Skip to content

9days

How to reduce no-shows: AI booking and reminders for a salon, dental office or contractor

How to reduce no-shows when you take appointments: book inside the first conversation, keep the date close, and remind before the day. What 51,529 appointments say about booking too far ahead, what one reminder was worth in a randomised trial and in a 2026 review of ten, how to test an AI appointment booking tool, and whether it is worth it. For a roofer, a salon, a dental or accounting office, or a real estate office.

Excelsior Mind23 min readUpdated

We help local business owners work out how to reduce no-shows: book an appointment inside the first conversation, keep the date close and send a reminder before the day. This guide shows how to reduce no-shows with AI appointment booking: what the research says about how far ahead a time is booked and what a reminder does, how to test a tool, whether it is worth automating, and what to do first this week.

It is written for owners who take appointments: a roofer or plumber with estimate visits, a salon, spa or gym, a dental or accounting office, a real estate office.

An invented example shows how an appointment is lost without anyone doing anything wrong. On a Sunday evening in June a homeowner in Wappingers Falls messages a landscaper about a patio. The owner is at dinner, sees it at nine, and does the right thing: he calls back first thing on Monday and has a good conversation.

Then the two of them look for a time. His Tuesday is gone. Wednesday the homeowner is away. Thursday he has a job. The first slot that works cleanly for both is the Wednesday of the following week, half past six. He writes it down, and the homeowner sounds delighted.

Nine days after that Monday call he drives out, parks at the curb, and waits twenty minutes.

Nothing in that story is a failure of service. He answered, he was pleasant and he was flexible, and he lost the evening anyway. In the schedule it is logged as an estimate visit, and if anyone counts visits this month, it will be counted.

In short

  • An appointment is not an outcome. It is a prediction, and the further away you put it the worse the prediction gets. In 51,529 appointments at one university eye clinic, the no-show rate at a lead time of six months was more than four times the rate at nought to two weeks.
  • One reminder, sent 72 hours ahead, moved attendance from 80.5% to 87.5% in a randomised trial of 1,848 people, and a 2026 review of ten trials found attendance 11 percent higher with reminders. A text did the job as well as a phone call, at roughly two thirds of the cost.
  • A calendar invitation is a different object from a text message that mentions Thursday. One of them puts an event with an alarm on the other person's phone and tells you whether they accepted it. The other one hopes.

How to reduce no-shows: three moves

To reduce no-shows, book the appointment inside the first conversation, keep the date as close as the calendar allows, and send a reminder before the day with an easy way to move it. The first two follow from how lead time affects attendance in one large study, and the third has randomised trials behind it, including a 2026 review of ten of them.

None of that research comes from your trade, so it shows a direction, not your own rate. The sections below give each move its evidence, then what an AI appointment booking tool does about it, and how to tell a good one from a poor one.

Why does the gap to the appointment hide the cost of no-shows?

The gap hides the cost of no-shows because a business with a calendar in it has two numbers on hand: how many inquiries came in, and how much work came out. The distance between them is a number worth knowing: the days between the moment someone asked for your time and the moment that time arrives.

That number has a name in the scheduling literature. It is called lead time, and there is a study that measures what it does. Michael McMullen and Peter Netland pulled every appointment out of the scheduling database at the University of Virginia Eye Clinic for a twelve month period. That was 51,529 of them, sorted by how far in advance each had been booked. Their paper, published in 2015, reports two clinics separately, one run by residents and one by faculty. The resident clinic missed 21.7% of its appointments and the faculty clinic 6.6%. That is a reminder before any of the rest: the level depends on who the customers are.

Both curves point the same way. In the resident clinic, appointments made nought to two weeks ahead were missed 9.1% of the time, and appointments made six months ahead 38.3%. In the faculty clinic the same comparison runs from 2.4% to 6.9%. The authors also ran the sums the other way: according to their model, the resident clinic's no-show rate would fall by nearly 60% if every appointment were scheduled within two weeks.

A patient is not a customer, and an eye clinic is not a roofer or a salon. The percentages above are not a benchmark for anything you do, and the study is from 2015, so it stays here for the argument. The direction is what should worry an owner: the most natural, polite and accommodating thing you do all week is push an appointment out until it suits everyone.

The evidence

How often people failed to show up, by how far ahead it was booked

How often people failed to show up, by how far ahead it was bookedResident clinic, booked 0 to 2 weeks ahead: 9.1%. Resident clinic, booked 6 months ahead: 38.3%. Faculty clinic, booked 0 to 2 weeks ahead: 2.4%. Faculty clinic, booked 6 months ahead: 6.9%. Share of scheduled appointments where the patient did not attend, from 51,529 appointments in one clinic's scheduling database over twelve months. The two clinics are shown separately because their baseline rates differ by more than a factor of three. The slope is the finding, not the level: both rise steeply with how far ahead the appointment was made.Resident clinic, booked 0 to 2 weeks ahead9.1%Resident clinic, booked 6 months ahead38.3%Faculty clinic, booked 0 to 2 weeks ahead2.4%Faculty clinic, booked 6 months ahead6.9%

Share of scheduled appointments where the patient did not attend, from 51,529 appointments in one clinic's scheduling database over twelve months. The two clinics are shown separately because their baseline rates differ by more than a factor of three. The slope is the finding, not the level: both rise steeply with how far ahead the appointment was made. Source: Michael J McMullen and Peter A Netland, Lead time for appointment and the no-show rate in an ophthalmology clinic, Clinical Ophthalmology 2015;9:513-516, University of Virginia School of Medicine.

This is an eye clinic, and a patient is not a customer. Nobody has run this study on estimate visits, salons or showings, and these percentages are not a benchmark for anything you do. What transfers is the direction, which is uncomfortable enough on its own. The appointment you push out to a date that suits everyone is measurably less likely to happen than the one you take tomorrow. The paper's own model makes the same point from the other side. It estimates the clinic's no-show rate would fall by nearly 60% if every appointment were booked within two weeks.

Why do longer waits lead to more no-shows?

Longer waits go with more no-shows, and the reasons are only partly proven. The chart measures the effect. It does not explain it. So it is worth being honest about which parts of the explanation have evidence under them and which are how people tend to work. The grid below gives four reasons: the person's situation moves, the commitment cools, they forget, and they book somebody else. The first is closest to what the chart measures, since more days is more time for a situation to move. The other three are mechanisms, offered as mechanisms.

What they have in common is that none of them is about the person being unreliable. The instinct after an empty evening is to decide something about the character of the people who did not come. A more useful reading is that they made a plan while they were keen, the plan sat for a week and a half, and life got to it.

Between the booking and the day

Four ways an agreed time quietly stops being one.

Their situation moves.

Nine days is long enough for another company's quote to arrive, for a job to change, or for a week to go badly. None of that is about you, and none of it can be got back. It is simply what happens to plans that are left out in the weather.

The commitment cools.

The person who agreed to Wednesday was in the mood that made them message you at a quarter past eight on a Sunday. By the following Wednesday they are a different person with a different week. The appointment is now a duty rather than an impulse.

They forget, in the ordinary way.

Not carelessly. It went into a mental note rather than into a calendar, because you wrote it in a text message and they were walking at the time. This is the one reason on this list that a machine can help with directly. That is why the next section is about reminders.

They booked somebody else in the meantime.

A nine day gap is nine days in which they can keep looking, and most people do. The appointment did not fail on the day. It failed on the Tuesday after they collected two other quotes from companies that had a Saturday free.

Raindrops in sharp focus on a car windscreen and, blurred beyond them, two parked cars under a grey and white cloudy sky, with the red lettering of a shop sign too soft to read
This is the half past six that nobody came to, seen from the driver's seat. The drive out, the twenty minutes waiting, and the nine days between the Monday call and this slot.

Does a reminder reduce no-shows?

A reminder helps reduce no-shows because, of those four reasons, one can be helped by software on its own. It happens to be the one with the best evidence behind it anywhere in this guide.

In 2007 four researchers at Sir Run Run Shaw Hospital in Hangzhou took 1,859 people with booked appointments at a health check-up centre and put them into three groups at random. One group got a text message 72 hours before their appointment. One got a phone call at the same interval, with the same content. The third got nothing, which was the normal practice. 1,848 of them were analysed.

The result, published in 2008, is one of the plainest numbers in this subject. Attendance was 80.5% with no reminder, 87.5% with a text and 88.3% with a call. Both reminder groups beat the control by a margin the paper reports as statistically significant, and the difference between the text and the call was not significant at all. The text cost less: 0.31 Yuan per person who turned up against 0.48 for the phone calls.

A newer and larger picture agrees on direction and is more careful about size. According to a systematic review and meta-analysis published in March 2026, which pooled 10 randomised controlled trials of reminders for hospital outpatient appointments (8,236 participants), attendance was 11 percent higher with reminders than without (risk ratio 1.11, 95% confidence interval 1.05 to 1.19). For text messages alone the ratio was 1.14, but its interval (0.99 to 1.31) just includes no effect; for telephone calls it was 1.11 (1.04 to 1.19). The paper reports high heterogeneity between the trials (I-squared of 83 percent), so read it as a direction.

Read all of it as a shape rather than a promise. These trials were run in clinics, not on estimate visits or haircuts. The people in Hangzhou had made a reservation for a routine health check-up, and the 2026 review covers hospital clinics. What carries over is the least glamorous sentence in this guide: one reminder, sent automatically, is cheap, and in these trials reminders went with higher attendance. It is also easy to switch off, because it feels like nagging.

The evidence

Attendance with one reminder, sent 72 hours before

Attendance with one reminder, sent 72 hours beforeNo reminder at all: 80.5%. One text message: 87.5%. One phone call: 88.3%. Share of people who attended, from the 1,848 randomly assigned participants the paper analysed. The 0.8 point gap between the text and the call is not a result: the paper reports no significant difference between the two (P=0.670). What is significant is either of them against no reminder at all.No reminder at all80.5%One text message87.5%One phone call88.3%

Share of people who attended, from the 1,848 randomly assigned participants the paper analysed. The 0.8 point gap between the text and the call is not a result: the paper reports no significant difference between the two (P=0.670). What is significant is either of them against no reminder at all. Source: Zhou-wen Chen, Li-zheng Fang, Li-ying Chen and Hong-lei Dai, Comparison of an SMS text messaging and phone reminder to improve attendance at a health promotion center: A randomized controlled trial, Journal of Zhejiang University Science B 2008;9(1):34-38.

China, 2007, routine health check-ups, and a population that had already gone as far as making a reservation and leaving two phone numbers for it. Seven points is what one reminder was worth there, and it is not a number to expect on an estimate visit or a haircut, because nobody has measured that. Two things do carry. The first is that a single reminder is the only intervention in this literature that is both cheap and reliably effective. The second is that the cheap channel performed as well as the expensive one, at roughly two thirds of the cost per person who turned up.

What does an AI appointment booking system do to reduce no-shows?

Under the category name, AI appointment booking is four small things in a row, and only the last one has a trial behind it.

It answers immediately, which is the part that gets sold. It offers times, which is where the real judgement lives, because a system that lists a fortnight of availability has handed the customer the choice that the evidence above suggests can hurt. It writes the slot down in both places. And then it reminds.

Notice what is not on that list. It does not persuade, qualify or have a personality. Each step is designed to happen every time, at nine on a Sunday evening as at eleven on a Tuesday morning, which is the part a person with a full week finds hard to repeat.

The mechanic, in four parts

Answer, offer, write it down, remind.

It answers while they are still holding the phone.

Seconds rather than the following morning. This part is the same argument the chat and voice pages make, and it is genuinely the smaller half of this one. That is because answering fast and then agreeing on a date nine days out has solved the cheap problem and left the expensive one alone.

It offers the soonest real slots, not a menu.

Read live from your calendar, so it cannot offer a time you do not have. And ordered so the nearest one is the easy one to take. That ordering is a small decision with the whole of the chart above behind it.

It writes the appointment on both calendars.

Yours, and theirs, as an actual event with an alarm on it rather than as a sentence in a message thread. The section below on what a booking technically is explains why those two things are not the same object.

It reminds them, which is the part with a trial behind it.

A confirmation immediately and a reminder before the day. It is the least impressive item on this list, and the only one where someone randomised eighteen hundred people to find out whether it works.

The mechanic

Sunday, a quarter past eight.

A staged illustration, not a client's messages. The times and the wording are invented. The sequence is the one the service page describes.

The exchange

  1. Hi, could someone come out and give us an estimate on a new patio?

    The homeowner / 8:15 pm

  2. Yes. I can get an estimator out tomorrow at 4:00pm or on Saturday at 10:00am. Which suits you better?

    The assistant / 8:15 pm

  3. Tomorrow works. Is it okay if my brother comes, he is the one who knows about the drainage

    The homeowner / 8:19 pm

  4. Of course. You are booked for tomorrow, Monday, at 4:00pm at your address. I have sent a calendar invitation to this number so it lands on your phone. The estimator will meet you there.

    The assistant / 8:19 pm

  5. got it, thanks

    The homeowner / 8:20 pm

What happened on your side

  1. The calendar is read

    Free and busy times only, not the contents of anything, and the two soonest genuine openings are offered.

    8:15 pm
  2. The slot is written

    Held on your calendar the moment they choose it, so the next person asking cannot be offered it.

    8:19 pm
  3. The invitation goes out

    A real calendar event with an alarm on it, sent to them, not a sentence in a thread.

    8:19 pm
  4. The reminder

    On the day, a few hours before the visit, with a way to move it. The single cheapest thing on this page and the only one with randomised trials behind it.

    12:00 pm

What reading your calendar should actually mean

A tool that says it reads your calendar may mean two quite different levels of access, and it is worth knowing which one you are being asked for.

The narrow version is a free and busy query. A widely used calendar service's developer interface has an endpoint for it, and its documentation says it returns free/busy information for a set of calendars. That is all: not the titles, not the attendees, not the notes, not the address of the job you are driving to. A wall of blocks.

The wide version is full read access to the events themselves. Some products ask for it because it lets them do more. It also means the vendor's systems can see who you are meeting and why. That may be a reasonable trade, and it is not one to agree to without noticing.

There is a second half to this that is easy to miss: the writing rather than the reading. A system that can only read cannot hold a slot, so two people asking at the same time can be offered the same half past six. The slot has to be written the moment someone takes it, not at the end of the conversation and not in a nightly sync.

What a booking is, technically, and why many of them are not one

The technical meaning of a booking sounds fussy, and it is the difference between an appointment that happens and one that may not.

When a booking system says it has confirmed an appointment, it has done one of three things. It may have sent a text message that contains a date and a time, which is a sentence. It may have attached a calendar file, which is a document. Or it may have sent an actual invitation, which is a transaction with a reply.

The format underneath is iCalendar, defined in RFC 5545. The specification says that if a calendar object has no METHOD property, a scheduling transaction must not be assumed, and the object is merely being used to transport a snapshot of some calendar information. In plainer words, an attachment with no method on it is a picture of an appointment. It may land nicely on a phone and it may not, and nothing comes back to say either way.

The same specification defines an alarm component, a reminder that lives on the customer's device rather than on your server. So the question to ask a vendor is not whether it sends a confirmation. It is whether the person on the other end gets something their calendar treats as an invitation, with an alarm attached, and whether you find out when they accept it.

The system

From a message to an event on somebody's phone.

Six hops, and the two timestamps are not the argument this time. The gap between the first hop and the last is what the chart above is about. This whole chain exists to make that gap a day rather than a fortnight.

The path from an inquiry at 8:15pm to a reminder sent on the day of the visit: The ask, The reply, The times, The hold, The invite, The reminderThe askSite, text or call8:15 PMThe replySeconds, not MondayThe timesFree and busy, read liveThe holdWritten to the calendarThe inviteAn event on their phoneThe reminderBefore the day, not after12:00 PM

Scroll to follow the chain

  1. The ask: Site, text or call
  2. The reply: Seconds, not Monday
  3. The times: Free and busy, read live
  4. The hold: Written to the calendar
  5. The invite: An event on their phone
  6. The reminder: Before the day, not after

In your numbers

How many of the people who ask actually end up in front of you?

Every channel: the website, the portal, the phone, the text somebody sends because they got your number from a sign.

8

Not an acknowledgement. An answer with a time in it.

70%

The ones that survive the back and forth and end up as a time both of you have written down.

60%

Your number, not the clinic's. If you have never counted, count the next twenty and come back.

80%

Appointments you actually keep, a year

140appointments

People askingyour 8
8a week
Over a year52 weeks
416asks a year
Answered the same dayyour 70%
291conversations
Reaching an agreed slotyour 60%
175appointments
Where somebody turns upyour 80%
140appointments

There is no second column here, and no line showing what this would become with reminders switched on. The missing row is the point. The trial on this page found one reminder worth about seven percentage points of attendance, in China, in 2007, among people who had made a reservation for a routine health check-up, and the 2026 review covers hospital clinics. Applying that to your own appointments would be the most flattering sum on this website, and it is a sum nobody has done. Every rate above is yours for the same reason. A default we typed in would quietly become the number you remember.

The time that suits everybody is almost never the time that happens. The soonest awkward slot beats the perfect one a fortnight out, and it is not close.

What should you do about the no-shows that still happen?

Even with the shortest lead times and a reminder, some appointments will not happen. The difference between businesses is in what happens in the following hour.

The hour after is the opportunity. Someone who missed a half past six knows at seven that they have missed it, and may be mildly embarrassed. A short message offering a new time, with no reproach in it, is a low-effort recovery. If it waits until the next morning it is competing with everything else in their inbox.

Write the message so it assumes something got in the way, because something usually did: two lines, one new time, and nothing that needs an apology to answer. A booking system can be set up to send this as well; ask a vendor to show it.

Then count them, not to punish anyone but because a business that does not know its own attendance rate cannot tell whether any of this worked. The number is simple: of the last twenty appointments you agreed, how many happened.

How do you test a tool to reduce no-shows before you buy it?

Four questions, each answerable inside a demo without knowing anything technical.

Book yourself an appointment from your own phone, in front of them. Watch what arrives. If it is a text message with a time in it, that is a sentence, not an invitation. Open whatever it sends, and see whether your own calendar offers you the option to accept it and sets a reminder by itself.

Ask which times it offered you, and why those. If it offered the next ten openings in a list, ask what happens if you change it to offer the soonest two. A product that cannot answer may not have thought about the part the research is about.

Ask what access it needs to your calendar, in the vendor's own words, and whether free and busy access is enough. If it needs to read the events themselves, ask what it does with them and where they are stored.

Ask what happens when two people take the same slot within a minute of each other, and ask to see it. That failure shows in front of a customer rather than in a log file.

The count, at least

Tell us how you take appointments today, and we will send back the two questions worth measuring first. One is your median gap between the ask and the slot. The other is how many of the last twenty were kept.

Ask for the two questions

It is a short reply from a person, it costs nothing, and the answer is useful whether you buy anything or not.

A barber shaving a customer with a razor, one hand at the customer's jaw and the other steadying his head, the customer tipped back in the chair with a white towel across his forehead and eyes, the barber's arm out of focus in the foreground
And this is the whole point of the machinery: one ordinary appointment at which two people are in the same room at the time they said they would be. Nothing about that hour is impressive. Everything expensive in this article happens because it did not.

Is automating booking and reminders worth it to reduce no-shows?

Automating booking and reminders is worth it when appointments are made often enough that a missed one is noticed within a week. These are the screening questions of our AI assessment, our own method rather than a published standard, and three of them keep the idea or cross it off:

  • Does the job happen often enough that a failure shows within a week? Appointments are made and missed weekly, and an empty slot is felt the same day.
  • Can the rule be written down for a new hire? The appointment types, how long each takes, your hours and how far ahead you will book fit on one page.
  • Does a wrong answer land where a person sees it first? A booking outside the rules goes to a person, and someone counts the no-shows each month.

Then the stop question, which asks what breaks if the job stopped: if nothing breaks, stop doing it and do not automate it. Then check whether a tool you already pay for does it, since a booking page or a calendar you use may already send a reminder you have not switched on.

Score the job's impact from 1 to 5. At 1 or 2, leave it alone: a text sent by hand the day before is enough. A confirmation and a reminder reach a customer, so they carry risk, and a person approves the wording before anything is sent.

An invented example: a salon books 30 appointments a week and 3 of them are no-shows. That is 3 x 52 = 156 empty slots a year. At 45 minutes a slot, 156 x 45 = 7,020 minutes, which is 117 hours of chair time a year (the owner's estimate), and the job passes the test. A roofer with 3 estimate visits a week and one miss a month has 12 empty visits a year: a text by hand the evening before is the answer, and the automation can wait.

The three questions are run on your own last twenty appointments before anyone looks at a tool.

What does it take to run booking and reminders that reduce no-shows?

We print no price on this page. What booking and reminders take from the owner is mostly four decisions and a calendar that tells the truth.

The four decisions are what the appointment types actually are, how long each one really needs including the drive, which hours you are willing to be booked into, and what the system should do when someone asks for a Sunday. Every one of those is a question about your business, and every one will still be true if you never automate anything.

The part people forget is keeping the calendar true. A booking system is only as good as the availability it can see. A business where half the commitments live in someone's head will get exactly the reliability that implies. That is the first month of work, and it is worth doing whether the software arrives or not.

Our setup sends messages only to people who agreed to get confirmations and reminders, and a reply of STOP ends them for that number. What the message says is the owner's to approve before the first one goes out.

What can AI appointment booking not do about no-shows?

AI appointment booking does not make anyone turn up. Reminders went with higher attendance in the trials above, and they are still only reminders.

It does not decide who is worth an appointment. A system that fills your week faster fills it with whoever asked. Booking speed and lead quality are separate problems with separate answers, and a booking assistant in front of an unfiltered inbox can fill a quiet Saturday with whoever asked, which is busier without being better.

It does not create time. If the week is genuinely full, the honest output is a longer lead time or a refusal. Some inquiries are better served by someone saying plainly that the soonest real slot is a fortnight away, and asking whether that still works.

It does not run the appointment. It is built to get two people to the same place at the same time with the address written down. What happens in the next forty minutes is the part you are paid for.

And it does not repair a calendar you do not trust. If you already override the calendar three times a week because it is wrong, a machine reading it will be wrong three times a week as well, in front of customers.

Three ways it is wasted

None of them are the booking software.

It fills the week with the wrong people.

Booking brings you more of whatever is already coming, and it does not read intent. If nothing sits between the inquiry and the calendar, then you can spend Saturday driving to three estimate visits for people who are eleven months from deciding. That is a worse week than the one you had before.

The calendar it reads is not the calendar you live in.

Many appointments in a small business are agreed by voice and never written down anywhere a machine can see. A booking assistant reading a calendar with holes in it will offer a slot you are already standing in someone's kitchen for. And it will do it politely.

The reminder is the first thing switched off.

It feels like nagging, someone complains once, and it quietly comes off. It is also the only part of this whole system with a randomised controlled trial behind it. Taking it out leaves you paying for the impressive half and throwing away the half that was measured.

Who this fits: the businesses that lose money to no-shows

Booking and reminders fit a business that sells its time in slots and loses money when a slot sits empty: a roofer, plumber or heating company with estimate visits, a salon, spa or gym, a dental or accounting office, a real estate office with showings and listing appointments.

Three kinds of business do something else. A business with a handful of appointments a month sends a text by hand the evening before. A business that already books every customer within days and reminds every one has little distance left to close. And a business where nobody will check the calendar and count the misses each month should start with the count, because it cannot tell yet whether anything would improve.

An invented example: a piano tuner with forty visits a year keeps a text on his phone and sends it from the driveway the day before.

Which of these groups your business is in decides the next step.

Who this is for

A good fit, a poor fit, and what to do instead

A good fit

A business that sells its time in slots: estimate visits, a salon, spa or gym, a dental or accounting office, a real estate office with showings.

A poor fit

A handful of appointments a month, a business that already books within days and reminds everyone, or nobody to count the misses each month.

Instead, for a poor fit

A text sent by hand the evening before, and the count of the last twenty appointments.

Either way

A person approves the booking rules and the reminder wording, and a person decides what happens to a miss.

What should an owner do first to reduce no-shows?

The first step to reduce no-shows is one job, taken a week at a time, ending with a decision written down.

  • Week 1: count the last twenty appointments you agreed. For each, write the days between the conversation and the slot, and whether it happened. Work out the median of the first column and the share of the second.
  • Week 2: for the next appointments, offer the two soonest real slots first, send one reminder by hand the day before, and write down what changed.
  • Weeks 3 and 4: run the test in "Is automating booking and reminders worth it to reduce no-shows?". If it passes, approve the booking rules and the reminder wording and read every booking for a month. If it fails, keep the hand reminder and write that decision down.

This is also how our AI readiness assessment starts. In its interview we ask for one job written in two minutes under input, do, output and your check, then handed to a colleague who does not know it, noting where they stall. Each job is then scored for impact and risk from 1 to 5, with four checks: it binds the business, it reaches a customer, it feeds another system, it can damage trust. The owner leaves with each job marked to automate, to keep a person on, or to leave alone.

An invented example: a salon owner writes her booking job in two minutes. Her colleague stalls at "how far ahead do we book colour", and that answer becomes the first booking rule.

A suggested schedule

One job, a week at a time, a decision written down

  1. Week 1

    Count the last twenty appointments: days between the conversation and the slot, and whether it happened.

  2. Week 2

    Offer the two soonest real slots first and send one reminder by hand the day before.

  3. Weeks 3 and 4

    The worth-it test; then approve the booking rules and wording and read every booking, or keep the hand reminder and write the decision down.

A schedule to start from, not a deadline.

Source: our method, the AI assessment

Common questions, answered honestly

How do I reduce no-shows?

Book inside the first conversation, keep the date close, and send a reminder before the day with an easy way to move it. According to a randomised trial of 1,848 appointments at one health check-up centre, attendance was 80.5 percent with no reminder and 87.5 percent with one text sent 72 hours ahead. According to a university eye clinic study, appointments made six months ahead were missed several times more often than ones made within two weeks. Neither study is about your trade, so measure your own rate before and after.

What is a good no-show rate?

There is no benchmark in this guide, because the rate depends on who the customers are. According to the eye clinic study, the resident clinic missed 21.7 percent of its appointments and the faculty clinic 6.6 percent. The useful number is your own: of the last twenty appointments you agreed, how many happened.

What should an appointment confirmation and reminder text say?

A confirmation text and a reminder text both carry the same things: the business's name, the day and the time, the address, and one way to move it. Nothing private: for a dental or accounting office, never the treatment or the matter. It goes to people who agreed to get reminders, and a reply of STOP ends them.

What is AI appointment booking, in plain terms?

AI appointment booking is software that does four things. It answers a request for your time straight away. It offers times that are genuinely free, by reading your calendar. It writes the chosen slot onto both calendars while the conversation is still happening. And then it sends the confirmation and the reminder. The AI part is only the conversation, understanding a message written the way people write and holding a short exchange about times without sounding like a form. The booking underneath is ordinary software, and that is a good thing.

A booking link moves the work to the other person: they open a page, look at a grid of times and choose one. That improves on phone tag. It also asks the person to do a task, where answering a message is a smaller step. Here the times arrive inside the conversation the person is already in, and the answer is a word rather than a form. The two also differ in what they can notice: a link shows a fixed grid, and a conversation can be set up to see a request for a Saturday you do not work, say so and offer the nearest thing you do have.

Will it reduce no-shows?

Two parts of it can, and it is worth knowing which. The reminder is the part with evidence: the randomised trial above, and the 2026 review of ten trials in hospital clinics. The other part is less obvious. Booking inside the first conversation tends to shorten the gap between the ask and the appointment, and shorter gaps went with far better attendance in the clinic study above. Neither is from your trade, so anyone quoting a no-show reduction for your business is quoting a number nobody has measured.

How does it know when I am free?

An AI booking tool knows when you are free by asking your calendar. A free and busy query returns only the time ranges you are busy, with none of the contents. Full access lets a system read the events themselves. Ask which one a vendor wants. For booking alone, free and busy plus permission to create an event is enough, and anything beyond that should come with a reason.

Can it book on the phone as well as by text?

Yes, and it is the same booking layer underneath. A caller can be booked mid-conversation by the voice agent, exactly as a website visitor can be booked mid-chat. That means one calendar rather than two, so the two channels read the same availability.

What if somebody wants a time I do not have?

Then it should say so and offer the nearest other times, and it should never offer a slot you have blocked out. The more interesting question is what you want it to do with a request for a Sunday or for eight in the morning. That is a decision about your life rather than a setting, so decide it on purpose.

Is a shorter lead time always better?

Not always, and the study is about attendance rather than revenue. An appointment two days out is more likely to happen, and it also gives someone less time to arrange the money, a partner or childcare. For some appointments those matter more than attendance does. The finding is not that everything should be tomorrow. It is that the fortnight you drifted into because it suited the calendar has a cost, and now you know roughly which direction it runs.

Can we talk about no-shows on a call if my business is outside New York?

We are based in New York, and the call works for an owner anywhere.

Ten questions

What did you learn? Check it in two minutes

Ten questions on what this page taught. It takes about two minutes.

Take the quiz
  1. 1 of 10

    At the University of Virginia eye clinic, what share of the resident clinic's appointments made six months ahead were missed?

    • A.9.1 percent
    • B.21.7 percent
    • C.38.3 percent
    • D.6.9 percent
    Show the answer

    C. 38.3 percent The resident clinic missed 38.3% of appointments made six months ahead, against 9.1% for nought to two weeks; see "Why does the gap to the appointment hide the cost of no-shows?".

  2. 2 of 10

    What carries over from the eye clinic study to a roofer or a salon?

    • A.The exact percentages
    • B.The direction: the further ahead it is booked, the more are missed
    • C.The size of the clinic
    • D.The time of day
    Show the answer

    B. The direction: the further ahead it is booked, the more are missed The percentages are not a benchmark for anything you do; the direction is the finding; see "Why does the gap to the appointment hide the cost of no-shows?".

  3. 3 of 10

    In the Hangzhou trial, what was attendance with one text reminder sent 72 hours ahead?

    • A.80.5 percent
    • B.88.3 percent
    • C.91 percent
    • D.87.5 percent
    Show the answer

    D. 87.5 percent Attendance was 80.5% with no reminder, 87.5% with a text and 88.3% with a call; see "Does a reminder reduce no-shows?".

  4. 4 of 10

    In the same trial, was the difference between the text and the call significant?

    • A.Yes, the call was far better
    • B.No, and the text cost less per person who turned up
    • C.Yes, the text was far better
    • D.The paper did not compare them
    Show the answer

    B. No, and the text cost less per person who turned up The difference was not significant, and the text cost 0.31 Yuan per attendance against 0.48 for the call; see "Does a reminder reduce no-shows?".

  5. 5 of 10

    What did the March 2026 review of 10 randomised trials find about reminders overall?

    • A.Attendance 11 percent higher with reminders
    • B.No effect on attendance
    • C.Attendance 50 percent higher
    • D.Lower attendance
    Show the answer

    A. Attendance 11 percent higher with reminders Pooled attendance was 11 percent higher with reminders (risk ratio 1.11); see "Does a reminder reduce no-shows?".

  6. 6 of 10

    According to RFC 5545, what is a calendar object with no METHOD property?

    • A.A scheduling request
    • B.An alarm
    • C.A snapshot of calendar information, not a scheduling transaction
    • D.A reminder text
    Show the answer

    C. A snapshot of calendar information, not a scheduling transaction Without a METHOD a scheduling transaction must not be assumed; see "What a booking is, technically, and why many of them are not one".

  7. 7 of 10

    What does a free and busy query return?

    • A.The titles and attendees of every event
    • B.Just the busy blocks, with no event details
    • C.The addresses of your jobs
    • D.Your whole calendar
    Show the answer

    B. Just the busy blocks, with no event details It returns free and busy information only, a wall of blocks; see "What reading your calendar should actually mean".

  8. 8 of 10

    A salon books 30 appointments a week and 3 are no-shows. How many empty slots is that a year?

    • A.36
    • B.90
    • C.156
    • D.1,560
    Show the answer

    C. 156 3 x 52 = 156 empty slots a year in the invented example; see "Is automating booking and reminders worth it to reduce no-shows?".

  9. 9 of 10

    A job scores 1 or 2 for impact. What is the verdict?

    • A.Automate it at once
    • B.Buy a second tool
    • C.Hire someone to do it
    • D.Leave it alone and send a text by hand the evening before
    Show the answer

    D. Leave it alone and send a text by hand the evening before At 1 or 2, leave it alone; a text sent by hand the day before is enough; see "Is automating booking and reminders worth it to reduce no-shows?".

  10. 10 of 10

    What happens in week 1 of the roadmap?

    • A.Count the last twenty appointments you agreed
    • B.Install a booking tool
    • C.Switch on automatic reminders
    • D.Offer the two soonest slots
    Show the answer

    A. Count the last twenty appointments you agreed Week 1 counts the last twenty appointments: the days between the conversation and the slot, and whether it happened; see "What should an owner do first to reduce no-shows?".

What can you do about no-shows this week?

The count needs nobody, us included. It needs only paper: the last twenty appointments you agreed with someone outside the business, each with two numbers, the days between the conversation and the slot and whether it happened. The median of the first column and the share of the second are the whole of this guide in your own handwriting.

In the invented example the median was nine days, from the Monday call to the slot, and the empty curb on the ninth day was not the fault of the person who did not come.

In the invented example at the top, the slot was agreed nine days out, on the Monday call, for a person who was certain on the Sunday. Nothing in the schedule records it as a loss, and the count of your own last twenty appointments is where it would show.

There is no price here. What booking takes is mostly four decisions and a calendar that tells the truth. The AI assessment is an hour, done with you, and the plan it leads to starts with the first booking path on one channel.

Ask us about AI appointment booking.

Sources

  • McMullen and Netland, University of Virginia School of Medicine, on appointment lead time and no-shows in an eye clinic, 2015, the paper. Read November 14, 2026; kept for the argument, the study in this guide on lead time.
  • Chen, Fang, Chen and Dai, a randomized controlled trial of text message and phone reminders at a health promotion center in Hangzhou, 2008, the paper. Read November 14, 2026; kept for the argument, the trial in this guide on reminders.
  • Majed Al-Turbag, Mary Mooney and Margarita Corry, Trinity College Dublin, A systematic review and meta-analysis of appointment reminders for enhancing hospital attendance, Journal of Hospital Management and Health Policy, published online March 6, 2026, the article. Read November 14, 2026.
  • A widely used calendar service's developer documentation, on its free and busy query, the reference, a live document. Read November 14, 2026.
  • IETF, RFC 5545, Internet Calendaring and Scheduling Core Object Specification, the standard, September 2009. Read November 14, 2026; kept for the argument, a standard in force.
  • Our own method: the screening questions, the scores, the four checks and the verdicts of the AI assessment.

Sources checked November 14, 2026.

Written with an AI assistant and checked by a person.

Know somebody who would argue with this? Send it to them.