Work Your Recall List

Most practices do not have a recall problem, they have a recall process problem. The list exists, it is accurate, and it is worked in whatever minutes are left over, which in a busy practice is none. The result is not a decision anyone made; it is a job that lost every week to jobs with a person attached to them.

This guide turns that into a routine. Eight lessons, about 3 hours to set up, then roughly 90 minutes a week. The output is a cleaned and segmented list, a written message, a protected slot, and two rates you can watch change.

It assumes you have read the recall list you already own or at least computed your contacted rate, because everything below is a change to a number and a change to a number you have never measured is indistinguishable from a change to nothing.

What this produces, and what it will not fix#

By the end you will have a weekly routine that runs whether or not anyone is enthusiastic that week, a message you have written once rather than improvised each time, and two rates measured monthly. Those are the outputs, and they are deliberately modest.

What it will not fix is a practice whose diary is full. Recall converts available capacity into booked capacity, so if you have no capacity the work produces frustrated patients and a longer wait list. Check that you actually have chairs to fill before spending 90 minutes a week filling them.

It also will not fix a list whose underlying intervals are wrong, which lesson 1 exists to catch before you start ringing people about an appointment they do not need.

Recall converts available capacity into booked capacity. If the diary is already full, this guide is the wrong work.

The commercial context is favourable and worth naming. In 2025 one third of dentists reported they were not busy enough, up from one quarter in the fourth quarter of 2024 [3]. A practice with unfilled capacity and an unworked recall list is holding both halves of the answer and has not put them together.

01Pull the list and clean it before you ring anybody#

The first pass is data, not people. A neglected list carries duplicates, wrong intervals, and contacts that no longer reach a human, and ringing through it uncleaned means your first measured rate describes your data quality rather than your process.

1. Run the recall-due report for the last 12 months and export it.
2. Drop anyone with a future appointment already booked. This is the most common single error and it produces the worst phone call in the set.
3. Drop anyone marked deceased, transferred out, or inactive.
4. Flag records with no mobile and no email, since those are letter-or-nothing and belong in a different queue.
5. Flag records whose recall interval disagrees with your current policy, and set them aside rather than fixing them in bulk.
6. Count what is left. That is your real list, and it is usually 15% to 30% smaller than the report said.

Step 2 deserves the emphasis. A patient who booked 3 weeks ago and gets a call saying they are overdue learns that the practice does not know what it is doing, and that impression is expensive and hard to reverse.

Step 4 is worth doing even though it feels like admin. A record with no reachable contact is not a patient who declined, and leaving those rows in the main list drags your contacted rate down for a reason that has nothing to do with your process. Counting them separately turns an unexplained bad number into a data-quality task somebody can actually do.

Expect the cleaning pass to take about 45 minutes the first time and around 10 minutes on each repeat, because most of what it removes stays removed. That first 45 minutes is the highest-return time in this whole guide.

Step 5 exists because interval drift is common and invisible. Practices that changed policy at some point, or merged another practice's records, carry a mix of intervals nobody reconciled, and the list quietly reflects two policies at once. Do not bulk-fix these; a clinician should look at them.

02Split the list by how overdue each patient is#

A patient 2 weeks past their interval and one 14 months past are different conversations with different success rates, and working them as one list means writing a message that suits neither.

Work the segments in that order, and treat the last row as a separate exercise entirely. Reactivation has its own message, its own success rate, and its own reporting, and mixing it into recall produces a denominator that grows forever and a rate that only ever falls.

There is a second reason to segment that has nothing to do with the message. The segments have very different value per contact, so a slot that runs out halfway through should have spent its time on the rows most likely to book. Working an unsorted export means the order is arbitrary, and an arbitrary order spends your best hour wherever the report happened to put things.

The practical value of segmenting shows up immediately in expectations. A 40% booking rate on the Due segment and a 6% rate on Lapsed are both normal, and a practice that reports them as one blended 18% has learned nothing it can act on.

A blended recall rate is two good numbers and two bad ones averaged into one that describes nobody.

Set a floor on the Lapsed segment too. Beyond about 24 months, contact details are unreliable enough that you are mostly measuring your data rather than your patients, and the time is usually better spent on the first two rows.

On the medical side, what counts as overdue is not yours to invent. The US Preventive Services Task Force publishes recommended screening intervals by topic, and those intervals are the reference an auditor or a payer will use rather than whatever your practice management system was configured with in 2019 [4]. Dental recall runs on a different cadence set by clinical risk, which is why the two lists should be segmented separately rather than merged.

03Decide the channel for each segment#

Channel is a cost decision more than an effectiveness one, which is the useful finding from the reminder literature even though that literature is about appointment attendance rather than about booking in the first place.

Text and telephone reminders produce statistically indistinguishable attendance, and 2 studies in the pooled review put the cost per text per attendance at 55% and 65% below the equivalent phone call [1]. That is a large enough gap to decide the default channel on, provided you are honest that it is evidence about a different job.

1. Default to text for the Due and Overdue segments, where the message is short and the ask is simple.
2. Use a call for Lapsing, where the conversation is longer and a person can answer an objection.
3. Use a letter only for the no-mobile-no-email flag from lesson 1.
4. Never use 2 channels on the same patient in the same week. It reads as pressure and it corrupts your attribution.
5. Record which channel each contact used, because lesson 8 compares them.

Step 4 matters more than it looks. A patient texted on Monday and rung on Wednesday will often book, and you will not know which one worked, so you will do both forever at twice the cost.

Step 5 is the one that gets skipped and then regretted. Recording the channel costs a single column in the outcome log and it is the only way to answer the question you will inevitably ask in month 3, which is whether the calls are worth the time they take compared with the texts.

If you are texting at any volume, the consent and registration questions are separate from everything in this guide and have to be settled first. They are worked through in the patient texting guide, and they are not optional.

04Write the message once, including what it must not say#

An improvised recall message is a different message every week, which makes lesson 8's measurement meaningless. Write it once, keep it short, and write down the boundaries as explicitly as the content.

• Names the practice in the first 6 words, so it is not mistaken for spam
• States what is due, in plain language rather than a clinical code
• Gives exactly 1 way to respond, and makes it the easiest one
• Contains no offer, discount, or promotion of anything
• Contains no clinical detail beyond what is due
• Includes opt-out wording if it is a text
• Fits in 1 message segment, so it is not split or truncated

The fourth line is the legally loaded one. Under HIPAA, marketing means a communication about a product or service that encourages the recipient to buy or use it, and communications for treatment, including case management and care coordination, are carved out of that definition [2].

A recall message sits comfortably inside that carve-out. Attaching an offer to it converts the whole message into marketing, which is the single most common way a practice turns a permitted communication into one that needs an authorisation it does not have.

The fifth line is the one staff push back on, because more clinical detail feels more helpful. It is not: a message that names a condition can be read by whoever picks up the phone, and the patient did not consent to that. Say what is due, not why.

The seventh line is mechanical and easy to breach. A text longer than 1 segment costs more, can arrive split out of order on some handsets, and gets truncated in the preview that decides whether it is read at all. Count the characters once when you write the message rather than discovering the problem from a delivery report.

Write 2 versions while you are at it, one for text and one as a call opener, and keep them saying the same thing. A caller who was texted last week and rung this week should hear the message they already read, not a different pitch, and staff improvising the spoken version is how the two quietly diverge.

05Book a fixed slot, and defend it#

This is the lesson that decides whether any of the others happen. Recall loses to every job with a person standing in front of it, so it needs a slot rather than an intention.

1. Pick one 90 minute block a week, at a genuinely quiet hour rather than an aspirational one.
2. Put it in the rota as a named commitment, not as a note.
3. Name one person who owns it, and one who covers when they are away.
4. Agree, out loud, what is allowed to interrupt it. The answer should be a very short list.
5. If it is missed 2 weeks running, that is a finding to discuss, not a failure to apologise for.

Step 1 is where most attempts die. Practices choose Monday morning because that is when the week starts, and Monday morning is the busiest hour in most practices. Mid-afternoon on a midweek day is usually the real quiet slot, and your call volume by hour will tell you which.

Step 3's cover person is not a formality either. A single-owner routine has a 100% failure rate during annual leave, and a fortnight missed in August is how a routine that was working in July stops existing by September. Name the cover when you name the owner, not when the leave is booked.

Step 4 is worth having as an actual conversation rather than an assumption. In most practices the honest list of things that may interrupt is a patient in distress and nothing else, and saying that out loud once is what gives the person doing the work permission to let the phone ring.

Step 5 reframes the failure usefully. A slot missed twice is information about your capacity, and the honest responses are to shorten it, move it, or accept that this work needs to be resourced differently. Quietly not doing it is the outcome this whole guide exists to prevent.

06Work the list in order and log the outcome#

Working in order matters because the segments have different value and a list worked in whatever order the export happened to produce will spend its best hour on its worst rows.

1. Start at the top of the Due segment and go down. Do not skim for easy names.
2. Log an outcome on every row, including no answer. A blank is not a record.
3. Use a small closed set of outcomes: booked, will call back, declined, no answer, unreachable.
4. Stop when the slot ends, not when the list ends, and start where you stopped next week.
5. Move anything marked unreachable twice into the lesson 7 queue.

Step 3 is the one that turns this into something measurable. Free-text notes cannot be counted, so a month of diligent work becomes unanalysable, and lesson 8 has nothing to compute against.

Step 4 is deliberate. A 90 minute slot that regularly runs to 3 hours will be cancelled within a month, and a slot that reliably ends on time survives. Working steadily through a list over 6 weeks beats an exhausting sprint that never recurs.

Step 1's instruction not to skim is the one people quietly break, and it is worth understanding why it matters. Picking the easy names first inflates your booking rate in week 1 and deflates it in week 4, so the trend you are trying to read in lesson 8 becomes an artefact of the order you worked in rather than anything about your patients.

Step 5 keeps the tail from silently expanding. Without it, the same unreachable rows resurface every week, get another attempt, and consume a slice of every slot forever.

Stop when the slot ends. A routine that overruns gets cancelled; one that finishes on time survives the quarter.

07Handle the patients who never answer#

Every list has a tail that does not respond to anything, and the tail is where practices waste the most effort. Decide in advance how many attempts each segment gets, and then honour the limit.

• Due and Overdue: 3 attempts across 3 different weeks, then park
• Lapsing: 2 attempts, then park
• Lapsed: 1 attempt, then park
• Mark parked patients with a date, so they re-enter next cycle rather than vanishing
• After 2 unreachable results, flag the record for a contact-details update at their next visit
• Never let an unanswered attempt silently become a fourth and fifth attempt

The last line is about how this fails in practice. Without a stated limit, a conscientious person keeps trying the same 30 people because they feel unfinished, and those 30 consume the slot that 200 other patients needed.

Parking with a date is what stops the tail becoming permanent. A patient who did not answer 3 times in March is a perfectly reasonable person to contact again in September, and a system that quietly excludes them forever has lost them for no reason.

The contact-details flag is the highest-value line on this list and the easiest to skip. A patient standing at the desk after an appointment will update a phone number in 15 seconds, and that is the only reliable moment it will ever happen. A practice that captures this consistently sees its unreachable share fall over 2 or 3 quarters without anyone doing dedicated data work.

08Measure two rates, then change exactly one thing#

At the end of each month compute the same two numbers, and resist computing more. Contacted rate is contacted divided by due. Booking rate is booked divided by contacted. They fail for different reasons and they have different fixes.

1. Compute both rates for the month, per segment, from the lesson 6 outcome log.
2. Write them on the same page as last month's, so you are reading a trend rather than a number.
3. Pick the lower-performing of the two and change one thing about it.
4. Change nothing else that month, so the comparison means something.
5. Repeat.

Step 3 is where the split earns its keep. A low contacted rate is capacity: the slot is too short, or it keeps getting eaten. A low booking rate is the message, the channel, or the timing, and adding staff time to it will change nothing.

Step 4 is the discipline everybody breaks. Changing the message and the channel and the slot in the same month produces an improvement you cannot attribute and therefore cannot repeat.

One caution about reading month 1. The first month of a routine that has not existed before will produce an unusually good booking rate, because the Due segment has a backlog of patients who were always going to say yes and simply had not been asked. Month 2 will look worse and will be closer to the truth. Practices that treat month 1 as the baseline spend month 3 wondering what broke.

The most common first finding is that the timing question is bigger than the wording question. Practices that try contacting at 2 weeks past due against 6 weeks past due usually see a larger difference than any rewrite produced, and that is a cheap experiment to run over 2 months. The wider measurement discipline this sits inside is in read your front office numbers.

What decays, and when to redo this#

The routine decays in three predictable places, and it is worth knowing which one is failing because they need completely different responses. The slot gets eaten by busier work, the message goes stale and stops being read, and the list quietly refills with the records that lesson 1 removed.

The slot is the one to watch. It fails gradually rather than suddenly: shortened once for a good reason, then again, then moved, and 3 months later nobody can say when it last ran. The monthly rates catch it, which is another reason to compute them even in months when nothing feels different.

The message decays differently and more slowly. It does not stop working so much as stop being noticed, particularly by the patients who have now received it 3 or 4 times across successive cycles. Those are the people your attempt limits keep bringing back, so they see the wording more often than anyone, and a small rewrite once a year is aimed mostly at them.

Re-run lesson 1's cleaning every 6 months, and re-read the message every 12. Redo the whole thing after any change to your recall policy, your practice management system, or the person who owns the slot, because all three invalidate parts of what you set up. If the diary is consistently full when the slot comes round, stop and read the second paragraph of this guide again, since that is a good problem and this is no longer the right work.