Most waitlist statistics on the web cannot be checked. They quote a number, credit a “study”, and link to another article that quotes the same number. When you follow the chain far enough it usually ends at a vendor blog post with no population, no sample size and no date.
We went looking for numbers we could stand behind. Twenty-three candidates went in; three came out. The rest failed for one of two reasons: no traceable primary source, or a population that does not describe a full-service restaurant with a line at the door — fast food drive-through data, or European fine dining, dressed up as a general benchmark.
The three that survived
| Benchmark | Population | Sample | Collected |
|---|---|---|---|
| 20 minutes on a waitlist before a guest cancels; 9 minutes for the guests who got a table | US full-service restaurant cohorts on one platform (~127,000 locations) | Observed platform behaviour, not a survey | Q3 2024, year on year |
| 42% of diners avoid a restaurant with a wait over 30 minutes, against 41% for bad Google reviews | US diners | 1,500 respondents, ±3% | 15–25 October 2024 |
| 8 in 10 guests want to get in and out without waiting | US consumers who ate at a tableservice restaurant in the last six months | Sample size not published by the association | 2025 report |
1. Twenty minutes to cancel, nine minutes to a table
Toast’s Q3 2024 Restaurant Trends Report found that guests placed on a waitlist stayed around twenty minutes on average before cancelling — three minutes longer than the same quarter in 2023 — while the guests who were successfully seated had waited an average of nine.
This is the only figure of the three that is observed behaviour rather than a stated preference, which makes it the most useful and the most specific: it describes restaurants on that platform, not restaurants in general.
Conflict of interest, stated plainly: Toast sells a waitlist and table product that competes with ours. The data is platform data with a declared base, and we cite it because the base is declared — but you should know who published it, and so should anyone quoting this page.
The operational reading is not “guests will wait twenty minutes”. It is that the guests who leave and the guests who stay have very different clocks, so a single average wait time hides the number that matters. Split them: how long did the parties who walked wait, and how long did the parties you seated wait?
2. A thirty-minute wait costs more than a bad review
TouchBistro, with The Harris Poll, surveyed 1,500 US diners between 15 and 25 October 2024, with a margin of error of about three points. Asked what would keep them away from a restaurant, the answers ranked: negative feedback from a friend 49%, a health inspection notice 48%, a wait longer than thirty minutes 42%, bad reviews on Google 41%, and a service fee 39%.
Two honest caveats. The wait and the reviews are four points and one point apart respectively from their neighbours, which is inside the margin of error — so this is not a strict ranking, it is a group of comparable deterrents. And it is a stated preference: what people say drives them away is not always what does.
What survives both caveats is the comparison itself. Restaurants spend real money managing their review profile and, usually, nothing at all managing the wait — and diners put the two in the same band.
3. Eight in ten want no wait at all
The National Restaurant Association’s State of the Restaurant Industry 2025 reports, on page 30, that roughly eight in ten guests want to get in and out without waiting. The population is US consumers who ate at a tableservice restaurant in the previous six months. The association does not publish the sample size for that consumer survey, and we are not going to pretend otherwise.
Read it as a statement about expectations, not about behaviour: people want no wait, and they still queue. It is the reason a quoted wait a guest can trust beats a shorter wait they were not told about.
The numbers we are not publishing yet
We run waitlists for restaurants, so the obvious move would be to publish our own averages. We are not going to, yet, and it is worth saying why in public.
Our platform data currently describes too few restaurants to generalise from. A benchmark built on a handful of rooms is not industry data — it is one operator’s Tuesday, with a chart around it. The first person to check the method would be right to throw it out, and to distrust everything else on the site afterwards.
So we wrote the gate down before we had the data, which is the only order that means anything:
- at least 30 distinct restaurants, and
- at least 10,000 completed waitlist entries — parties that were seated, served, marked as no-shows or cancelled — and
- at least 90 days of history.
Below that, no first-party number gets published here. Above it, everything we publish will be aggregate: no restaurant identified, no guest data, nothing that could be traced back to a room or a person. This page is where those numbers will live when they exist, next to the third-party ones, under the same rules about population and sample size.
How to build your own version of this table
Industry benchmarks tell you which questions matter. Your own numbers tell you what to do on Friday. Three are worth measuring from the first week:
- The wait of guests who left before being seated, separated from the wait of guests you seated. One number is a lost cover; the other is a satisfied one. Averaging them together is how a room convinces itself the line is fine.
- Quote accuracy: the share of parties seated more than five minutes after the time you promised. This is the number guests actually experience, and the wait-time quoting guide has the method.
- Table turn time by service period, because a wait quote built on last month’s average turn is wrong on both of your busiest nights. The table turn time guide covers how to measure it without a stopwatch.
Track those three for a month and you will have something no report can give you: a benchmark whose population is your own dining room. The waitlist KPI guide shows how to keep them on one page your team reads.