Chicago hotels sold a record summer. The revenue barely moved.

Tuesday, 9 a.m. On a revenue desk in a Chicago hotel, the week opens on a rate calendar: a dozen room types, ninety days out, a grid the manager nudges by hand. Two screens over, the channel report shows how many of last night's arrivals came through an online travel agency at a double-digit commission. Neither screen agrees on what a room is worth.

That gap is where hotel revenue management lives, and in 2025 it stopped being a rounding error. US hotel occupancy fell 1.2 percent to 62.3 percent while average daily rate rose only 0.9 percent to $160.54, so revenue per available room slipped to $100.02, its first full-year decline since 2020. Rooms did not empty out. Rate and mix failed to cover the difference.

Get your rate calendar and channel mix mapped, and see which nights you sold too cheap

Chicago proves it from the other side. Downtown filled a record 3.56 million room nights this summer, up 4.3 percent over 2024 and past the old 2019 high. Revenue from those nights rose 0.8 percent. Sell four percent more rooms, collect under one percent more money, and you have the shape of a revenue-management problem.

What hotel revenue management actually is

Strip away the software names and revenue management is four decisions, repeated every day for every future date. What base rate a room type carries. How that rate moves as the date fills or stalls. Which channels get which rates and inventory. And which length of stay or segment earns priority when demand is tight. RevPAR, the number that grades all of it, is occupancy multiplied by average daily rate, so a full hotel at the wrong price and a half-empty one at the right price can post the same line. For years the reflex was to chase occupancy, because empty rooms are visible and unsold nights feel like failure. The 2025 numbers argue for rate: when demand is flat, rate and channel are the only levers left, and most independents still pull them by hand.

Where the revenue actually leaks

Three rooms drain it, and none is an empty guestroom. The first is the stale calendar: a rate set weeks out and never revisited is a guess frozen in place, and when a McCormick Place citywide or a Soldier Field concert moves demand, the hand-tuned grid reacts last. The second is channel cost: an OTA booking carries a double-digit commission, so two rooms at the same rate return very different money by who sent the guest, and a hotel that has not read the commission math on each channel is pricing blind. The third is the discount reflex, cutting rate to protect occupancy, which is how a record summer becomes a flat revenue line.

The trap under all three is treating a full house as the win. Costs no longer sit still while you chase heads in beds: AHLA found property-level costs rising faster than revenue in 2024, with operations, sales, and IT each up nearly five percent. A sold-out night at a cut rate can lose money the occupancy report never shows.

The order we would run it

When we map a hotel's revenue week, we fix it cheapest and highest-trust first, not by buying a model and hoping.

1. One rate everyone trusts. A single source of truth for the calendar, so the front desk, the channel manager, and the OTAs never quote three prices for the same night.
2. Rules before models. Turn the moves the desk already makes into plain rules: raise on compression, hold on soft dates, cap the discount. That catches the stale-calendar leak without a black box.
3. Watch channels live. Flag parity breaks and the true net rate per channel, so a cheap OTA night is a decision, not an accident.
4. Add the model last. Machine-learning pricing earns its keep only once the calendar is clean and the rules are honest; fed a messy history, it automates last year's discounting.

An operator walks away with a map of which decision is leaking and a short list of rate moves the desk can make without a meeting. Most of the recovery in that same record Chicago summer sat in the first two steps.

Back to that Tuesday desk. The rate calendar and the channel report will keep disagreeing, because they answer different questions and nobody has made them share an answer. A record summer already showed what happens when it stands: the rooms fill and the revenue does not follow. The first fix is not a smarter model. It is one number the whole desk trusts.