Nobody decided to let them go. They stopped being anyone's job.
How to build a recall list you can defend, sequence it across six touchpoints, and read the numbers it produces without panicking or celebrating too early. With the messages written for you.
Six parts, in the order you will use them.
Everything here is a starting point. Please adjust the windows, the timings and the wording to match how your clinic works and how often your patients would naturally expect to hear from you.
To suggest a change or leave a note, mark your edits in a different colour so they are easy to spot.
A patient enters the recall pathway when both of the following are true.
Their last appointment sits outside the recency window for the service they came in for.
They have no future appointment booked, and no status on the record that explains why.
Discharged, moved away, referred on, opted out. No action required. The patient is accounted for and does not enter recall.
Nobody decided anything. This is the population your recall list is drawn from, and in most clinics it is far larger than anyone expects.
That second box is the whole exercise. Pull every patient in your system with no next appointment and ask what each one's status is. If they all carry one, you do not have a recall problem. If most of them carry nothing, the size of that group is the size of your problem.
There is a version of this answer that proves a clinic has it under control, and it sounds like this: ten thousand patients on the books, five thousand formally discharged, two and a half thousand moved away or deceased, and of the remaining two and a half thousand we are in active communication with a thousand. Very few clinics can describe their patient body that way. The usual answer is a shrug and a rough monthly figure.
How many patients should have a next appointment and do not?
The answer is not “it is about forty a month”. The answer is “four hundred and sixty-seven last month”. If nobody in the room can produce the second kind of answer, that is the first thing to fix, before any message goes out.
Follow-up covers everything that happens between appointments: clinical, administrative, supportive and commercial. Recall is the narrow slice that says a patient attended some time ago, has nothing booked, and probably should have.
Their last visit is outside your window and nothing is booked. The message is an offer to pick things back up.
The same mechanism aimed much further back: the patients nobody has spoken to in a year or more. Cheaper to work than any advertising, and almost always ignored.
A patient who walks out today with nothing booked is a different case. Catch them inside three days and they usually come back. Miss that window and the likelihood drops sharply. Run that as a same-week exit list, not as recall.
A patient who cancelled is on a shorter, faster pathway with higher intent. That has its own guide, the Clinic Cancellation Guide, and its own set of messages.
“Contact everyone who has not been in for a while” is not a strategy. Everyone in the system and everyone worth contacting are two different lists, and the gap between them is where most recall problems start.
Your system does not know that a patient has died, moved away, is a staff test account, or already has an appointment next Tuesday, unless something explicitly tells it so. Left unfiltered, “everyone” quietly becomes a list padded with people you should never have messaged. Each of those is a small trust cost at best. At worst it is a complaint with your clinic's name on it.
Deceased patients, and the concrete answer to how that gets flagged before you rely on it.
Under-18s, or anyone whose recall needs a guardian workflow.
Staff, staff family, and test or dummy accounts.
Funding types you cannot currently action through this channel.
Patients who already have a future appointment booked.
Duplicate patient records.
Two internal reports covering what is meant to be the same population over the same window will often disagree. This is rarely anyone being wrong. It is usually two different filters: duplicates stripped or not, one site or a rolled-up total, a cumulative figure against one week's slice. Each is locally correct and mutually inconsistent.
The fix is not to pick the more convenient number. Write down which filter produced which number, next to the number itself. A report without its method attached is not wrong, but nobody reading it in a month can tell the difference.
Recall is not a cancellation. A patient who cancelled yesterday still has the appointment in mind, so that pathway is front-loaded into the first 48 hours. A patient who last came in eight months ago is not thinking about you at all, so the arc is longer, lighter and more spread out. Contact opens promptly, then tapers over about six weeks.
Most clinics stop at two or three attempts and record the rest as not interested. That is where the majority of recoverable patients are lost. Six to twelve touchpoints across more than one channel is the working range, and patients routinely reply on the fifth or seventh as though it were the first message they had seen. Persistence is not pestering when the person already chose you once.
The moment a patient replies on any channel, switch the others off. Nothing damages a recall programme faster than a patient who has already answered on WhatsApp receiving the scheduled email and then the call.
A patient who attends every twelve months is not lapsed at month six. Re-engage them at month thirteen. A rule that treats every patient as lapsed at a fixed interval will chase people who are exactly on schedule.
“Your check-up is due” informs. “Reply YES and I'll hold Thursday at 2pm” asks for a decision. The second gets answered; the first gets read and forgotten.
Generic “we miss you” copy reads as marketing whatever you call it. A line that names the treatment, the clinician or the problem they came in with reads as a person who remembers them.
You have crossed into spam when a patient tells you they will get in touch if they need anything. Until then you have not. For every patient who says that, there are many more who reply saying thanks for checking in.
The difference between outreach and marketing is not the volume, it is whether the message could have been sent to anyone else on the list without changing a word.
Across roughly 140 consecutive patient replies to live recall and follow-up programmes, about 56% arrived by SMS, 30% by WhatsApp, 12% by email and the rest by phone. That is the mix patients reply on, not the mix most clinics send on.
Two things follow. The first is that email should not carry the sequence on its own: it is the channel patients are least likely to answer on, and the one where a non-reply tells you least. The second is that the copy has to change shape with the channel. A four-paragraph message with a formal sign-off reads as a letter, and patients answer letters with silence. On WhatsApp and SMS they answer in six words, so write something that can be answered in six words.
The most avoidable failure in recall is the patient who replies asking who you are. It happens whenever a message opens with a warm greeting and gets to the clinic's name at the bottom, or never. A patient who cannot place you in the first line does not read the second. Open with the coordinator's first name and the clinic name, every time, on every channel.
Patients reply to a person. They open with the coordinator's first name, and they are still using it months later. Pick one name per clinic, keep it consistent across every channel, and do not rotate it between whoever is on shift. A message signed by the clinic rather than by a person gets answered less.
"Let me know when suits" puts the work on the patient. Two named slots gets a one-line reply choosing one, which is the single most common booking pattern in live programmes.
Six touchpoints. WhatsApp and SMS carry short messages; the one email is the only place a paragraph belongs. Fields written as `{{first_name}}` are filled automatically at send. Replace `xx` with your own coordinator's first name and keep it the same throughout.
The sequence is the easy half. What separates a recall programme that books patients from one that annoys them is what happens in the hour after somebody answers.
Most replies are short and most are not a straight yes. In live programmes the largest single group is patients choosing one of the offered slots, and the second largest is patients who are away and will come back to you. Everything below that is a question, a correction, or something that should never have been sent.
| What they say | What to do |
|---|---|
| They pick one of your slots | Book it and confirm the same day. A held slot that goes quiet for 24 hours is the most expensive silence in the sequence. |
| “I'm away until the 11th” | Snooze to a real date, confirm you will come back then, and stop every other message in the meantime. This is one of the most common replies and a scheduled follow-up landing mid-holiday undoes it. |
| A question instead of a booking | Answer the question first, in full, and offer times underneath it. Price, suitability, who they would see, parking, what to bring: these are buying questions, not objections. |
| “I'm fine, I don't need anything” | Stop, mark the record, and do not re-add them at the next campaign. This is the line, and it is the only definition of it that matters. |
| “Opt out” or “STOP” | Honour it immediately and across every channel, not just the one it arrived on. Then check why they were on the list at all. |
| Anything clinical | Hand it to a clinician the same day and take it out of the recall thread. Patients do disclose new symptoms, deterioration, pregnancy and serious diagnoses in reply to a friendly check-in, and a coordinator must never be the one answering. |
| “Who are you?” | A sender identity failure, not a patient problem. Fix the opening line of the template before the next send. |
| “I'm staff, take me off” | An exclusion failure. Fix the record, then fix the filter that let it through. |
A clinical disclosure goes to a clinician the same day, out of the recall thread, with the patient told who is picking it up.
An opt-out is honoured immediately, everywhere, before anything else in the queue. Both of these outrank the booking.
Patients who reply late apologise for it. Clinics that reply late lose the patient to whoever answered first, and the reply usually says so plainly: they found someone local, they got seen elsewhere, they sorted it themselves.
Set a response-time standard for replies to recall messages and measure it. Anything slower than a few hours in working time and the sequence is generating goodwill you are not collecting.
The most common mistake in reading a recall report is comparing raw counts across populations or windows of different sizes, then drawing a conclusion from the comparison.
Ten opt-outs sounds worse than five, right up until you learn the ten came from a population twice the size, at which point they are identical. Any number on its own tells you almost nothing until it is divided by the population it came from. Before comparing anything, convert to a rate: opt-outs over messages sent, bookings over patients engaged. Never the numerator by itself.
Low single digits of messages sent, for a well-targeted list. Materially higher is usually a targeting or tone problem, not proof that patients do not want to hear from you.
A modest minority of those contacted. Most patients will not respond to a single touch, which is why the sequence has six. On its own this does not mean the campaign failed.
The smallest slice, taken from those who engaged. Protect this one from noise: it sits downstream of the other two, so it moves whenever they do.
Opting out of WhatsApp is one tap. Unsubscribing from email carries more friction. That alone pushes the WhatsApp rate above email even when both messages landed equally well. Compare a channel to its own history, not to another channel.
These ranges exist to catch when something is unusually wrong, not to set a bar every campaign must clear. A number outside the normal range is a prompt to investigate, not a verdict on the strategy.
Only once all four are ruled out does a below-benchmark number become a genuine question about the message, the offer or the audience, rather than a question about the plumbing.
Five levels. Most clinics who believe they have recall are at level one or two, which is the gap between the perception of recall and the reality of it.
A Scotland-based clinic turning over £1.2m grew 15% in a year without spending another pound on marketing. Just under half of that growth came from one segment: patients who had not visited in six months or more.
Those patients were already in the database. The acquisition cost had been paid years earlier. The only thing standing between them and a second visit was whether anyone in the building remembered to ask.
Questions, edits, or anything you'd like us to adapt — write to team@coherenthealthcare.com.