Where data decides what a carrier promises

A laker's arrival depends on things no single system holds: the queue at the Welland Canal locks, wind on Lake Erie, the draught allowed at the discharge berth, the loading rate at the previous port and whether a terminal has a gang on shift. I build arrival forecasts from those inputs and publish them as a likely time with a window, then judge the forecast as a forecaster would, by calibration: how often the window held, by route and by season. Accuracy at the median matters less to a grain elevator than knowing whether a ninety-per-cent window really holds nine times in ten, because the elevator orders rail cars and labour against it. I also set the rules for overriding a forecast. A commercial manager may change one, but the change, the reason and the eventual arrival are recorded and reported monthly, and a pattern of hopeful overrides is treated as a data problem rather than a sales one.

The Great Lakes and St. Lawrence navigation season ends when the locks close for winter maintenance around the turn of the year, and ice can make the final weeks slower and more dangerous than any schedule assumes. For each vessel I forecast the probability of completing a proposed late-season voyage and reaching its winter berth with a margin before closure, using ice outlooks, lock transit history in cold weather, the vessel's hull and power and the berth sequence. The forecast governs which fixtures we offer; it never instructs a master, who keeps authority over the safety of the voyage. The same work continues ashore. Our salt customers are municipalities that need stockpiles in place before the first storm, so we forecast depot drawdown from winter outlooks and move salt by water before the locks close. The real skill is judging how far a run of mild winters should be trusted when the next may not be mild.

Shippers want to watch their cargo on its way, and most of what they ask for is safe to give. A vessel's position is broadcast by its transponder and shown on public tracking services, so withholding it protects nothing. What does need protection is commercial inference: a customer that can see where our vessel sails after discharging its cargo learns whom else we carry for, in what volume and when. I design shipper access around that distinction. Each shipper sees its own voyages, its own arrival windows and its own cargo documents; no shipper sees the next fixture, another customer's tonnage or a vessel's season rotation. Thirty-four shippers now use a single data-sharing agreement built around the questions they may answer with our data, and any request outside it comes to my team rather than being met by a well-meaning manager with a spreadsheet.

Canadian and US customs authorities both require carriers to send cargo and conveyance information electronically before a truck or vessel arrives, with lead times that differ by mode and cargo. I do not treat that filing as a clerical step at the border. It is the last link in a data chain that begins when a shipper places an order, and most errors are created at that first link. At Thibervaux I moved capture of commodity descriptions, consignee details and weights to order intake, checked against each shipper's approved descriptions, and stopped cross-border dispatch without complete data. At Morvaise the same principle governs cargoes loaded in US ports for Canadian discharge, and the road division's lanes into Michigan, Ohio and New York. I agree data definitions with our customs brokers and our trade compliance lead. I do not give customs advice, and I point colleagues to the people who own each filing requirement.

Telematics produces data about people as well as machines: a driver's braking, a watch officer's course changes, a truck's position at three in the morning. I govern it through a purpose register that names each data element, the uses it may serve, who may see it and when it is deleted. A use outside the register needs a recorded decision, not a manager's request. Drivers and crews can see the events recorded about them and add context before anyone acts on them. As a carrier operating across provincial and national borders, Morvaise is subject to the federal private-sector privacy law for employee information, which confines collection to purposes a reasonable person would consider appropriate in the circumstances. I apply that as a design test at the start rather than a legal review at the end. I am firm about retention, because most video and location data loses its legitimate use within days and keeping it longer only adds exposure.

Steelmakers, grain companies and municipalities increasingly ask for the greenhouse-gas emissions of the cargoes we carry, because transport forms part of the emissions they report. I give them figures built from fuel actually burnt, read from flow meters on every vessel and fuel cards in the road fleet, rather than from an average factor per tonne-kilometre. Each cargo's statement includes its share of the ballast voyage the vessel made to reach the loading port, allocated by a published rule, together with an uncertainty range. I work with our sustainability lead, who owns the company's own inventory, and with our auditors, who review the method each year. The difficult conversations are with customers who would prefer a lower figure. On routes with long ballast legs, a factor-based number can sit well below the measured one, and I would rather explain a higher figure we can reproduce than defend a lower one we cannot.