Hidden Processing Fees, Held Payouts, and Absent Support: Reading the Fine Print of Hotel Software

Hotels pay somewhere between 0.7 and 3.5 percent of card volume to process payments, a payment platform can hold a slice of every deposit for 30 to 90 days, and the support terms that decide how long a broken check-in stays broken are often not in the contract at all. The fix is a fine-print agent: a language model that reads every software and merchant agreement into a single register, a parser that turns each month's statements into an effective rate and a payout lag, and a ticket ledger that measures support against what was promised.

The quote said 1.1 percent. That number was true, and it was also about a third of the story.

A payments consultancy that sells fee audits pulled apart an integrated hotel payments quote in May 2026 and made the point bluntly. A rate written as "1.1%++" means a 110 basis point markup sitting on top of interchange and network assessments, which they called "3-4x higher than we typically see for any type of integrated or embedded setup." They also flagged a line in the merchant terms that lets the vendor switch the hotel's payment provider "at its sole discretion." Neither detail is hidden, exactly. Both sit in documents most owners sign once and never open again.

I signed a lot of those documents. As founder and CEO of Collective Retreats, ten locations by the time it was acquired in 2025, the software orders and merchant agreements came to me for signature, and the fee schedule was always an exhibit at the back. Before that I spent years as VP of Sales Strategy and Operations at Starwood, across more than 800 hotels, where a basis point on distribution cost was a number someone defended in a meeting. Independent hotels rarely have that person. So the exhibit at the back goes unread.

The Problem: The Real Terms Live in Three Places Nobody Reads Together

Review sites tell the story in the operators' own words. On Capterra's Cloudbeds page, which carries a solid 4.3 overall from 351 reviews and a 4.0 for customer service, the complaints cluster in a familiar pattern: urgent issues that wait, support that runs through tickets and email, phone support that costs extra, "hidden fees (specifically for credit card processing)," and in one 2023 review, payouts that "suddenly stopped" for ten days. There are plenty of warm reviews of individual reps on the same page. That's the point. The experience depends on terms the hotel never wrote down.

Those terms are split across three kinds of paper.

The software agreement and order form hold the subscription, the annual escalator, the renewal and notice window, and whatever support tier was bought. The merchant agreement, often a separate click-through accepted during onboarding, holds the processing price, the payout schedule, reserve rights and the processor-switch clause. And the monthly statements hold what actually happened. Each one is readable on its own. Nobody reads them against each other, which is where the money goes.

Prostay, a PMS vendor with its own product to sell, put a number on part of it in a June 2026 cost analysis: most independent hotels "miss 35 to 60 percent of the real five-year cost" by judging a PMS on its sticker subscription. Its example of an escalator is worth doing in your head once. Seven dollars per room per month, with three annual 7 percent increases, becomes 8.61 dollars, a 23 percent rise over three years that nobody voted for.

Even the vocabulary is loose. A July 2026 guide from Smart Order opens by admitting that "hotel PMS transaction fee" is not a standard term, which is exactly why two quotes using it can't be compared. And reliability is already a switching issue: 48 percent of respondents in the HotelTechReport 2026 PMS study said they'd change vendors over it. Only 11 percent of hotels in Otelier's 2026 Hotel Operations Index describe their stack as fully integrated, so most are juggling several of these agreements at once.

The Data: What Payments, Payouts and Support Really Cost

Start with the card fees, because they're the biggest line. Michele Fitzpatrick, CEO of Eviivo, wrote in PhocusWire that hoteliers are currently paying between 0.7 and 3.5 percent in card processing fees, and that "simply moving away from a heavily marked-up blended rate can claw back 20 to 30 basis points of cost." That range is wide because hotels sit on both ends of the card mix. A guest tapping a debit card at the desk costs little. A corporate card keyed in for a deposit taken months before arrival costs a lot.

The cost of the card itself is mostly fixed. Visa and Mastercard publish their interchange tables (Mastercard's 2026-2027 U.S. merchant rates run to dozens of pages), and every processor pays the same interchange for the same card. What varies is the markup, and how visible it is. Helcim, a processor that sells interchange-plus, lays out the difference in its pricing guide: interchange-plus passes through "the true cost of each transaction" and adds a visible markup, while flat or blended pricing folds everything into one number.

The only number that settles which deal is better is the effective rate. Add every card-related fee on the month's statement and divide by card volume. Here's an illustration, not a client figure. A hotel processing 4 million dollars a year in cards that moves from a 2.9 percent effective rate to 2.6 percent keeps 12,000 dollars. That's the 30 basis points the Eviivo piece talks about, and on most statements it's buried in forty lines of codes.

Every processor pays the same interchange for the same card. The markup is the only part you can negotiate, and it's the part the quote makes hardest to see.

Payout holds are the second cost, and they bite cash flow before they bite margin. Stripe's reserves page is a plain description of how platforms do it. A fixed reserve holds a percentage of each transaction until a set release date. A rolling reserve holds a percentage on a moving window, for example 30 days. "Normally, this is 30 to 90 days to give time for customers' refunds and chargebacks to process." Its worked example uses a 25 percent reserve. On a hotel taking advance deposits for a summer season, a 25 percent rolling hold on spring bookings is payroll money sitting in someone else's account.

Hotels look risky to an underwriter for a fair reason. Money comes in long before the service is delivered, and a cancellation wave turns into refunds and chargebacks all at once. So the reserve isn't the scandal. The scandal is a reserve imposed without the contract saying when, how much, or for how long.

Support is the third cost. It never shows up on a statement. It shows up as a night auditor on hold, or a GM refreshing an inbox while the payment terminal declines every card. Plain, a support software company, published 2026 SLA benchmarks for business software that make a useful yardstick, even though hotel vendors aren't obliged to meet them.

Then there's the price you pay to keep the software at all. Vertice's SaaS inflation index hit a record 16.4 percent in June 2026 before easing to 12.7 percent in August, and Zylo's 2026 SaaS Management Index found 79 percent of IT leaders saw price increases at renewal, with 77 percent reporting unexpected costs after signing. Those surveys skew toward larger companies. A 40-room hotel has less bargaining power than any of them.

The Mechanism: A Fine-Print Agent That Reads the Paper Every Month

Nobody at an independent hotel has time to reread four contracts and reconcile three statements each month. Software does, and the pieces are ordinary. I'd call the combination a fine-print agent. It has three parts and one rule.

The first part is a contract extractor. Every agreement the hotel has signed with a PMS, payments, channel, POS or guest-messaging vendor goes in as a PDF, including the click-through merchant terms (download them from the vendor's site with the version date). A language model reads each one and fills a fixed register: fee lines and their basis, escalator and cap, term, renewal and notice dates, payout schedule, reserve triggers and limits, processor-switch rights, support hours and channels, response targets, service credits, data export on exit. Every extracted field carries the page and clause it came from. Where the contract is silent, the field says "not stated," which is often the most useful answer in the register.

Language models are good at this kind of reading. They find a reserve clause in section 14.3 of a 40-page document faster than a person, and they don't skip the exhibit at the back. They're less good at deciding what a clause means when it's ambiguous. So a person checks each extracted field against the page reference before it goes live, once per contract, and anything material goes to counsel.

The second part is a statement analyzer. Each month it takes the merchant statement and the payout report, as CSV where the platform allows it and PDF where it doesn't, and computes four numbers: effective rate overall and by card type, markup in basis points over interchange and network fees, fees that appear on the statement but not in the contract register, and payout lag (days from transaction to deposit) against the schedule in the merchant terms. The model's job here is narrow. It maps each cryptic fee code to a category. The arithmetic is plain code, and it reconciles to the statement total to the cent or it raises an exception.

The third part is a support ledger. Every ticket the hotel opens with a vendor gets copied in, either by forwarding the vendor's emails to a monitored mailbox or by pulling the ticket portal's export. The ledger records opened time, priority, first response and resolution, and compares them with the SLA the extractor found. When a target is missed and the contract carries a service credit, it drafts the claim with the timestamps attached.

The rule is the one that makes it worth running. Nothing the agent finds changes a contract or a setting on its own. It writes a monthly variance note for the GM and owner, and 120 days before any renewal notice deadline it writes a renewal brief: what the contract says, what the statements and tickets show, and what to ask for.

The Framework: Read, Measure, Price, Renegotiate

Read first. The register is where most hotels find their first surprise, before any statement is parsed. Look for three blanks in particular: no written support response time, no limit on reserves, and no notice before a processor change. Each blank is a term the vendor can set later.

Measure for three months before acting. One month's effective rate swings with the card mix, especially in a month with a big group deposit or a wave of OTA virtual cards. Three months gives you a baseline you can defend, and three months of ticket timestamps tells you whether support is a pattern or a bad week.

Price the gap in dollars. A markup 40 basis points above what comparable integrated setups charge is a number. Eleven days of average payout lag against a two-day schedule, multiplied by average daily card volume, is a number. Six missed P1 responses with a service credit clause attached is a number. Vendors respond to numbers in a way they don't to complaints.

Renegotiate at the right time, which is before the notice window closes. After it, you're asking for a favor.

The most useful answer in a contract register is often "not stated." Every blank is a term the vendor gets to set later.

One more thing on the guest side, since it lands on the same front desk. The FTC's rule on hidden fees for short-term lodging took effect on May 12, 2025, and as Bryan Cave Leighton Paisner's summary explains, it requires mandatory fees to be in the total price shown to guests. If a vendor adds a card surcharge or convenience fee to the guest folio through your booking engine, the hotel carries the disclosure obligation. Put the booking engine's fee settings in the register too.

Implementation: Standing It Up in Five Weeks

This runs on what an independent hotel already has: the contracts in someone's inbox, statement downloads from the payments portal, and the vendor support emails. No vendor has to agree to anything.

Week three is where the effort goes. Statement formats are ugly, and fee codes differ between processors and sometimes between months. Give the model a fixed list of categories (interchange, network assessment, processor markup, per-item, monthly, chargeback, PCI, other) and make it show its mapping for every new code. A person approves each new code once. After that it's automatic.

Keep the agent's access read-only. It needs the documents and the exports. It doesn't need a login that can change payment settings, and it shouldn't have one.

If you'd rather have someone else build the register and run the first three months, our technology audit and roadmap work starts with exactly these documents. The contract negotiation piece covers what to ask for once the gaps are on paper, and the vendor consolidation playbook covers what to do with tools the register shows you don't need.

Across the ten hotels I've owned, the contract that cost the most was never the one with the highest sticker price. It was the one whose renewal date nobody had written down.

Frequently Asked Questions