Dental lead response time: how to measure and fix it

Define what you are measuring first

Response time is only a useful number if everyone agrees what starts the clock and what stops it. Most practices that "measure response time" are quietly measuring four different things and averaging them, which produces a number that cannot be acted on.

We suggest these definitions. If you prefer different ones, use yours — just write them down before you start, because the comparison over time is worth more than the absolute figure.

• The clock starts when the enquiry first reaches your practice: the moment the phone rang, or the moment the form was submitted. Not when it was assigned. Not when someone opened it.
• The clock stops at first meaningful two-way contact — the patient actually spoke to someone, or replied to a text. An outbound call that rang out is not a response. An automated "we got your message" is not a response either; it is a holding acknowledgement, and it is worth tracking separately.
• A lead is a new or returning patient enquiring about care they have not yet booked. An existing patient confirming Thursday is not a lead and will distort everything if you leave it in.
• Count in wall-clock time, not business hours. This is the one practices resist most. A Saturday 7pm enquiry answered Monday at 9:15am has a response time of about 62 hours, not 15 minutes. The patient experienced 62 hours. Measure what the patient experienced, then segment separately to understand the causes.

Step 1: pull the raw data

You need two exports and, ideally, a third. Everything here is data your practice already holds.

1. The call log from your phone system. Whatever you run — a VoIP platform, a call tracking number on your ads, the carrier portal — will export a CSV with the timestamp, the direction, the caller number, the duration, and whether it was answered. Pull a full ninety days. Thirty is too few to see hour-of-day patterns; a year invites seasonal noise you cannot interpret yet.
2. Web form submissions. Your site's form tool, your inbox, or your CRM. You need the submission timestamp and the contact identifier. If your forms only email you, the email timestamps are your data — export the folder.
3. Your outbound activity. The same phone export usually contains outbound calls, which is how you find the response. If you also reply by SMS, you need that log too, or your measured times will look worse than reality.

Put all three in one spreadsheet, one row per enquiry, with columns for: enquiry timestamp, channel, contact identifier, first outbound attempt timestamp, first connected contact timestamp, and a call-type column you will fill in shortly. If your practice runs call tracking numbers per campaign, keep the source column — it becomes useful later.

Step 2: strip out what is not a lead

This is where most measurement attempts go wrong, and it is the step that takes the longest. A raw call log is mostly not leads.

• Remove existing-patient calls. Rescheduling, results, billing, and hygiene recalls. Match on number against your patient list where you can; eyeball the rest.
• Remove spam, sales calls, and wrong numbers. Very short duration plus an unrecognised number is a decent first filter, but check the short ones by hand — a missed new-patient call is also short.
• Remove internal and supplier calls. Labs, suppliers, referring practices. These are a bigger share of a dental line than people expect.
• Deduplicate. The same person calling three times in ten minutes is one enquiry with a response time measured from the first attempt. Counting it three times makes your data look better and is the single most common self-deception in this exercise.
• Keep the missed calls that never came back. Do not delete them because there is no response timestamp. They are the most important rows in the file — see the next section.

Expect to lose a large share of your raw rows. That is normal and it is the point. What remains is your actual new-enquiry volume, which is usually smaller than the practice believed and more valuable per row than the practice assumed.

Step 3: compute the median, not the mean

Compute both. Then use the median, and understand why they differ.

The mean — the ordinary average — is pulled around by extremes in both directions. A dental enquiry log has severe extremes: a cluster of enquiries answered in under a minute because the desk was free, and a tail of enquiries answered two or three days later because they arrived on a Friday night. Averaging those produces a number that describes no actual patient. Worse, the tail can be dragged around by a handful of rows, so your mean will swing month to month for reasons that have nothing to do with how your practice is performing.

The median is the middle row when you sort every enquiry by response time. It tells you what a typical enquirer actually experienced, and it moves only when your practice's behaviour genuinely changes. That stability is the entire reason to prefer it.

Compute four numbers and keep all four:

• Median response time. Your headline figure. This is what a typical enquirer waited.
• 90th percentile. Sort by response time and read the row nine-tenths of the way down. This is your bad-day number, and it is where the lost cases live. If the median is fine and the 90th percentile is measured in days, you have a queueing problem, not a staffing problem.
• Never-responded count. The rows with no response timestamp at all, as a share of total leads. Practices are consistently shocked by this one. It cannot be averaged into anything — report it separately, as a count and a percentage.
• Median inside business hours only. Computed on the subset that arrived while you were staffed. The gap between this and your headline median is exactly the size of your after-hours problem, stated in hours.

Step 4: segment by hour and by call type

A single median for the whole practice tells you whether you have a problem. Segmentation tells you where it is. Two cuts do almost all the work.

By hour of day and day of week. Build a grid: days down the side, hour bands across the top, median response time and enquiry count in each cell. You are looking for two things — the cells where volume is highest, and the cells where response time is worst. They are rarely the same cells, and the overlap between them is where you should spend money first.

By call type. Tag every remaining row as urgent, elective, or routine. Urgent is a patient reporting a problem needing prompt attention as your practice defines it. Elective is a new enquiry about cosmetic, implant, orthodontic, or other planned treatment. Routine is a new-patient check-up or hygiene enquiry. This tagging takes an hour and is the highest-value hour in the whole exercise, because the three types have completely different economics and completely different acceptable response times.

The Monday morning queue problem

This is the most reliably broken thing in a dental practice's response data, and it is structural rather than anyone's fault.

Enquiries arrive across the weekend. Nobody works them. On Monday morning the practice opens and the live phone starts ringing immediately — Monday is typically the busiest phone morning of the week, carrying its own weekend backlog of patients who waited to call. Your front desk now faces two queues at once: the live one that is ringing, and the accumulated one sitting in a voicemail box and a form inbox.

The live queue always wins, because a ringing phone is louder than a list. So the weekend enquiries get worked in gaps — late morning, over lunch, or on Tuesday. By then, a Saturday-evening implant enquirer has had roughly two days of silence, during which she almost certainly contacted another practice. And the practice will still describe its Monday as a good, busy morning, because every live call got answered.

Three things make it worse than it looks in the raw numbers:

• The queue is worked last-in-first-out. Whoever is picking through the voicemail box tends to start at the top, which is the most recent message. The Friday-evening enquiry — the oldest and coldest — is worked last, which is exactly backwards.
• Weekend arrivals skew elective. People research veneers and implants on Saturday night, not on Tuesday at 10am. So the queue that gets worked last is disproportionately full of the highest-value enquiries you received all week.
• The tail is invisible in the mean. If you only ever looked at an average response time across the month, the weekend tail is smeared into it and disappears. Segmenting by arrival window is what makes it visible.

The diagnostic is simple: compute median response time for enquiries that arrived Friday 5pm to Sunday midnight, and compare it against enquiries that arrived Tuesday to Thursday during staffed hours. The ratio between those two numbers is the size of your queueing problem, and it is a number you can act on. If the answer is uncomfortable, the way the coverage layers stack is the next thing to read.

Step 5: read the table honestly

You now have a headline median, a 90th percentile, a never-responded count, and a grid of segments. Before deciding anything, three cautions.

• Small segments are noise. A cell with four enquiries in it has a median that means nothing. Widen the window or merge the cell rather than making a decision on it.
• A good median can hide a bad business. If your median is four minutes and your never-responded count is one in six, the median is not the story. The unanswered rows are.
• Correlating response time with bookings is tempting and mostly unsound at practice scale. You do not have enough rows to separate the effect of speed from the effect of the enquiry type, the source, the clinician, or the season. Use the segments to find the operational gap; do not build a conversion model out of ninety days of one practice's data and then trust it.

What the exercise reliably gives you is a map of where enquiries wait and how long. That is enough to act on, and it is enough without inventing a benchmark. We would rather you have your own honest number than a confident industry figure that nobody can source.

Fixes, in order of cost

Work down this list. Most practices find something in the first two items, and the first two cost nothing.

1. Reorder the Monday queue. Work the weekend backlog oldest-first, and assign it to a named person for a fixed block before the phones get loud. Free, and it usually moves the 90th percentile more than anything else on this list.
2. Give the form inbox an owner and a schedule. "Someone checks it" means nobody does. A named owner and defined check times fixes the worst segment in most practices' data at zero cost.
3. Send a real holding message on web forms. An immediate automated acknowledgement that says specifically when a human will make contact is not a response, and you should keep measuring it separately — but it buys patience, and it stops the enquirer assuming the form never arrived.
4. Fix the overflow path. Look at how often a second line rang while the first was engaged. If concurrency is your problem, more staffing hours will not help; call routing will.
5. Cover the after-hours window. If your after-hours median is measured in tens of hours and your elective enquiries land there, this is the gap worth spending money on — whether with an answering service or a structured intake system. Which one depends on whether a message is enough or a booked consultation is needed.
6. Add follow-up for enquiries that did not book. Response time only covers first contact. An enquiry answered in three minutes and never contacted again is still a lost case — see dental lead follow-up.

Keeping it measured

Re-run this monthly, on the same definitions, and keep the four headline numbers in one row per month. Twelve rows of consistently-defined median, 90th percentile, never-responded count, and after-hours median will tell you more about your practice's front end than any dashboard you could buy.

Two rules make the tracking worth having. Never change your definitions mid-year — if you must, keep the old definition running alongside for three months. And never replace the median with the mean because the mean looks better in a partners' meeting.

Questions practices ask

Related reading