In this article
7:16pm. The table for four that was booked for 7:00 hasn't called and still isn't there. By the door, a family of four who walked in half an hour ago without a reservation is starting to get restless. You're standing at the host stand, and in the next few minutes you have to make a call your reservation system will never make for you: how much longer do you hold that table? Most restaurants — and nearly every reservation platform that offers guidance on it — answer with the same number: 15 minutes. Fixed, for every table, every night. That number isn't a bad guess. It's just the wrong question to answer with a single answer.
This isn't about fewer no-shows or smarter overbooking
Good tools already exist for cutting no-shows before they happen — reminders, deposits, confirmation links — and for working out in advance how many reservations you can safely book above capacity, based on your own no-show rate. This article is about what happens after both of those decisions have already been made: the deposit wasn't taken, the reminder didn't work, and the guest simply isn't there. That's no longer a planning question. It's a decision you make in real time, under whatever pressure exists at the door at that exact moment — which is exactly why one fixed number can never answer it well.
Why "always 15 minutes" is a compromise nobody actually chose
Fifteen minutes isn't the number you keep running into by accident. Reservation platforms and POS systems cite it as the go-to default, and most venues simply adopt it because it sounds reasonable: long enough to cover traffic, a parking problem or a misremembered time, short enough not to wait forever. The trouble is that "reasonable" here is the average of two situations that have nothing to do with each other.
| Situation | What 15 minutes actually means there | What it should be |
|---|---|---|
| Quiet Tuesday, nobody's waiting on that table | An arbitrary cutoff with no reason behind it — you're giving the table away to nobody | Generous — 30 to 45 minutes is fine, there's nothing to lose |
| Packed Friday, three parties queued at the door | Fifteen minutes in which a waiting family can decide to eat somewhere else | Short — 10 to 15 minutes, every extra minute costs a real chance |
Research into restaurant no-shows generally points to a rate somewhere between 2% and 7% of reservations, with a sharply worse outlier in the UK, where roughly a quarter of bookings are estimated to be no-shows — costing the sector there billions a year on its own. Globally, the annual cost of no-shows to restaurants is commonly put at around $16 billion. Those numbers explain why the 15-minute rule exists in the first place: you need a cutoff somewhere. They don't explain why that cutoff should be identical for every table, every night.
Four scenarios, four very different answers
What's actually at stake in the moment comes down to two things: is anyone waiting, and what is this reservation worth?
There's nothing to lose by waiting. Releasing early gives an annoyed guest nothing in return.
Every extra minute is a real chance the party actually standing at the door leaves. Fast release almost always wins here.
The relationship with this guest is worth more than the table. A short call costs nothing and prevents the worst version of this mistake.
The hardest call on the list: a large reservation is worth a short delay to the waiting guests too — but not an unlimited one.
A framework, not a measured statistic — meant to replace the fixed 15-minute rule with a judgment call that fits what's actually happening. Plug your own numbers into the calculator further down for a concrete number of minutes.
The two things that actually decide it
Underneath the framework above sit just two variables, and both are readable at any point in the night. The first is replacement demand: is there someone right now who could take that table if you released it? If the answer is no, waiting costs you literally nothing — there's nobody to seat there anyway. If the answer is yes, and several parties are already queued and have themselves been waiting a while, every extra minute is a real chance one of them walks. The second variable is what's at stake: party size times this specific reservation's average spend. A 2-top ordering coffee and a sandwich is a small bet either way. A table of eight on a tasting menu isn't — and it deserves more patience when nothing else can immediately replace it, not less.
| Variable | What it does to your grace period |
|---|---|
| Number of parties currently waiting | Each additional waiting party shortens the reasonable hold time — there's more to lose per minute |
| How long the longest-waiting party has already waited | The longer they've waited, the higher the risk they leave the queue themselves — that speeds the clock up further |
| Party size × average spend | Higher stakes justify more patience, unless the queue makes that choice impossible |
| Regular guest or first visit | For a regular, the relationship carries more weight than an unfamiliar name on the list |
That's the whole model: not a fixed number, but the output of what actually matters at that moment. The same restaurant can release a 2-top after 10 minutes and hold a table of eight for 40 minutes on the exact same night — and both calls can be the right one.
Run the numbers: a table of eight on a €95-per-head tasting menu already represents roughly €760 in tonight's revenue alone, before counting whatever that group is worth as future regulars. Handing those same eight seats to a party of two who happened to walk by isn't an even trade — unless a party of comparable size is also waiting. That's exactly why "large reservation" and "long grace period" belong together, for as long as nothing on the floor can replace that seat at equal value.
Why one phone call changes everything
The number that actually shifts this whole calculation isn't a minute count — it's whether the guest has already made contact. Someone who calls to say they're stuck in traffic hands you two things at once: certainty that they're coming, and a realistic time to expect them. That's fundamentally different from silence, where all you can do is guess between "running late" and "not coming at all." A short, friendly call or text from your side right as the reservation time passes — "we're still holding your table, everything okay?" — costs a minute and turns a guess into a fact. Restaurants that run active waitlists commonly report that a basic text-notification system recovers roughly 60 to 70% of guests who would otherwise have walked away; the same logic runs in both directions. A guest who knows you're still waiting is more likely to stay put a little longer too.
There's also a real difference between a guest who's late and a guest who isn't coming, and it's bigger than most fixed rules allow for. A misremembered time, traffic, a kid who wouldn't get in the car — all of those are situations where the guest is genuinely on the way, just later than planned. A reservation made for 7pm that still hasn't shown by 9pm with zero contact is a very different scenario from that same reservation at 7:18. The fixed 15-minute rule treats both identically. The model above — with the waitlist as an output, not a cause — doesn't have to.
Live: the fixed rule versus your actual night
Drag the number of waiting parties and watch the recommended minutes change
Calculated for a table of 4 at a €35 average spend, with the longest-waiting party already waiting 12 minutes — the same numbers as the calculator below.
Work out your own grace period
Enter tonight's actual situation — the party, how many guests are already waiting and for how long, and whether this is a regular — and see the recommended number of minutes instantly, along with what's at stake if you get it wrong in either direction.
Grace-period calculator
Every figure is illustrative — plug in your own numbers
At zero waiting parties, the model pulls the grace period toward the generous end — there's nothing to lose. As waiting parties stack up, or the longest wait stretches on, the recommendation steps down toward the short end. A regular guest or a large reservation pulls it back up a little, but never past what the queue can reasonably absorb at that moment.
What to actually say at the host stand
The model above tells you when. The most uncomfortable five seconds of the night are still what you say in the moment — to the waiting guests, and later to the latecomer who does show up. Three situations come up over and over.
You release the table to the waiting party
Don't say "they never showed up" — that puts a judgement on a guest the waiting family doesn't know and doesn't need to know about. A simple "your table's ready, right this way" is enough. No explanation needed, no apology, no delay to a moment that's finally good news for them.
The latecomer arrives after the release
This is the moment most owners dread, and it's almost never as bad as it feels. "I'm so sorry, we had to release your table, but I'll have you seated within [X] minutes" — followed by an honest effort to make that true — repairs the night nine times out of ten. What actually breaks the relationship is getting defensive or putting the blame back on the guest, even when it belongs there.
You call ahead to check
Keep it short and blame-free: "Good evening, we had you down for 7 — are you still on your way?" That one line costs the guest nothing to answer and gives you the one piece of information the model above can't guess at.
How this fits with what you already do
This week: agree with your team on a default when nobody's waiting (generous, around 30 minutes) and a default when the waitlist is full (short, around 10 to 12 minutes) — so nobody at the host stand has to rely on gut feel alone.
This month: connect this to what you already run. The reminders and deposits you already use lower how often you ever have to make this call in the first place; the model above decides what you do on the nights it happens anyway. If you run an active waitlist, feed it in explicitly as an input — how many parties are waiting, and how long.
This quarter: if your reservation platform enforces one fixed hold time, check whether it can flex by time of night or party size. Most systems allow it, but it's rarely switched on by default.
The bottom line
Fifteen minutes isn't a bad number — it's just one number standing in for two completely different nights. A quiet Tuesday deserves patience because there's nothing to lose; a packed Friday deserves speed because every minute has a party at the door who'd genuinely rather stay. Treat those two situations separately instead of averaging them, and the call at the host stand doesn't get easier — but it finally gets right.