Revenue Up, Profit Flat: Where Undistributed Expenses Are Eating RevPAR Gains

U.S. hotels grew revenue 2.6 percent in 2025 and still gave back three tenths of a point of GOP margin. Much of the leak sits in the undistributed lines that no department head owns. An owner-level cost intelligence layer pulls the general ledger, vendor invoices and utility interval data into one model, flags subscription creep and duplicate tools, benchmarks each line against CBRE and HotStats, and forecasts flow-through before the month closes.

Revenue rose 2.6 percent across the 2,216 U.S. hotels in CBRE's 2025 sample. Gross operating profit rose 1.5 percent. That gap is the whole article. CBRE's analysis in LODGING puts the result plainly: GOP margin slipped from 35.1 to 34.8 percent, EBITDA margin from 23.3 to 22.8, and EBITDA itself grew three tenths of a percent in a year when rates and occupancy both edged up.

Look at where an owner's statement hides it. On a USALI report the undistributed block is five lines with one number each and no vendor names: administrative and general, information and telecommunications, sales and marketing, property operations and maintenance, utilities. The GM can talk to every one of them on the monthly call. Very few can tell you which forty invoices built the I&T number, or why utilities jumped in a month when occupancy didn't.

That's the problem this piece is about, and the fix is a piece of software, a fairly modest one.

The Problem: The Lines Nobody Owns

Operated departments have owners. The rooms director answers for rooms labor, the F&B director for food cost, and both get measured on departmental profit every month. Undistributed expenses sit above GOP with no department to charge them to, which is exactly why they drift. CBRE's own framing is that these costs are mostly fixed, so when they grow faster than revenue the margin damage is severe.

I spent years as VP of Sales Strategy and Operations at Starwood, inside a portfolio of more than 800 hotels and $3 billion. At that scale you learn quickly that a line growing a point faster than revenue every year is a decision, even when nobody in the building made it. A subscription renews. A second tool gets bought because the first one was hard to use. A utility tariff changes and the bill gets paid on time, which is all AP was asked to do.

The trend has a long tail. CBRE found hours worked per hotel fell 7.4 percent between 2019 and 2024 while labor dollars rose 22.1 percent, per its 2025 operating cost review, which also had insurance up 17.4 percent in 2024 and franchise fees up 3.9 percent. HotStats data cited by Hospitality Net in September 2026 shows U.S. GOPPAR in 2025 still 10 percent below 2019, and administrative and general costs up 7.8 percent year to date in 2026 among comparable hotels, about a quarter of the total undistributed increase.

Revenue won't bail this out. CBRE's Hotel Horizons forecast cited in the same LODGING piece projects RevPAR growth of 1.2 percent in 2026 and 1.7 percent in 2027.

When I built Collective Retreats out to ten locations, every new site arrived with its own utility accounts, its own connectivity contract and its own set of software logins before it sold a single night. Multiply that across a few years of renewals and the undistributed block stops being overhead. It becomes a portfolio of small contracts that nobody reviews as a whole.

The Data: Where 2025 Margin Went

Here is the undistributed block from CBRE's 2025 numbers, set against revenue. Every line except A&G grew faster than the top line.

Two things stand out. Technology costs keep compounding, and the growth isn't where most people look for it. In CBRE's review of IT spending, I&T rose 4.6 percent from 2023 to 2024 against a 4.0 percent rise in total hotel expenses. Phone and internet service costs actually fell, by 1.7 and 1.0 percent. The increase came from operated-department systems: reservations, PMS, key coding, point of sale. Back-office systems rose only 1.3 percent, and IT labor was just 16.3 percent of the department, per the same analysis on CoStar, which covered 4,834 properties.

The second is utilities. CBRE estimated 2024 utility cost at $2,478 per available room across all hotels, $9.68 per occupied room, and the 2025 jump of 5 percent came after a 3.6 percent compound annual rise from 2019 to 2023. Its earlier work, published in LODGING and Asian Hospitality, found commercial electric rates up more than 20 percent since 2020 and gas settled about 33 percent higher.

The limited-service number deserves a second look. At roughly 4 percent of revenue, utilities in a select-service box cost more as a share of the top line than they do at a resort, and the tools most of those hotels use to watch them are a paper bill and a spreadsheet.

The undistributed block is five numbers on the owner's statement and several hundred transactions underneath it. Margin gets lost in the transactions, and the statement is built to hide them.

The Mechanism: An Owner-Level Cost Intelligence Layer

The fix is a cost intelligence layer that sits above the property's accounting system and reads what the accounting system already holds. It doesn't replace the GL, the AP platform or the BI tool. It joins them. Four inputs and four jobs.

The first input is the general ledger at transaction level, coded to USALI. The monthly P&L is a summary; the layer needs every journal line, the vendor, the department and the account. Most cloud accounting systems used by hotels can export it on a schedule. The second is vendor invoices, which is where the detail lives that the GL drops: the per-room license count, the per-booking fee, the auto-renewal date, the price escalator. If you already run AP automation, the extracted invoice fields are sitting there. If you don't, a language model with structured output can read a year of PDFs into a vendor, product, unit, quantity, unit price and term schema in an afternoon, with a person checking every dollar field.

The third is utility interval data. A monthly bill tells you what you spent. Fifteen-minute interval data from the meter, which many utilities offer through a customer portal or a standard download, tells you when, and that's how you see a chiller running at full load at 3 a.m. in a half-empty shoulder week. The fourth is the operating driver set from the PMS: occupied rooms, arrivals, group versus transient, covers. Without drivers, a cost is just a number. With them, it's a rate per occupied room that can be compared month to month and hotel to hotel.

Then the jobs. One is classification. Every transaction is mapped to a USALI line and to a vendor-product record, so "I&T" turns into a list of 30 to 80 named subscriptions and services with owners, unit counts and renewal dates.

Another is detection. The layer flags three kinds of creep automatically: subscriptions whose unit count exceeds active users or rooms, two or more products doing the same job (two guest messaging tools, two review platforms, a survey tool bundled inside the CRM nobody turned off), and price changes at renewal that exceed the contract escalator. Zylo's 2026 SaaS Management Index puts average license utilization at 54 percent across organizations, and the company reports that 51 percent of applications are miscoded or miscategorized in expense and financial systems. Those are vendor figures from outside hospitality, so treat them as a reason to look, and hold your own data to the final word.

A third is benchmarking. Each undistributed line, normalized per available room and per occupied room, is compared with the CBRE Trends averages for the hotel's chain scale and location type and, for subscribers, with the HotStats comp set. The USALI 12th revised edition is the reference for how these lines are defined, so check that any comp data you use was reported to the same edition. The point isn't to hit the average. It's to know which two lines sit furthest from it and why.

The last is forecasting, which is where it pays. The layer projects each line forward on its drivers and its contracts (renewals, escalators, tariff changes, scheduled rate increases) and computes expected flow-through for the month and the year before the books close. The GM sees the margin problem in week two instead of on the 20th of the following month.

The Framework: Flow-Through Before the Month Closes

Flow-through is the share of each new revenue dollar that reaches GOP. It's the one number that tells an owner whether growth is worth anything, and the 2025 averages make the point without any modeling. Take CBRE's starting margin of 35.1 percent, grow revenue 2.6 percent and costs 3.0 percent, and you land at 34.8 percent margin with about 25 cents of every new dollar flowing through. The table runs the same arithmetic for other cases.

Read the last row twice. If revenue grows at the pace CBRE forecasts for 2026 RevPAR and costs keep growing at the 2025 rate, GOP falls in absolute dollars. That's what HotStats meant in its August 2025 analysis when it observed that margin growth starts to show only in markets where revenue grows faster than about 3.5 percent.

The gap between the first and third rows is one percentage point of expense growth. On a hotel doing $20 million in total revenue, with about $13 million of costs above GOP, that's roughly $130,000 a year. Capitalized at 8 percent, before incentive fees, it's worth about $1.6 million. A good share of it sits in lines nobody is watching.

So the framework is three steps, run monthly. Rank the undistributed lines by their gap to revenue growth. Within the worst two, rank vendors by dollar growth year over year. Then attach a forecast: what flow-through looks like for the rest of the year if nothing changes, and what it looks like if the top three flagged items are fixed. My Six Sigma Green Belt training left me with one durable reflex, which is to never chase a variance you can't decompose. The layer exists to decompose it.

A line that grows a point faster than revenue every year is a decision, even when nobody in the building made it.

Tool consolidation is usually the fastest money. The evidence base is thin in hospitality and worth stating honestly.

Implementation: Building It on the Tools You Already Have

None of this needs a new platform. A single hotel or a small portfolio can stand up a first version in eight weeks on its existing accounting system, a data store it may already pay for, and the BI tool its controller already uses.

Weeks one and two are data. Export 24 months of GL detail at transaction level with vendor and account, 12 months of vendor invoices as PDFs or AP records, 12 months of utility bills, and whatever interval data the utility offers. Pull monthly drivers from the PMS. If you've already built a hotel data warehouse, this lands in it; if not, a cloud database or even a well-structured set of tables will do for a first pass.

Weeks three and four are classification. Map every transaction to a USALI line and every vendor to a product record. Use a language model to read invoices into a fixed schema and to propose the mapping, then have the controller approve it. Expect the model to be quick on vendor and product and to need correction on units and terms. That's normal.

Weeks five and six are detection and benchmarks. Load active user counts for every per-seat tool (most admin consoles export them), set the duplicate-function rules, and load the CBRE Trends averages for your chain scale, plus HotStats if you subscribe. Interval data goes into a simple baseline model by hour, day type and occupancy, so off-hours load stands out. Our energy management research covers the next step, from detection to automated control. The USALI 12 energy reporting framework is worth reading before you design the utility tables, so the consumption fields match how owners will be asked to report.

Weeks seven and eight are the forecast and the pack. Build the flow-through forecast from drivers and contracts, and produce a one-page monthly view: each undistributed line against revenue growth, the top ten flags with dollar value, and the projected year-end margin with and without the fixes. Wire it into the month-end close pack so it arrives with the P&L instead of after it.

After launch the layer runs on a schedule. New invoices classify on arrival, renewals surface 90 days before the notice date, and interval anomalies post within a day. The controller's job shifts from building the report to deciding which flags to act on. For owners who'd rather build the scorecard with help, our AI and technology scorecard and reporting service starts from this ledger.

One caution. Brand-mandated technology sits inside I&T for franchised hotels, and a lot of it can't be cut mid-term. Classify it separately so it doesn't swamp the discretionary flags, and handle it through the process in our brand mandate ledger piece.

Undistributed expenses rose faster than revenue on four of five lines last year, and they'll do it again this year by default. The default changes only when someone can see the transactions.

Frequently Asked Questions