Airbnb guest refund math: how much to comp when a stay goes wrong

The boiler died on a January night, in the middle of a three-night stay, with the outside temperature at minus four. The guest messaged me at 22:40: no hot water, and now no heat. I spent twenty minutes pretending it might be the pilot light before I accepted the truth — the unit was uninhabitable until a plumber could come in the morning, and the guest needed somewhere warm to sleep. The question every host fumbles in that moment is the same one: how much do I owe them? Most of us guess, guess low, and then watch the platform impose a bigger number anyway.

This post is the formula I use now instead of guessing. A severity factor, four worked scenarios with real dollar figures, and the part nobody tells you — what Airbnb, Booking.com, and Vrbo do to your payout if your first offer is a lowball.

The formula that replaces guessing

Every mid-stay problem reduces to three numbers you already know:

Refund = nightly rate × affected nights × severity factor

The nightly rate is what the guest actually paid per night, not your list price — pull it from the payout breakdown, after any length-of-stay discount. Affected nights is how many of the booked nights the problem touched, and it can be fractional: a fault that ran from 8 a.m. to 1 p.m. on one day is roughly a quarter of a night, not a whole one. The severity factor is the only judgment call, and the next section pins it down.

The reason to write it as a formula is not precision theatre. It is that a number you can defend in one sentence — "two nights, half the value gone, here's 50% of two nights" — survives contact with an angry guest and a platform mediator. A number you pulled from your gut does not. When Airbnb asks you to justify a refund offer, "nightly rate times affected nights times an impact factor" is a sentence that ends the conversation. "It felt about right" invites a counter.

The severity factor: putting a number on "how bad"

Severity is the share of the stay's value the problem destroyed. Here is the scale I use, calibrated over a few hundred stays and a few painful escalations.

The single mistake hosts make here is treating climate control as an amenity. In a 33°C heatwave or a sub-zero week, heating and cooling are not nice-to-haves — they are the difference between a habitable unit and a non-habitable one. A broken AC in October is a minor annoyance. The same fault in August is a major one. The calendar moves the factor, not just the fault.

The second mistake is anchoring the factor to your own embarrassment instead of the guest's lost value. You feel terrible that the dishwasher broke; the guest barely noticed because they ate out every night. Pay for what they lost, not for how bad you feel.

Four scenarios, four payouts

Numbers make this concrete. Here are four real shapes of problem, run through the formula.

Scenario one — the bulb. A burnt-out bulb and an empty coffee-filter box. Severity zero. You do not offer cash for this, because offering cash trains the guest to expect cash, and it signals that a trivial thing was actually a big thing. You apologise, you send the cleaner with a bulb and filters, and you move on. The right response to a 0% problem is operational, not financial.

Scenario two — the cold shower. Hot water out from 8 a.m., plumber done by 1 p.m. The formula says a quarter-night at 25% — roughly $9. Nobody feels made whole by $9, so this is where you round up to the relationship, not down to the math. A $40 credit, or "coffee and breakfast on me, here's a $40 voucher for the café downstairs," costs you almost nothing and buys back the review. The formula sets the floor; goodwill sets the ceiling.

Scenario three — the heatwave. AC dead for two of three nights, 33°C outside, fixed on day three. Habitability is genuinely degraded, so severity is major — 55%. Two nights × $185 × 0.55 ≈ $204. I would offer $200 plus the $25 box fan I couriered over (handled separately, as an expense, not folded into the refund). Offer it before the guest asks. A proactive $200 reads as a host who has their act together; the same $200 dragged out of you after a complaint reads as the minimum you could get away with.

Scenario four — the boiler. No heat in January, can't fix until morning, guest cannot stay the night. The two remaining nights are a 100% loss: 2 × $160 = $320 refunded. Then there is the part that surprises hosts — if the guest rebooks a hotel at $210 against your $160, the platform may make you cover the $50/night gap for the nights they had booked with you. That is another $100 of exposure on top of the $320. Total: about $420 on a $480 booking. The lesson buried in scenario four: when a unit goes uninhabitable, refund the unused nights in full immediately and voluntarily, because the alternative is the platform doing it for you and adding the rebooking gap on top.

Refund vs comp vs future-stay credit

Three instruments, three correct uses.

1. Cash refund — for anything that destroyed real value. A refund is the only instrument the platforms recognise and the only one that defuses an escrow dispute. Use it for moderate-and-up severity.
2. Comp — a small in-kind gift for minor and cosmetic problems: a dinner voucher, a bottle of wine, a late checkout for free. Comps preserve the review without setting a cash precedent. Use them at the bottom of the scale, where a cash refund would feel disproportionate in both directions.
3. Future-stay credit — "20% off your next booking." This works only for direct and repeat guests. To a one-time platform guest from another country who will never return, a future credit is worth exactly zero, and offering it reads as a dodge. If your booking came through Airbnb or Booking.com, do not reach for the credit; it is the instrument that says "I don't want to pay you."

The decision rule: match the instrument to whether the guest lost money or lost goodwill. Lost money gets cash. Lost goodwill gets a comp. Nobody on a platform booking gets a credit.

What each platform does if you lowball

This is the section that pays for the article. The platforms are not neutral referees — each has machinery that activates when a guest is unhappy and you are stingy, and that machinery almost always lands on a bigger number than the one you would have offered.

Airbnb. The guest opens a Resolution Center request or reports a travel issue under the Rebooking and Refund Policy. They have 72 hours from noticing the problem to file. You then have 72 hours to respond to a money request before the guest can ask Airbnb to step in. Once Airbnb steps in on a genuine issue — no heat, no AC in extreme weather, not clean, materially not as described, unsafe — Airbnb can refund the guest from your payout and, separately, rebook them somewhere comparable and bill you the difference. The imposed number includes rebooking costs you never control. Airbnb's version of scenario four is almost always more expensive than yours.

Booking.com. There is no escrow Resolution Center holding the money, but there is a lever that is arguably sharper: the review goes live fast and ranks your property in a crowded market, and Booking.com customer service will lean on you to refund to keep a guest from posting a one-star. You are often the merchant of record on the card, which means the guest's fallback is a chargeback — a "service not rendered" dispute that costs you the stay total plus a processing fee regardless of who wins. On Booking.com the pressure is reputational and financial rather than automated, but it is not lighter.

Vrbo. Sits between the two. The Book with Confidence Guarantee can reimburse guests for specific failures and then pursue the host, on a slower and more case-by-case basis than Airbnb's policy, but with the same end state: a third party deciding your number for you.

The throughline across all three: your first offer is the cheapest offer you will ever make. Every hour and every dollar you hold back raises the odds that someone else sets the figure, and they set it higher.

The escalation tax

I learned this the expensive way. A guest's AC failed on a four-night $180 stay during a hot spell. I offered $20 — a number anchored to my irritation, not to the formula. The guest declined, escalated, and Airbnb refunded 50% of the entire stay: $360 from my payout. Then the review landed: three stars, "host was slow to fix a serious problem." The formula would have told me to offer around $200 the same evening. I would have paid less, kept a five-star, and kept the Superhost-qualifying review streak intact.

That gap — $360 imposed versus $200 voluntary, plus a review that cost more than either — is the escalation tax. It is the premium you pay for treating a guest's lost value as a negotiation instead of an arithmetic problem. The tax is not on the refund. It is on the delay and the lowball.

One opinionated take

Hosts treat a mid-stay refund as a loss to be minimised, and that framing is exactly why they overpay. The refund is not the cost. The cost is the imposed number plus the review plus the hour you spent negotiating a figure you could have computed in thirty seconds. Run the formula the moment the problem lands, offer the result before the guest asks, and round up to the relationship when the math comes out cold. You will pay less cash and keep more stars than the host who fought it. The cheapest refund is the fast, fair one you offered first — and if you want the rest of your operation to run on numbers instead of nerves, start where the calendar and the money are sorted in one place.