Reminders that actually get answered

A reminder gets answered when it is specific about the appointment, easy to act on, and early enough that there is still room to move things. Most reminders fail one of those three. They arrive an hour beforehand, they say less than the diary entry already said, or they ask the reader to ring a number during working hours to reschedule.

There is real evidence about this, and some of it is stronger than the marketing built on top of it. This post takes that research as far as it honestly goes, marks where it stops, and turns what is left into a sequence that holds the diary. The mailbox a written reminder goes out from is the subject of AI email automation.

Being specific about the appointment and easy to act on. A message that names the date, the time, the place and the person, and offers a one-tap way to move the booking, gives the reader something they can do in the ten seconds they are looking at it. A message that only says an appointment is coming asks them to remember to act later, which is the behaviour that produced the missed appointment in the first place.

The cost of getting this wrong is easy to underestimate. A London NHS dental hospital recorded a 14.5% did-not-attend rate across its new-patient clinics (Journal of Dentistry, 2025), and a slot missed on Tuesday cannot be sold again on Wednesday. The wider argument for closing that leak without charging people upfront is in how to cut no-shows without deposits.

Early enough that the person can still do something about it. A reminder arriving an hour beforehand can only inform. One arriving a day or two ahead can be acted on, because there is room on both sides of the diary to move things. One workable shape is two messages, at forty-eight hours and again at twenty-four, and the timing is agreed with you before anything runs.

Worth saying plainly that the research does not hand you a single correct interval. The strongest trial evidence on wording sent its messages five days ahead of the appointment. The interval that matters is the one that leaves your client enough room to rearrange their week, which makes it a question about your clients.

Yes, and by more than a text message is usually credited with. In the first of two randomised controlled trials across an NHS hospital trust, covering 10,111 appointments, a reminder stating the specific cost of a missed appointment to the health service produced a did-not-attend rate of 8.4%, against 11.1% for the existing message (Hallsworth and others, PLOS ONE, 2015). The lever tested there was a public-service cost, which a private clinic does not have. What carries over is that a concrete detail outperformed a general one.

The trials are worth reading properly. Trial One allocated 10,111 appointments at an NHS hospital trust to one of four SMS reminders, all issued five days ahead, and the version stating the specific cost of a missed appointment to the health service came out lowest at 8.4%, against 11.1% for the existing message. Trial Two ran that same message against gentler variants across a further 9,848 appointments and it won again, and the authors projected roughly 5,800 fewer missed appointments a year for that trust at no additional cost (Hallsworth and others, PLOS ONE, 2015). Take from it that the content of a reminder is a lever you can pull. Treat the exact wording as belonging to the setting it was tested in, because a private clinic asking a patient to consider the cost to the NHS is quoting somebody else's argument.

No, though the evidence for text is the strongest of the channels studied. A Cochrane review of four randomised trials covering 3,547 people carries two findings. Three of those trials showed text reminders improving attendance over sending nothing, at a risk ratio of 1.10. Two of them showed text and a phone call having a similar impact, with a lower cost per attendance for the text (Car and others, Cochrane Database of Systematic Reviews, 2012). The practical answer is to reach people where they already read, which for most lists means text and email together.

The reviewers grade both of those findings as moderate quality and call for higher-quality research before firmer conclusions are drawn (Car and others, Cochrane Database of Systematic Reviews, 2012). They still point somewhere practical. If a text performs about as well as a phone call and costs less per attendance, the hours your team saves belong on the appointments that genuinely need a conversation.

The test is whether each message still has a job of its own. A confirmation when the booking is made, a reminder while there is still time to move it, and a short prompt on the day are three messages doing three different things. A fourth that repeats the third spends goodwill and buys nothing. Every sequence stops the moment the person confirms, reschedules or opts out, so nobody is chased for something they have already done.

It should not require a phone call during working hours to change an appointment, because that is the friction the message was sent to remove. It should not arrive after the person has already rescheduled, which means the sequence has to read the live diary and stop itself when the booking moves. And it should not be the first time the client hears your cancellation terms, which is a conversation that belongs to the moment of booking.

Reminders are only as good as the diary behind them. Confirmations, reminders, rescheduling and recall outreach are ordinary work for an AI employee, running in wording you approved and at timings you set, with an audit trail showing what was sent and when. For practices that have to evidence delivery, that trail is most of the value.

Count last month's empty slots and multiply by what one of them is worth. That arithmetic is the whole business case for a reminder sequence, and nobody outside your practice can do it for you. We build the sequence as part of an AI employee role.