitoring the performance of the model and for our sales team to have quality presentation assets when approaching prospects.
Our team had the opportunity to showcase our demo at a conference event and here is what the VETA model demonstration product looked like:
While the model demonstration tool was under development, I continued my work on designing the vessel tracking and data validation software.
Here is an interactive prototype of the solution: