Your comp set is lying to you

Picture the Monday revenue meeting again. The STR report is on the screen, the RevPAR index is the first number anyone reads, and it says 103. The room nods. We are winning. The rate gets held, maybe nudged, and the meeting moves on. Nobody in the room asks the only question that matters about that 103: who is in the set it is measured against, and when was the last time anyone checked?

For most hotels the honest answer is years. The comp set was built once, often at opening or at acquisition, sometimes by a brand template, occasionally by whoever set up the STR subscription. It was a reasonable set on the day it was drawn. Then the market moved and the set did not. A competitor renovated up a tier. Two new builds opened down the street. One hotel flagged to a brand that put it in a different demand pool entirely. The set on the report is a snapshot of a market that no longer exists, and every index read off it is wrong by an amount nobody can name.

How a comp set goes stale, or gets gamed

A hotel comp set is a small list of properties, usually five or six, that a hotel agrees to be measured against. STR and Lighthouse build the index off that list. The index is only as honest as the list, and the list rots in two different ways.

The first is drift. Markets are not static and comp sets are. A set that was a tight match in 2021 is loose by 2024 because the properties in it changed, not because anyone made a bad call. A competitor spends twenty million on a renovation and jumps a price tier. Their ADR climbs, the set average climbs with it, and your index against that set falls even though you did nothing wrong and lost no share. The reverse happens too. A property in your set lets itself go, discounts to survive, and drags the set average down. Now your index reads 108 and the room celebrates, when all that happened is one competitor got desperate. The number moved. Your business did not.

The second way is worse, because it is on purpose. Comp sets get gamed. A revenue leader under pressure to show index improvement can engineer it by swapping the set rather than the strategy. Drop the strongest competitor, add a weaker one, and the index climbs without a single additional room sold. This is not rare. It is one of the oldest moves in revenue management, and it travels well up the org chart because the people reading the index at corporate and at ownership usually cannot see the set behind it. They see 112 and they see a hero. The hotel under-indexed against its real competitors all year and the report said the opposite, because the report was measuring against a set chosen to flatter it.

Either way, drift or design, you end up reading a real report against a fake market. The STR data is clean. Lighthouse is clean. The pull is correct to the dollar. The set is the lie.

Why a wrong comp set flows into the operating number

Here is where this stops being a revenue-management curiosity and becomes a COO problem, the same way a misread catering pace does. The index is not a vanity metric that lives in the revenue meeting and dies there. It is an input to decisions that move the P&L.

Walk the chain. The index says you are over-indexing, so the team holds rate into a softening period instead of pricing to fill, and occupancy slips while the report still says you are winning. Or the index says you are under-indexing against a set that has quietly inflated, so the team chases it down with discounts, sacrifices rate and the flow-through that comes with it, and erodes GOP to win a comparison against the wrong hotels. Every pricing call made off a distorted index is a real decision about real revenue, and the distortion does not stay in the rate. It shows up in occupancy, in ADR, in the flow-through to GOP, in the variance to budget that the owner asks about on the call. A bad comp set is one of the quiet reasons a property can post RevPAR up and GOP flat and have nobody able to explain why.

It travels up too. The index gets lifted out of the revenue meeting and put in the owner’s report, the asset-management review, the board pack. By the time it lands in front of ownership it has lost all its context. There is no asterisk that says this is measured against a set last validated three years ago, against two properties that have since renovated, with the strongest competitor dropped last spring. There is just a number that says winning or losing, and capital decisions get made on it. Whether to fund the renovation. Whether to push for new management. Whether the property is performing. A comp set is an operating control, not a revenue toy, and most hotels do not govern it like one.

What a defensible comp set looks like

A defensible comp set is not a perfect one. It is a governed one. It has an owner, a definition, and a review cadence, the same three things every number that drives a decision needs.

It is built on real substitutability, not history or proximity. The right test is demand: when your hotel sells out, where does that guest actually book instead? That is your competitor, regardless of flag, address, or who was in the set three years ago. A property that no longer competes for the same guest does not belong in the set no matter how long it has been there.

It is reviewed on a schedule, not when something breaks. Once a year at minimum, and the moment a major supply event hits the market, a competitor renovates, a new build opens, a property changes brands. The review is a deliberate act with a record, not a quiet swap nobody documents.

And every change is logged with a reason, so the set cannot be gamed without leaving a trail. When the index moves, you can separate the part that came from your performance from the part that came from the set composition changing under it. That single discipline, knowing which part of an index move is real, is the difference between a comp set that informs decisions and one that launders them. It is the same trust problem we wrote about in the five systems that disagree: the number is only as good as the definition behind it, and the definition has to be visible to be trusted.

The set is part of the operating system

The deeper issue is that the comp set sits outside the place where operators actually work. The STR and Lighthouse data lives in one system. The set definition lives in someone’s memory or a subscription setting nobody audits. The pricing decisions live in the RMS. The consequences land in ProfitSage and the owner’s report. Four places, no connection, and the one input that determines whether the index means anything at all is the one input nobody governs.

That is the same gap underneath almost every reporting problem we see, and it is what the Operator Intelligence model is built to close: one connected source of truth, so the index, the set, the pricing, and the operating result sit in the same place. An intelligence layer that reads the way an operator already thinks, so the index arrives with its set and its history attached, not as a naked number to be taken on faith. And an operating cadence that makes reviewing the set a habit, not a fire drill the week before a board meeting. A governed comp set is a small piece of that, but it is the piece that determines whether everything you read off STR is signal or noise.

You can do this yourself with the steps above. Pull your set, test every property against where your sold-out demand actually goes, document what you change and why, and put a review on the calendar. Some teams should, and the discipline is worth more than the tooling. If you would rather the set, the index, and the operating number it feeds lived in one place that stays honest without someone policing it by hand, that is the kind of thing we build. Either way, stop reading a clean report against a market that moved on. The data is not lying to you. The set is.