The real cost of a missed call at a small practice

A missed call at a practice is not a missed message

In most businesses a missed call is a delay. Somebody leaves a voicemail, you call back Tuesday, the work is still there waiting for you. A practice is not most businesses.

The person calling you is in the middle of deciding something. A molar started hurting on Saturday. Their old orthodontist retired. A new insurance card arrived and they are working down a list of in-network offices in whatever order the plan's website printed them. They did not call to speak to you specifically. They called to get a problem handled, ideally today.

So when the phone rings out, nothing happens that anybody in your office can see. No voicemail, no note, no lead in the system. The caller taps the next number on the list, somebody there picks up, and the appointment they were going to give you is on another practice's schedule before your front desk is free again. There was no second chance because there was never a first contact.

That is what makes this expensive and invisible at the same time. A no-show shows up. An unpaid claim shows up. A stalled treatment plan shows up. A person who called, got nothing, and went elsewhere shows up in no report you will ever run. The only trace is a line in the phone system log — which nobody reads, and which is exactly where this article starts.

The four moments the phone goes unanswered

Practices differ in almost every way and are identical in this one. The dead spots are the same four everywhere, and they are structural — not a discipline problem, and mostly not a hiring problem.

• While the front desk is with a patient. Somebody is at the counter checking out or rescheduling for their daughter. The person standing in front of you wins, and they should. That is the right call, and it costs you the other one.
• Lunch. Your desk eats at the same hour your callers do. Working adults phone a dentist on their own break — precisely the window when one person is covering the phone, or nobody is.
• After hours and weekends. Pain does not check your posted hours. Neither does a bracket that breaks Friday night, a crown that comes off Saturday morning, or a parent with ten quiet minutes on a Sunday evening.
• The Monday and post-holiday surge. Everything that happened while you were closed arrives at once when you open. Staffing is flat across the week; demand is not. You get a queue at the moment callers have the most urgency and the least patience.

Only the first of those falls inside a normal working hour with a normal working body available. The other three are hours when covering the phone with a person is expensive, impossible, or both — so any fix that assumes a human at a desk is addressing one of the four moments.

Work out your own number

Nobody can tell you what your missed calls cost. Not a vendor, not a webinar, not a statistic on a slide, and not this article. The inputs are internal to your practice: three of the four already sit in systems you own, and the fourth is a twenty-minute conversation with whoever runs your schedule.

1. U — unanswered inbound calls in a normal week. Your phone system knows. Nearly every VoIP platform reports inbound total, answered, abandoned and after-hours separately. Pull four weeks and pick a typical one — not the week after a holiday. If the system cannot produce that report at all, that is finding number one.
2. N — the share of those that were new-patient inquiries. Existing patients call back. Vendors call back. Strangers do not. Estimate N honestly; your desk has a good sense of how much of the ringing comes from numbers the practice has never seen, and some phone systems will flag unknown numbers outright.
3. B — the share of new-patient inquiries your desk normally books. Take this from the calls you did answer, not from an industry figure. It depends on your fees, your availability next week, and how your phone gets answered.
4. V — what one patient is worth to you. Not the first visit: the whole course of care, the hygiene recall, and the family members who follow a happy patient. Your practice management system can produce a version of this, and your accountant can sanity-check it.

Then multiply straight through. U times N gives the new-patient inquiries that went unanswered. Times B gives the patients you would have booked. Times V gives money.

Run it twice — once with the most pessimistic estimate of N and B you can defend, once with the most generous — then look only at the pessimistic answer. If even the low version would justify hiring somebody, you have your answer and you needed nobody's benchmark to get it. If the low version is small, that is useful too: your phone is not the leak, and you should go look at recall or treatment acceptance instead.

The arithmetic is deliberately conservative. It ignores the existing patient who could not get through to reschedule and quietly stopped coming. It ignores referrals. It ignores the marketing you paid for to generate the call that then rang out unanswered — the part that usually stings once it is written down.

Why the usual fixes underperform

Every practice has tried at least two of these. None of them are stupid. They each solve one slice of the problem and stop short of the thing the caller actually wanted.

The fix	What it actually solves	What it leaves on the table
Voicemail	Gives a caller already committed to your practice a way to reach you out of hours.	A caller with four other numbers in front of them does not leave one. A mailbox cannot answer a question or hold a slot — it converts a booking into tomorrow's callback list.
A call-back list	Makes the misses visible, which is more than most practices have.	It is a to-do list competing with patients in the chair, and it loses. By the time anyone works it, the caller has been seen elsewhere. It records the problem; it does not fix it.
An answering service	A human voice picks up out of hours and a message gets taken.	They cannot see your schedule and cannot book. The caller has to be phoned back and asked to pick a time again — two conversations to do the work of one — and the service does not know your answers.
Hiring a second front-desk person	Real capacity during the hours they are at the desk, plus hands for everything else.	A fixed monthly cost aimed at a spiky problem that peaks outside the hours they work. They take lunch, get sick, and do not work Sunday night — three of the four moments still uncovered.

The common thread is easy to miss: every option above ends in we will get back to you. The caller did not want to be got back to. They wanted an appointment. Anything that turns a booking into a callback has handed you a second chance to lose the same person, and second chances convert worse than first ones.

What a good answer looks like

Judge whatever you put on the end of that line — software, service or human — against five things.

• It picks up immediately, every time. No queue, no lunch, and no difference between Tuesday at ten and Sunday at two in the morning. A system that answers most of the time is a faster version of the problem you already have.
• It can see the schedule and write to it. Reading real availability out of your practice management system and booking into it — plus rescheduling and canceling — is the line between a message-taker and a receptionist.
• It speaks the caller's language. In most markets that means English and Spanish handled natively in the same call, not a hold and a transfer to whoever is bilingual and at their desk today.
• It knows your real answers to the five questions everyone asks. Are you taking new patients. Do you take my plan. Where are you and where do I park. What happens at a first visit. Can I be seen today. Those answers are yours, and they change — somebody has to own keeping them current.
• It hands off to a human when it should. An emergency, an upset caller, anything outside what it was built for, and above all a caller who asks for a person. Quick and warm, not a menu maze. Every call should leave a transcript somebody can read afterward.

If it also texts a confirmation after the call — and it should, because a spoken time gets forgotten and a text does not — treat that text as its own program with its own rules: consent captured on the call, the caller's own number only, STOP and HELP handled, carrier registration done before the first send. We wrote up what actually gets an A2P 10DLC campaign approved.

What to measure once it is fixed

Three numbers. The first comes from the same phone report that gave you U.

• Answer rate by hour of the day. Answered over total inbound. You want the curve to go flat — no lunch dip, no Monday trench, no cliff at closing time. Do not panic when total inbound appears to rise: you are now counting calls that were always happening and were never visible.
• Appointments booked outside office hours. The cleanest signal you will get, because before the change it was zero by definition. If it is still near zero after a month, the problem is in the booking path, not the phone.
• Where the callers who used to vanish are landing. Tag the source on new patients and ask two things: are agent-handled calls converting like desk-handled ones, and where in the schedule are they landing. If everything lands in one narrow window, the availability rules need work.

Then keep listening. Read the transcripts weekly for the first month — not to catch the system out, but because every question it fumbled is a gap in the answers you gave it, and most gaps take an afternoon to close.

What it should never do

An honest note, because these failure modes are worse than a missed call.

• It should not diagnose. When a caller describes symptoms and asks whether they need to come in, the job is to route them to the right kind of visit and escalate anything urgent to a clinician — not to assess. That line belongs in code, not in a politely worded instruction.
• It should not discuss a chart with an unverified caller. Knowing a patient's name does not identify anybody. Scheduling actions get gated behind verification, and anything clinical goes to a person. For a practice this is a design question rather than a manners question — we walked the whole call path in putting an AI on the front desk without breaking HIPAA.
• It should not pretend to be a person. If a caller asks whether they are talking to a human, the answer is no, immediately, followed by an offer to get them one. Everything else about the call can be excellent and it will not matter; a voice that lies about what it is spends trust you took years to build.

Start with the log, not the software

Order matters. Pull the phone report before you talk to anybody, including us. Get U for a normal week. Sit down with your front desk for twenty minutes and get honest estimates for N and B. Ask your practice management system for V. Multiply. Now you have a number that belongs to you.

If it is small, leave the phone alone and go find your real leak. If it is not, the shape of the fix follows from the four moments: coverage in the hours a human cannot economically cover, able to book rather than take a message.

That is what our AI receptionist is built for — answering in English and Spanish, booking straight into the practice management system, escalating emergencies to a person, and sending one consent-gated confirmation text at hang-up. If you would rather talk it through with your own numbers in hand, a discovery call is free and thirty minutes long, or you can tell us what your week looks like and we will come back with a scope.