Data governance for hotels, in plain operator English

Picture the ownership review. The asset manager asks a simple question. What was group revenue last quarter. The revenue manager reads one figure off the PMS pull. The controller reads a slightly different one off ProfitSage. The catering director, looking at the sales-and-catering system, has a third. All three numbers are close. None of them match. And for the next twenty minutes a meeting that was supposed to be about the business becomes a meeting about which number is real.

Every operator has sat in that meeting. The instinct is to blame the tools, or the people, or the export that ran wrong. The actual cause is something nobody wants to say out loud, because it sounds like a consultant word. The cause is that the property has no data governance. Nobody agreed on what “group revenue” means, nobody owns the source it comes from, and nobody can show how the number was built. So three honest people read three honest numbers and the room loses an hour.

What governance actually means here

Forget the word for a second. “Data governance” is a phrase that has been ruined by people selling frameworks and committees. For a hotel group it means four plain things, and you can hold all four in your head.

One agreed definition per metric. There is exactly one thing called “occupancy,” one thing called “RevPAR,” one thing called “F&B contribution,” and it means the same thing in the flash report, the dashboard, and the board deck. Not three close cousins. One.

One owner per data source. Every system that feeds a number (the PMS, the POS, the sales-and-catering system, the labor system) has one named human who is responsible for it being right. When a code is wrong, you know whose phone to call. Not a department. A person.

Rules for how codes and segments get used at the property. The PMS market segments, the POS menu categories, the rate codes, the event types: somebody decided what each one is for, wrote it down, and the property follows it. New code, you ask first. You do not invent a “social-2” segment on a slow Tuesday because the dropdown was confusing.

An audit trail. For any number on any report, you can answer one question without panic. Where did this come from. Which source, which definition, which date it was pulled, what was excluded. So the number holds up in front of an owner instead of folding the moment someone pushes on it.

That is governance. Four things. One definition, one owner, agreed rules, an audit trail. None of it requires a committee or a forty-page policy. It requires somebody to make the decisions and write them down where the next person can find them.

Why ungoverned source data corrupts everything downstream

Here is the part operators underestimate. The damage does not start in the report. It starts at the property, in the source systems, in the boring fields nobody governs.

A front desk agent codes a corporate group as transient because the group block was already cut off. A banquet captain rings a plated dinner under the wrong revenue category in the POS because the right one was three taps deeper. A new sales manager, trained by nobody, starts using an event type that means something different from what the last person meant by it. Each of these is a five-second decision by someone doing their actual job. None of them is malicious. None of them is even obviously wrong in the moment.

But every report you build sits on top of those five-second decisions. Pace, mix, segment analysis, the flash, the forecast, the board deck. They all pull from the same PMS codes and the same POS categories. If the source is dirty, every number downstream is dirty, and the polish you add on top (the nice dashboard, the clean chart) only makes the wrong number look more trustworthy. We have written before about the five PMS fields nobody trusts, and the reason they cannot be trusted is exactly this. They are ungoverned at the point of entry, so they corrupt silently for months before anyone notices the totals do not tie.

This is why ungoverned source data is worse than missing data. Missing data, you notice. A blank is loud. A miscoded segment is quiet. It flows all the way through to the ownership review looking like a real number, and by the time someone catches it, three quarters of reports have already been built on it and three decisions have already been made off it.

Hotel data governance without the committee

So how does a hotel group actually govern data without turning it into a bureaucracy nobody follows. You keep it operator-sized.

Start with the metrics that show up in front of an owner. You do not govern everything on day one. You govern the ten or fifteen numbers that appear in the flash, the ops call, and the ownership review, because those are the ones that cost you credibility when they disagree. Write a one-line definition for each, in plain English, the kind a GM could read. “Group revenue is definite group rooms plus contracted catering tied to a block, recognized on the stay date, excluding cancellations.” That sentence is governance. It is also the thing that ends the twenty-minute meeting.

Then name an owner for each source. The PMS has an owner. The POS has an owner. The sales-and-catering system has an owner. Their job is not to be a data scientist. Their job is to keep the codes clean and to be the person you call when a number looks off. One name per system, written down.

Then write the coding rules and put them where the property can see them. Not a binder in corporate. A short reference the front desk, the banquet team, and the sales office can actually find. What each segment means. What each POS category is for. When to ask before inventing a new code. Most miscoding is not defiance, it is a person guessing because nobody ever told them.

Then keep the audit trail by default, not on request. Every number that hits a report should carry its lineage with it. Which source, which definition, pulled when, excluding what. When that travels with the number automatically, “where did this come from” stops being a fight and becomes a click. The number holds up because you can show your work.

None of this is glamorous. That is the point. Governance is the part of data work that never demos well and never wins the meeting, and it is the part that decides whether everything you build on top of it is true.

Where this sits in the operating model

Governance is not a project you finish. It is one of three things working together. We call it the Operator Intelligence model. One connected source of truth, so the PMS, the POS, and the sales-and-catering system stop disagreeing because they finally feed one place. An intelligence layer that reads the way an operator already thinks, in the metrics you actually use, not raw tables. And an operating cadence that turns it into a weekly habit instead of a quarterly fire drill.

Governance is what makes the first one trustworthy. A single source of truth that is full of ungoverned codes is just a faster way to be wrong at scale. The definitions, the owners, the coding rules, and the audit trail are what let one connected source actually be the source, the thing the ownership review can lean on without re-litigating every number. Without governance, the platform is plumbing with no water you would drink.

You can do this yourself. Pick your fifteen metrics, write the definitions, name your owners, publish the coding rules, and make lineage travel with every number. The hard part is not knowing the four things. It is the discipline to keep them current through a PMS change, a menu reset, and a year of property turnover. If you would rather it just held, defined hard and maintained, so the next ownership review is one number everyone trusts instead of three nobody does, that is the kind of thing we build. Either way, stop letting a five-second coding decision at the front desk decide what the owner sees three months later.