Hotel Non-Refundable Rate: The 2026 Break-Even Discount Math
A 60-room city hotel we reviewed this year was selling a non-refundable rate at 15% below its flexible rate, on every date, all year. The rate plan was derived from BAR, so it moved automatically and nobody touched it. It produced 1,900 room nights in twelve months and everyone treated it as a win, because the bookings were guaranteed and the cash arrived early.
Run the arithmetic and the picture changes. On the dates where that hotel sold out anyway, the 15% discount bought a guarantee worth roughly 3%. The other 12% was a gift. Across the compression nights alone it gave away close to AED 70,000 of rate for protection it did not need. The non-refundable rate was not the problem. The number attached to it was.
Table of Contents
- Why every published discount range is wrong
- What a non-refundable booking is actually worth
- The break-even discount formula
- A worked example across three date types
- Why “push non-refundable in peak season” is backwards
- The derived rate plan trap
- How to measure your three inputs
- The rate ladder that holds up in 2026
- A 30-day implementation checklist
- Frequently Asked Questions
- Conclusion
Why every published discount range is wrong
Search for guidance on the hotel non-refundable rate and you will collect a pile of ranges. One source says 5% to 15%. Another says 8% to 12%. A third says 10% to 20%, a fourth says 10% to 25%. They cannot all be right, and none of them tell you which number belongs to your hotel, on your dates, in your market.
The ranges are not made up. They are the observed behaviour of thousands of properties, most of which picked their discount by copying a competitor. What is missing is the reasoning that would let you defend the number to an owner. A non-refundable rate is not a marketing gesture. It is the purchase of an insurance policy, and you are the one paying the premium.
The premium is the discount. The cover is the revenue you would have lost if that booking had cancelled and you could not resell the room. So the only question worth asking is the one nobody answers: how much is that cover actually worth on this date?
The three questions the ranges skip
- How often does a booking like this one cancel when it sits on flexible terms?
- When it cancels, how much of the rate do you get back by reselling the night?
- How much does your flexible cancellation policy already claw back in penalties?
Answer those three and the correct discount falls out of the arithmetic. Skip them and you are guessing with your ADR.
What a non-refundable booking is actually worth
Take one room, one night, one guest. The guest is choosing between your flexible rate and your non-refundable rate on the same screen. Set aside for a moment the idea that the discount creates new demand. On an OTA listing the two rate plans sit side by side, so the majority of non-refundable bookings are the same guests who would have booked the flexible rate, taking the cheaper option because it is there. That is the base case, and it is the one your model has to survive.
If the guest books the flexible rate, you do not have the money. You have a probability. Some of the time they arrive and pay in full. Some of the time they cancel, and then two things can happen: you resell the night, or you do not. If your policy has teeth, you may also retain a penalty.
If the guest books the non-refundable rate, you have the money. All of it, discounted, and it stays yours whether they arrive or not.
The value of the non-refundable rate is the difference between certainty and that probability. It is not the whole rate. It is only the slice of the rate that a cancellation would genuinely destroy. Everything you discount beyond that slice is money you handed to a guest who was going to show up anyway.
The break-even discount formula
Three inputs, one line of arithmetic.
- c = the probability this booking cancels under flexible terms, expressed as a decimal.
- r = your recapture rate. The share of the original rate you recover by reselling the night after a cancellation. A night you resell at the same price recaptures 1.0. A night you resell at a 30% discount recaptures 0.7. A night that goes empty recaptures 0.
- f = the share of the rate your flexible policy retains as a penalty when the guest cancels late. Free cancellation until arrival means f is 0.
The maximum discount you can justify on a non-refundable rate is:
d* = c × (1 − f − r)
In words: the deepest defensible discount is the chance of cancellation multiplied by the share of the rate you would not have got back anyway. That is it. If a cancellation costs you nothing because you resell the room at full price, the guarantee is worth nothing and the correct discount is close to zero. If a cancellation means an empty room and no penalty, the guarantee is worth almost the entire cancellation risk and the discount can run deep.
Reading the formula properly
Notice what is not in it. Cash flow is not in it. Forecast accuracy is not in it. Both are real benefits, and both are the reason a hotel might accept a discount slightly above d*, but neither belongs in the rate calculation, because neither shows up in RevPAR. If you want to pay 2 points of ADR for cleaner cash flow, decide that consciously and write it down. Do not let it hide inside a number you never derived.
Notice also that f and r work in the same direction. Tightening your flexible cancellation window raises f, which shrinks d*. The order of operations matters: fix the flexible policy first, then price the non-refundable rate against it. Hotels that do it the other way round end up paying twice for the same protection.
A worked example across three date types
A 48-room independent hotel. One rate plan structure, three very different nights. Watch what the formula does to a flat 12% discount.
Night one: a compression Saturday
BAR sits at USD 260. The city has an event, the comp set is filling, and anything that cancels inside three weeks gets rebooked within days, often at a higher rate. Cancellation probability on the flexible rate is high, call it 20%, because guests who book compression dates early keep shopping. But recapture is 0.85. Free cancellation until 24 hours out, so f is 0.
d* = 0.20 × (1 − 0 − 0.85) = 0.20 × 0.15 = 3%
The hotel is discounting 12%. On a USD 260 rate that is USD 31.20 given away per booking to buy protection worth 0.20 × 0.15 × 260 = USD 7.80. Net destruction: USD 23.40 per non-refundable booking. Sell 400 of those a year on compression dates and the rate plan burns USD 9,360 of pure margin, on the exact nights where margin is easiest to hold.
Night two: a shoulder-season Wednesday
BAR at USD 190. Demand is soft but not dead. A cancellation 10 days out gets resold maybe half the time, and usually at a lower rate, so recapture lands around 0.45. Cancellation probability 18%. Policy still free until 24 hours out, so f is 0.
d* = 0.18 × (1 − 0 − 0.45) = 0.18 × 0.55 = 9.9%
Here the 12% discount is close to fair. Slightly generous, defensible, not a crisis. This is the date type the published ranges were built on, which is why they feel right often enough to survive.
Night three: a dead January Tuesday
BAR at USD 130. If a booking cancels you are not reselling that night, and you know it. Recapture 0.10. Cancellation probability 15%, lower simply because fewer speculative bookers are shopping this date. Policy free until arrival, f is 0.
d* = 0.15 × (1 − 0 − 0.10) = 0.15 × 0.90 = 13.5%
The 12% discount is under-priced. This is the night where a guaranteed booking is worth real money, and the hotel is being stingier here than on the Saturday where the guarantee is worth almost nothing. The rate plan is upside down.
What a tighter flexible policy does
Same shoulder Wednesday, but now the flexible rate carries a one-night penalty for cancellations inside 48 hours, and the average stay is three nights, so f is roughly 0.33 across late cancellations.
d* = 0.18 × (1 − 0.33 − 0.45) = 0.18 × 0.22 = 4%
The moment your flexible rate stops being free insurance for the guest, the non-refundable rate needs to give away far less to earn its place. This is the single most under-used lever in independent hotel dynamic pricing strategy, and it costs nothing to pull.
The summary table
| Date type | BAR | Cancel probability (c) | Recapture (r) | Penalty retained (f) | Max justified discount (d*) | Verdict on a flat 12% |
|---|---|---|---|---|---|---|
| Compression Saturday | USD 260 | 20% | 0.85 | 0 | 3% | Destroys USD 23.40 per booking |
| Shoulder Wednesday | USD 190 | 18% | 0.45 | 0 | 9.9% | Roughly fair |
| Dead January Tuesday | USD 130 | 15% | 0.10 | 0 | 13.5% | Too shallow, leaves demand unbought |
| Shoulder Wednesday, 48h penalty | USD 190 | 18% | 0.45 | 0.33 | 4% | Wildly over-generous |
Why “push non-refundable in peak season” is backwards
The standard advice is to lean on non-refundable rates when demand is strong, to protect yourself against the cancellations that peak dates attract. Half of that is true. Peak dates do attract more speculative bookings, and they do cancel more.
The conclusion does not follow. Cancellation frequency is only one of the three inputs, and on a compression date the other two crush it. Recapture is near 1.0, because a cancelled room on a sold-out night is resold within hours, frequently at a higher rate than the one you lost. A cancellation on a high-demand date is not a loss. It is often an upgrade to your ADR.
So the guarantee is worth the least precisely when the temptation to sell it is highest. Meanwhile the nights where a cancellation is a genuine, permanent, unrecoverable hole in your RevPAR are the soft nights, and those are the nights where most hotels quietly close their non-refundable plan because “we do not want to discount in low season.”
Read that back. Most hotels discount hardest for protection they do not need, and refuse to discount at all for protection they do need. The correct instinct is closer to the opposite of the received wisdom, and it follows the same displacement logic that governs group business. If you have not run those numbers, our walkthrough on hotel displacement analysis uses the same framework applied to group rates.
The one exception
If your peak dates are so compressed that you are already overbooking to cover cancellations, and the walk cost of getting it wrong is punishing, a guaranteed booking has a second value: it lowers the variance in your forecast and lets you overbook less aggressively. That is a real benefit, and it can justify a point or two above d*. It does not justify twelve.
The derived rate plan trap
Almost every channel manager and OTA extranet lets you build the non-refundable rate as a percentage off the flexible rate. Booking.com’s setup flow works this way, and so does every property management system worth using. It is a genuinely good feature. It means you only maintain one rate and the rest follow.
It is also how a single unexamined number metastasises across 365 nights. Derive a 15% discount once, and you have committed to a 15% discount on your busiest Saturday of the year, on the night of the city marathon, on New Year’s Eve, and on every date where you would have sold out at full price by Tuesday.
The fix is not to abandon derived pricing. The fix is to put date-level controls on top of it:
- Close the non-refundable plan on compression dates. A stop-sell on the plan, not on the room. The flexible rate keeps selling at full price.
- Run two derived tiers, not one. A shallow tier for normal demand and a deeper tier for identified soft dates, switched by pickup rather than by calendar season.
- Re-derive quarterly. Your recapture rate moves with your market. A discount set in a soft year is wrong in a strong one.
- Watch the OTA promo stack. A 12% non-refundable rate that also inherits a mobile rate and a Genius or member discount is not a 12% discount any more. It is whatever the compounded number turns out to be, and nobody in the building has computed it.
That last point is where most of the damage happens on Booking.com listings, because the discounts stack multiplicatively and the extranet does not show you the end result until the reservation lands.
The semi-flexible middle
There is a rate plan between the two that most independents never build. A semi-flexible rate: a small discount, free cancellation up to 14 or 21 days out, nothing after. It captures the guest who wants a hedge but knows their plans, and it does it at a fraction of the discount a full non-refundable rate demands. Booking.com even lets you keep some flexibility inside a non-refundable booking by automating date changes rather than refunds, which preserves the revenue while removing the reason a guest hesitates.
How to measure your three inputs
The formula is trivial. Getting honest numbers into it is the actual work, and it is why most hotels never do it.
Cancellation probability (c)
Do not use one number for the property. Cancellation behaviour splits cleanly along three lines, and your PMS already has the data:
- By channel. OTA bookings cancel substantially more than direct bookings at most independent properties. The platforms are designed for comparison shopping and guests treat a free-cancellation reservation as a shortlist entry, not a commitment.
- By lead time. A booking made 90 days out has 90 days in which plans can change. A booking made on Thursday for Saturday has almost no exposure.
- By date type. Event dates and holiday weekends attract speculative bookings. Ordinary midweek nights do not.
Pull twelve months of reservations, tag each one cancelled or stayed, and cross-tabulate by those three dimensions. You want a cancellation rate for each cell, not a headline average. The headline average is the number that hides every mistake you are about to make.
Recapture rate (r)
This is the number almost no hotel has, and it is the one that decides everything. You are asking: when a booking cancels, what fraction of that rate do I actually get back?
Approximate it honestly rather than perfectly. Take a sample of cancellations from last year. For each one, ask what your final occupancy was for that date and what the rate did afterwards.
- Sold out on the night, rate held or rose: recapture is close to 1.0, sometimes above it.
- Sold out but only after you dropped the rate: recapture is the ratio of the resale rate to the original rate.
- Finished below full occupancy: recapture is 0 for that room, because the room that went empty is the room that cancelled. There is no other way to count it.
Average within each date type. Most independent hotels who run this exercise find their recapture on soft dates is far lower than they assumed, and their recapture on compression dates is close to total. Both findings point the same direction, and both point away from the flat discount they have been running.
Penalty retention (f)
Read your own flexible cancellation policy, then read what your channel manager is actually sending to each OTA. These differ more often than anyone expects. If your extranet says free cancellation until 6pm on the day of arrival, f is 0 and your non-refundable rate is carrying the entire burden of a policy decision you never revisited.
The rate ladder that holds up in 2026
Booking windows are compressing and guests shop harder after they book than before it. A rate ladder built in 2019 is not going to hold. The structure below survives contact with 2026 demand because every rung is priced off the formula rather than off a convention.
| Rate plan | Position vs BAR | Cancellation terms | Open when | Closed when |
|---|---|---|---|---|
| Flexible (BAR) | Reference rate | Free until 48h, one-night penalty inside 48h | Always | Never |
| Semi-flexible | Minus 3% to 5% | Free until 14 to 21 days out | Always | Never |
| Non-refundable, shallow | Minus d* for normal demand, typically 4% to 7% | No refund, date change permitted | Normal pickup | Compression dates |
| Non-refundable, deep | Minus d* for soft demand, typically 10% to 14% | No refund, date change permitted | Pickup behind pace inside 30 days | Any date pacing at or above forecast |
Two things about this ladder are worth stating plainly. First, the non-refundable rate is now a demand tool rather than a permanent fixture, which means it belongs in the same conversation as your stay restrictions and gets opened and closed on the same pickup triggers. Second, the deep tier is deeper than most hotels dare to go, and that is deliberate. On a night you will not otherwise fill, a guaranteed booking at minus 14% beats an empty room at any rate you like.
Booking.com’s own guidance to partners is that adding a non-refundable plan can reduce cancellations by at least 9% and lift total bookings by around 5%. Take the direction of that claim seriously and the size of it with the caution any platform’s self-reported uplift deserves. It tells you the plan is worth having. It does not tell you what to price it at, and the platform has no incentive to help you with that part.
A 30-day implementation checklist
- Export twelve months of reservations with booking date, arrival date, channel, rate plan, status, and rate.
- Compute cancellation probability by channel, by lead-time band, and by date type. Do not stop at the property average.
- Sample 40 cancellations and estimate recapture for each. Group by date type and take the average.
- Read your flexible policy as the guest sees it on every channel. Establish f. It is probably 0.
- Tighten the flexible policy before touching the non-refundable rate. A 48-hour window with a one-night penalty is the standard starting point.
- Recompute d* for each date type using the new f.
- Rebuild the non-refundable plan as two derived tiers, shallow and deep.
- Put a stop-sell on the non-refundable plan for every identified compression date in the next 12 months.
- Check the compounded discount on each OTA, including mobile rates and loyalty programme discounts, and confirm the stacked number still sits at or below d*.
- Track non-refundable share of room nights, realised ADR by rate plan, and cancellation rate monthly. Re-derive quarterly.
If the data pull alone looks like a month of evenings, that is the honest cost of doing this properly in-house. It is also exactly the sort of work an outsourced revenue management engagement absorbs in the first two weeks, and the reason the numbers usually move before the end of the first month.
Frequently Asked Questions
What discount should a hotel non-refundable rate be?
There is no single correct figure, and any source that gives you one without asking about your recapture rate is guessing. The maximum defensible discount is your cancellation probability multiplied by the share of the rate you would not recover by reselling the night. For most independent hotels that lands between 3% and 6% on compression dates and between 10% and 14% on genuinely soft dates.
Are non-refundable rates worth it for independent hotels?
Yes, when they are priced and gated correctly. The plan earns its keep by converting a probability into cash on nights you cannot refill. It destroys value when it runs as a flat, always-open discount on dates that would have sold at full price. The plan is not the issue. The flat number is.
Should I offer non-refundable rates during peak season?
Usually not, or at least not at the same discount. On a date where a cancellation gets resold within days at the same rate or better, the guarantee you are buying is worth very little, so the discount you pay for it should be very small. Close the plan on your true compression dates and let the flexible rate carry them.
Does a non-refundable rate cannibalise my flexible rate?
Partially, yes, and you should assume it does when you build the model. On an OTA listing both plans appear on the same screen, so a share of non-refundable bookings are guests who would have paid the flexible rate. That is why the break-even calculation starts from the switcher case. Any genuinely incremental demand the plan creates is upside on top, not the justification for the discount.
How do I calculate my hotel’s recapture rate?
Sample your cancellations from the last twelve months. For each one, check what your final occupancy was on that date and what happened to the rate. If the night sold out at the same rate or higher, count recapture near 1.0. If it sold out only after you cut the rate, use the ratio of the two rates. If the night finished below full occupancy, count recapture as 0, because the empty room and the cancelled room are the same room.
Should the non-refundable rate be prepaid?
Not necessarily. Non-refundable and prepaid are two separate decisions that usually get bundled. The revenue protection comes from the terms, not the timing of the charge. Prepayment adds cash flow and removes payment risk, and it also removes a segment of guests who will not hand over money three months early. Decide it on its own merits.
Where does the non-refundable rate fit against total revenue?
A guaranteed room booking also guarantees the ancillary spend attached to it, which is a benefit the room-level formula ignores entirely. If your property carries meaningful food, beverage, spa, or parking revenue per occupied room, a cancellation costs you more than the ADR and your true d* is slightly higher than the formula suggests. We cover that arithmetic in our piece on total revenue management for hotels.
Conclusion
The hotel non-refundable rate is one of the few pricing levers where the industry has settled on a convention instead of a calculation. Everyone offers one. Almost nobody can tell you why the discount is the number it is, and the number is usually copied from a competitor who copied it from someone else. Meanwhile it runs on autopilot, derived from BAR, applying the same premium on the night you sell out as on the night you sit half empty.
Three inputs fix it: how often the booking cancels, how much you get back when it does, and what your flexible policy already retains. Multiply, and you have a discount you can defend to an owner, a board, or yourself.
If you want that arithmetic run against your own twelve months of reservations rather than an industry range, that is precisely the work we do. Get in touch and we will look at your cancellation and recapture data, rebuild the rate ladder around it, and show you what the current discount is costing on your compression dates. Month to month, no lock-in, and you keep the model either way.
