Booking recovery. The table nobody cancelled stays held until the last minute.
A no-show doesn't leave a gap in the bookings list. It leaves a table or room assigned to someone who isn't coming, and keeps it taken until it's too late to give it to anyone else.
The damage is how late you find out. A table cancelled at 6pm can be rebooked. The same table, found empty at 8:30pm, is lost twice over: the cover that never arrives, and the couple at the door who give up and leave.
The reminder goes out on its own; cancellations don't. The system asks for confirmation, reads the reply on whatever channel it comes through and records the cancellation. Cancelling a booking the customer hasn't cancelled, and offering the slot to someone else, remain decisions for a person.
There's one metric, agreed up front: the no-show rate as a share of all bookings. We measure it as it stands today over the previous thirty days, set the threshold the work has to beat, and compare afterwards.
This page covers a single piece of the system. The other pieces, and how we choose which one to start with, are on the services page.
The slot is back on sale while there's still time to sell it
What this piece buys you is notice: knowing two hours ahead that a slot has come free, instead of finding out once service has started. A booking that falls through without warning costs you twice. It ties up the table, which stays unavailable to someone who would have taken it, and it ties up the attention of staff, who keep holding it for someone who isn't coming.
Asking for confirmation by hand works, which is exactly why hardly anyone manages it every day. Twelve phone calls in the hour you needed to prepare for service is real work, and the first busy day is the day it gets skipped.
An empty table costs you the lost cover. An empty table you find out about too late also costs you the customer who could have had it.
How much you get back, and what sample that's based on
The most solid evidence for this mechanism comes from healthcare, where missed appointments have been counted for decades. A systematic review published in BMJ Open in 2016, covering 21 controlled studies and 16,076 patients, found a 15% no-show rate among people who got a notification against 21% among those who didn't: a quarter fewer.
The same authors add a practical detail: several notifications work better than one. In their words, “sending multiple notifications could improve attendance further”. That sample, though, is clinical appointments, not tables or rooms, and a patient who misses an appointment doesn't have another venue to go to instead.
For restaurants, the figure doing the rounds in Italy comes from an analysis of 212,000 bookings across 300 restaurants between January and August 2025: an average no-show rate of 12.8%, down to 5.4% where a deposit is taken, and a further 32% reduction with automatic reminders.
That analysis comes from a company that sells booking technology, and it reached the trade press as a press release, with no study published alongside it. It's still the broadest figure available for the Italian market, and these two sentences are the reason we don't treat it as a promise.
What the system does, step by step
Every booking goes through the same route, whatever channel it came in on, and every step leaves a log entry recording what happened and when.
| Step | What happens | What you get |
|---|---|---|
| Reminderbefore the booking | A message goes out on the channel the customer booked through, with the date, time, number of people and a single question: can you confirm? |
A reply instead of silence, because it's a yes-or-no question that takes one word to answer. |
| Reading the replyon any channel | Confirmations, cancellations and time changes are recognised even when people word them their own way, and the booking is updated. |
The diary gets updated either way, even when the customer replies in their own words instead of tapping the button provided. |
| Unconfirmed listwho hasn't replied | Anyone who doesn't reply stays booked and goes onto a separate list, sorted by time, which a person checks before service. |
Before service you know exactly who hasn't confirmed, without relying on anyone's memory. |
| Slot back on offeras soon as a cancellation comes in | A cancellation frees up the slot in the diary and flags it straight away, along with the time by which it needs to be rebooked to be worth it. |
The table comes back to you while it's still worth something, with a note of how long you have to rebook it. |
| Rebookingdecided by a person | The system suggests who to contact from that day's turned-down enquiries and waiting list; a person sends the offer. |
A ready-made list of who to call first, and the final say on the offer stays with you. |
No customer cancelled by mistake, and the rule that guarantees it
A customer cancelled by mistake doesn't come back, and this system is built so that can't happen. Any system reading human replies will make mistakes, and its two kinds of mistake don't cost the same. Reading a cancellation as a confirmation leaves a table held, which is the problem you started with. Reading silence or an ambiguous message as a cancellation removes a customer who would have turned up.
That's why the system is deliberately one-sided: only an explicit cancellation cancels anything. Silence leaves the booking valid. A reply the model can't classify is shown to a person instead of being interpreted, and stays marked as unread until someone looks at it.
Every row records whether it was handled by the model or passed to a person, so the share of ambiguous cases is counted instead of disappearing. If it rises, you see the problem before it turns into an angry customer on the phone.
The engine is built and has been tested on its test branches, with no real customers involved. That's a functional check, and we describe it as exactly that.
What goes out on its own, and what waits for a person
The reminder, reading the reply, recording the cancellation and flagging the freed-up slot all happen on their own. They're all actions on a booking the customer has already made: they repeat an agreed fact rather than adding a new one.
Anything that commits the business waits for a person: cancelling a booking nobody has cancelled, offering a freed-up slot to another customer, applying any financial terms, taking or keeping a deposit. The reasoning behind that line is on our page about the principles we build by.
The owner turns the automatic part on and off channel by channel, with a switch, without anyone touching the system. Some businesses want the reminder to go out every time; others only switch it on for busy days.
Reminding someone of a booking repeats a decision that's already been made. Cancelling it makes a new one.
The message says it comes from a system
From 2 August 2026, Article 50 of the EU AI Act applies, and for anyone using a system like this it means a duty to inform: people have to know they're dealing with a system, and they have to know before the exchange starts, not afterwards.
A reminder invites a reply, so it opens a conversation. It says so in one line, within the message itself, and that disclosure covers the whole exchange. What this really means for a small business is covered in the AI Act and Italian SMEs, and the systems we use are listed on the AI transparency page.
The disclosure makes little difference to the result, because people who've booked expect to hear back. What it does is remove the only serious risk the feature carries: a customer who believes they were writing to the owner.
Same mechanism, different name in every trade
The mechanism is identical; what gets lost isn't. It's worth looking at your own case, because that's where you see what it's costing you today.
| Sector | What gets lost | Where you see it |
|---|---|---|
| Hospitalityaccommodation and events | Two bookings for the day after tomorrow that have never confirmed: if they don't turn up, the rooms sit empty with no chance to rebook them. |
|
| Restaurantsand bars | The table for four, booked, never cancelled and never taken up, held for forty minutes while people are waiting at the door. |
|
| Appointment-based businessespractices, garages, personal services | The hour set aside for someone who doesn't turn up, which could have gone to one of the customers pushed back to next week. |
What this piece doesn't do
It doesn't replace your booking software, channel manager or diary. It works on what currently happens outside those systems, where confirmations are asked for over the phone and cancellations arrive on a channel the diary can't read.
It doesn't bring in new bookings. It works on the ones you already have, which is why the return is calculated on your current volume, without depending on a campaign that has yet to work. Enquiries that never even become bookings today are a different problem, and that's the job of Inbox AI, the first reply to every enquiry.
It doesn't introduce deposits for you. It's the lever with the biggest measured effect, and it remains the owner's commercial decision, with consequences for customer relationships that no tool can weigh up.
Questions and answers
Is a reminder enough to cut no-shows?
It makes a difference, and how much depends on the business. The most solid evidence comes from healthcare: a systematic review published in BMJ Open in 2016, covering 21 controlled studies and 16,076 patients, found a 15% no-show rate among people who got a notification against 21% among those who didn't, a quarter fewer.
In Italian restaurants, an analysis of 212,000 bookings credits automatic reminders with a 32% reduction. Neither sample is a restaurant or hotel like yours, so the direction holds, and the size of the effect is measured on site, over the previous thirty days.
What happens if the customer doesn't reply to the reminder?
Nothing automatic, and that's by design. Silence is never treated as a cancellation: the booking stays valid and goes onto a list of unconfirmed bookings that a person checks before service.
The same goes for a reply the model can't classify, which is shown to someone rather than interpreted. Cancelling a booked table because of a misread reply costs more than the empty table you were trying to avoid.
Do we need deposits? Is that something the software does?
No, a deposit is a commercial decision, and it's worth saying so because in the same analysis of 212,000 bookings it's the lever with the biggest effect: no-shows fall from 12.8% to 5.4% where there's a deposit or card pre-authorisation, averaging €20 per person.
The system can request it, record it and remind people about it. The decision to introduce it stays with the owner, and it affects the relationship with customers in ways no tool can judge on their behalf.
Do we have to change our booking software?
No. Your booking software, channel manager and diary stay where they are. The work goes into what currently happens outside those systems: asking for confirmation, reading the reply on whatever channel it comes through, recording the cancellation and flagging the freed-up slot while there's still time to rebook it.
We only suggest replacing a tool if the analysis shows it's the bottleneck, never out of preference.
How do you measure whether it's working?
By the no-show rate as a share of all bookings, agreed as the metric before we start. We measure the current figure over the previous thirty days, set the threshold the work has to beat not to count as a failure, and compare afterwards.
The second number is how many freed-up slots were actually rebooked, which is the most honest count because it counts people who turned up, not messages sent.
Notes on sources
- The figure of a quarter fewer no-shows comes from Robotham et al., Using digital notifications to improve attendance in clinic, BMJ Open, 2016: a systematic review and meta-analysis of 21 controlled studies, with 8,345 patients who got a notification and 7,731 who didn't. It measures healthcare appointments, not tables or rooms. We point this out because the sample isn't the reader's.
- The figures of a 12.8% no-show rate, 5.4% with deposits and a 32% reduction with reminders come from an analysis of 212,000 bookings across 300 Italian restaurants, January to August 2025, reported in the trade press. It comes from a company that sells booking technology, and no study is available alongside the press release: we include it because it's the largest Italian sample available, and we flag the limitation.
- This page doesn't report results achieved for a client, because this piece hasn't yet been delivered to a client. The tests mentioned are functional checks run in a test environment, which is why you won't find a recovered revenue figure here.
- We don't publish a figure for how many freed-up slots can be rebooked: it depends on how much notice the cancellation gives, the day of the week and demand at the venue. The number is worked out from your previous months, before any quote.
The other pieces in this group
Missed bookings and customers who don't come backFifteen minutes, with your case in front of us.
How many bookings were no-shows last month? And how many of the cancellations were rebooked? If the answer is that nobody knows, that's already the most useful thing to find out, and it can be counted within thirty days. In fifteen minutes on the phone we'll look at it together and tell you where it makes sense to start, even if we never end up working together.
You'll speak to Mattia Esposito, who then builds the system: there's no salesperson in between. If you'd rather measure things yourself before talking, the Diagnostico (in Italian) is twenty questions and five minutes.