Data governance: who owns each piece of data and what to govern first

The board of a multipurpose port terminal reviews the quarterly report. Operations presents a quay productivity of 28 moves per crane hour; the parent company, with the same vessels and the same quarter, calculates 24. Finance brings credit notes for services billed to the wrong customer, and a claim from a bulk customer over a coal shipment that has three different weights. The meeting is spent establishing which figure is correct.

Nobody in that room lied. Operations divides the moves by the crane’s net hours and the parent company by its gross hours; the customer exists twice in the systems, with its name spelled two ways; and the coal has the weighbridge weight, the draft survey weight and the bill of lading weight, with no contract clause saying which one prevails. The case is a composite of situations that are common in the industry, and its lesson is simple: ungoverned data rarely fails for lack of systems. It fails because nobody has the authority to decide what it means, who is accountable for it and what happens when two departments disagree.

Governing is not the same as managing

Data governance is the exercise of authority over the company’s data: who decides what each piece means, who is accountable for its quality and how disagreements are resolved. Data management carries out those decisions: it captures, integrates, cleans, documents and protects. Four roles sustain it, and none of them is a new position:

• The data committee, which in a mid-sized company can be a standing item on the monthly executive committee agenda.
• The data owner, the head of the department that produces the data: the chief financial officer for billing, because whoever knows whether a rate was applied correctly is whoever collects it.
• The data steward, who watches the rules day to day and fixes errors where they originate.
• The technical custodian, who runs databases, access and integrations, but does not decide what the data means.

The most common confusion is handing the owner’s responsibility to the technical custodian. IT ends up deciding which weight counts for billing a bulk shipment, or which customer record survives when there are two, using technical criteria on a matter that is commercial or operational.

The same data arrives from several sources

In a multipurpose terminal it shows up across all three types of cargo. The shipping line sends the type of each container by EDI with its ISO 6346 code, and the terminal groups those types in its terminal operating system (TOS) to apply its yard strategy; if a code is not in the mapping table, the planner corrects it by hand before the vessel arrives. In bulk, the weighbridge, the draft survey and the bill of lading give three legitimate weights that rarely match. In general cargo, the manifest may declare pallets while the discharge tally counts boxes.

That last difference has legal consequences in Colombia. The carrier must report to the DIAN, the country’s tax and customs authority, any surplus or shortage in the number of packages and any excess or deficit in the weight of bulk cargo within 24 hours after the end of a vessel’s discharge, and has five days to justify them (Decree 1165 of 2019, articles 151 and 152). None of those sources is wrong; the problem appears when nobody has decided which one prevails for each use.

Where to start

You do not govern everything at once. We recommend six stages, one domain at a time: assign owners, prioritize, define meanings, unify records, measure quality where the data originates, and operate the model. Two questions are enough to prioritize: how much money or risk the domain moves, and how often it produces discrepancies. In our reading, a terminal usually starts with customers, billing and productivity.

The size of the problem can be measured in an afternoon. In a study published in Harvard Business Review, 75 executives reviewed the most recent records created in their area, and on average 47% had at least one critical error (Harvard Business Review). The sample is not from the port sector, but the exercise can be repeated in any company before approving a budget.

The regulator is already doing the reconciliation

The Superintendencia de Transporte, Colombia’s transport regulator, publishes the official port traffic statistics with the data that port companies submit. In its 2024 bulletin it explains that it verified what was submitted, requested the bills of lading or equivalent documents to identify inconsistencies, and warned that the figures are provisional (Superintendencia de Transporte). When the regulator asks for the supporting documents, it is doing the reconciliation the company did not do.

And then, artificial intelligence

Gartner predicts that through 2026 organizations will abandon 60% of artificial intelligence projects that are not supported by AI-ready data (Gartner). A model that forecasts yard occupancy learns from what it is given: if container types are grouped incorrectly or customers are duplicated, it learns the error. Data governance determines whether a model learns from the real operation or from its contradictions.

Before buying a platform, the question for the committee is who is accountable today for the data that drives operations and billing. If the answer is a department, or nobody, that is where to start.