Freight data that can bear its use

A consolidation or routing model is a set of assumptions about freight that has not yet arrived. I ask my analysts to label every input as measured, declared or estimated, and to show how a recommendation changes when the weakest input moves. Dimensioning scans, scale readings and arrival scans are measured; an outlet's cubic declaration is not. Dispatchers keep the right to override, and we record the reason, because weather, road closures and a hub manager's knowledge of a customer will not appear in the data. Those override notes are the most useful material we have for improving a model. I report savings only after a line has run for a full quarter on the new plan. A projected departure saving is a hypothesis, and I would rather present a smaller measured result than defend a larger modelled one.

Our network touches drivers, outlet staff and millions of consignees, and franchised outlets handle much of that information first. Under the Personal Information Protection Law, financial account details and location tracks are sensitive personal information, which calls for a specific purpose, necessity and separate consent. I begin with an inventory of who holds what and why, then remove copies that serve no current purpose. Drivers verify their documents once with a central desk, and outlets see a status rather than an identity number. Location data from vehicles is collected while a driver is on a job, not across the rest of the day. I also treat decisions about people as decisions: a driver barred after a cargo loss should be told the reason and have a way to respond. Data protection here is mostly practical housekeeping, carried out across 1,710 independent businesses.

Our cross-border trucking service into Vietnam and Laos, and our foreign shipper customers, regularly ask for data abroad. Chinese rules on sending personal information outside the country require a lawful basis and, depending on the volume and type of data, a security assessment, a filed standard contract or a certification route. Some exemptions were introduced in 2024, including transfers genuinely needed to perform a contract with the individual. I do not interpret those rules myself; our legal team does. My contribution is a field catalogue. Each data element is classed as exportable, exportable after review, or onshore only, with a reason. Most requests can be met with shipment references, milestones and exception codes. When a customer needs a name or telephone number abroad, we ask what the field is for and look first for an onshore alternative.

Shippers increasingly ask for the emissions of their consignments for their own reporting. In a consolidation network that is an allocation problem, because one trailer carries freight for dozens of customers. We calculate by lane, using fuel-card records for our 1,240 tractors and distance-based estimates for subcontracted vehicles, allocated by tonne-kilometre and following the approach set out in ISO 14083. Each statement shows the share of the figure resting on measured fuel. In 2025 that share was 61 per cent across the network and lower on lanes run mainly by owner-operators. I have declined requests for figures per parcel, and for statements describing a shipment as carbon neutral, because our data cannot support either claim. A shipper is better served by a figure with an honest margin than by a precise number that an auditor later unpicks.

E-commerce platforms, freight-matching applications and shippers' own systems all want our data, usually through an interface whose terms they wrote. I start from the service. A platform selling our delivery needs tracking events and delivery status; it does not need driver identities, lane-level volumes or our outlet economics. We publish a standard event set, agree in contract what the receiving party may do with it, and refuse bulk extracts offered as a convenience. Aggregated lane volumes are commercially sensitive, and they can reveal a shipper's sales to a competitor using the same platform. When a platform changes its own terms, my team reviews what we send before the next release rather than after it. Sharing less has occasionally cost us a featured listing. I would accept that again before I would hand over data whose later use we cannot describe.