A front desk call volume model: when calls go unanswered, and why

Why daily totals mislead

A practice that receives 120 calls a day across nine open hours averages about 13 calls an hour. That sounds manageable for two people. But calls don't arrive evenly. They pile up when the office opens, when patients take their lunch break, and late in the afternoon when people leave work. Meanwhile the front desk is busiest with in-person patients at the very same times.

So the question is never "can we handle 120 calls a day?" It is "can we handle the 24 calls that arrive between 8:00 and 9:00 while also checking in the first patients of the day?"

The inputs you need

You can build the model with a phone system report and a staffing schedule. Most business phone systems can export a call log with a timestamp and an answered or missed status.

• Calls by hour of day: total inbound calls in each hour, averaged over at least four weeks so one odd day doesn't skew it.
• Average handle time: how long an answered call takes, including any after-call work like entering a booking. Your phone report may show talk time; add a minute or two for the follow-up work.
• Staff on the desk by hour: how many people are scheduled at the front, hour by hour, including lunch coverage.
• Phone share of each person's time: the fraction of each hour a person can realistically spend on the phone after in-person work. This is an estimate; time it for a few days if you can.

Building the model, hour by hour

For each hour, calculate demand and capacity in minutes.

Demand minutes = calls in the hour x average handle time.

Capacity minutes = staff on the desk x 60 x phone share.

If demand exceeds capacity, the hour is over capacity, and some calls will ring out, go to voicemail or sit on hold until the caller hangs up. A rough estimate of calls at risk is (demand minutes - capacity minutes) / average handle time.

Here is an illustrative example for a two-person desk with a 4-minute average handle time:

• 8:00-9:00: 24 calls = 96 demand minutes. Two staff at 40% phone share = 48 capacity minutes. Gap 48 minutes, about 12 calls at risk.
• 10:00-11:00: 14 calls = 56 demand minutes. Two staff at 60% phone share = 72 capacity minutes. No gap.
• 12:00-1:00: 18 calls = 72 demand minutes. One staff on lunch coverage at 70% = 42 capacity minutes. Gap 30 minutes, about 8 calls at risk.
• 4:00-5:00: 17 calls = 68 demand minutes. Two staff at 45% (check-outs) = 54 capacity minutes. Gap 14 minutes, about 4 calls at risk.

From calls at risk to dollars

Not every missed call is lost revenue. Some callers want directions, some are vendors, some will call right back. To get to dollars, chain three assumptions, each one labeled and, ideally, measured from your own call log.

In the illustrative example, the three problem hours add up to 24 calls at risk per day. Suppose, for illustration, that 35% of calls are from people trying to book or reschedule, that half of those callers do not call back, and that the average visit collects $165. Then: 24 x 0.35 x 0.5 = 4.2 lost visits a day. At $165, that is about $693 a day, or roughly $14,500 a month across 21 working days.

Those percentages are placeholders. The point of the model is not the specific number but the structure: it tells you which hours to fix, and it shows that even small changes in coverage at the right hour move real money.

What the model tells you to do

1. Move work out of peak hours. Confirmation calls, insurance follow-ups and callbacks made by the desk should happen in quiet hours, not at 8:00 AM.
2. Stagger lunches against the call curve. If the lunch hour is a peak, cover it fully and send people to lunch at 11:00 and 1:30 instead.
3. Answer missed calls immediately, automatically. A caller who gets a text with open times within a minute is far more likely to book than one who gets a voicemail greeting.
4. Track after-hours separately. Calls before opening, after close and on weekends are a pure gap: capacity is zero. Count them on their own line.
5. Re-run the model quarterly. Call patterns shift with seasons, new providers and schedule changes.

Checking your model against reality

Once you've built the model, compare its predictions with your phone system's actual missed-call counts by hour. If the model predicts 12 calls at risk at 8:00 and your log shows 3, your phone-share estimate is probably too low. If it predicts zero and your log shows 10, your handle time is probably too short or the desk is pulled away more than you thought. Adjust until the model matches your history, and then you can use it to test changes before you make them.