Overbooking policy: the math, the walk order, and the traps

Every hotel overbooks or pays for not overbooking. No-shows and last-minute cancellations are statistically certain — the only choice is whether you price that certainty deliberately or eat it as empty rooms sold to nobody. A real overbooking policy has three parts: a level set by arithmetic, a day-of-arrival playbook owned by a named person, and a feedback loop. Most hotels have none of the three; here's all of them.

Why overbook at all

Suppose your Tuesdays average six no-shows and late cancels. Refusing to overbook doesn't make that risk disappear — it converts it into six unsold rooms at 100% loss, every Tuesday. Overbooking is simply selling against the no-shows you can predict, and accepting a small, priced risk of walking a guest in exchange. The mistake isn't overbooking; it's overbooking without the math.

The optimal level: newsvendor logic in hotel terms

Two costs trade off. An empty room costs you its lost contribution — say ADR $189 minus $22 variable cost ≈ $167. A walk costs far more — say $450 all-in (built honestly below). Each additional overbooked room pays off only if enough no-shows materialize to absorb it:

Keep overbooking room b while P(no-shows ≥ b) × $167 > P(no-shows < b) × $450 — i.e. while you're at least $450 ÷ ($450 + $167) ≈ 73% confident that b no-shows will actually occur.

Worked example: Tuesday history over the last year says no-shows run mean 6, spread roughly 2. You're ~84% confident of at least 4 no-shows, only ~69% confident of at least 5. The threshold is 73%, so overbook by 4 — deliberately below the average, because a walk costs nearly three times an empty room. That asymmetry is the whole policy: the more a walk truly costs you, the further below your average no-show count you sit. Run the same calculation per day of week and season; Tuesday's number is not Saturday's.

Cost the walk honestly

The $450 isn't a guess — build it: the room you buy at a nearby (comparable or better) hotel at walk-in rates, transport both ways, the apology amenity or rate adjustment, and the part hotels omit — goodwill: the review you'll absorb, the loyalty points compensation, and for a repeat corporate guest, real lifetime-value damage. If elites or contracted corporate guests could plausibly be reached by a walk, their number is much higher than $450 — which is exactly why the walk order below exists. Understating walk cost is how spreadsheets talk hotels into overbooking levels their front desks pay for.

Day-of-arrival execution

1. A named owner per shift. The overbook position is reviewed at the morning huddle and handed off explicitly at shift change. "The desk will figure it out" is not a policy.
2. The walk order, decided in daylight: volunteers first (a comped future night finds takers surprisingly often), then one-night stays, then third-party prepaid singles. Never loyalty elites, group-block guests, long stays, direct repeat guests, or anyone arriving with a problem already (late flight, accessibility needs).
3. The walk kit, prearranged: a standing rate agreement with two nearby comparable-or-better hotels, transport arranged and paid, a recovery script ("we've reserved and paid for your room at X, here's your transfer, and tomorrow night here is on us"), and a follow-up owner the next morning. A well-executed walk recovers most guests; an improvised one loses them and their review.

Set it per rate and room type, not just house level

A house-level number hides two traps. Room types don't substitute downward: an overbooked suite guest can't be "upgraded" into a standard king, so overbook entry categories and keep premium types at or under capacity. And rate types behave differently: prepaid reservations barely no-show; flexible OTA bookings no-show and late-cancel at multiples of direct rates. Weight your no-show forecast by the mix on the books — the forecasting guide's cancellation-and-no-show adjustments are the same discipline pointed at arrival day.

Monitor it like a forecast

Three numbers keep the policy honest: walks per 1,000 arrivals (industry practice keeps this in low single digits; zero forever means you're under-overbooking and eating empty rooms), no-show forecast accuracy (track it like any forecast — if Tuesday's mean drifts from 6 to 3, yesterday's level is today's walk machine), and a walk-cost log — what each walk actually cost, fed back into the ratio. Review monthly in the same sitting as forecast accuracy.

The failure modes

• Copying someone else's number. "We always take 5 over" imported from a previous property imports their no-show behavior, not yours.
• Flat levels on compression nights. On a citywide sellout the walk hotel is also full, transport is scarce, and every walked guest is someone who paid peak rate — walk cost triples exactly when your overbook level should be at its most conservative.
• Nobody owning 6 p.m. The level was set by someone who went home at five; the walk decision lands on the newest desk agent. Ownership is the policy.
• Overbooking what you can't substitute. Suites, accessible rooms, connecting rooms — if there's no acceptable swap, the level is zero.

Put it in writing

The complete overbooking SOP — the level-setting worksheet, walk order, walk kit checklist, recovery scripts, and the monthly review agenda — ships in the RM Playbook, and the no-show math plugs straight into the free calculators. The daily overbook-position check belongs in the morning routine, right after the pickup review.