Forty calls between 5:45 and 7:15 on a Friday, averaging a bit under three minutes each, is close to two hours of talking compressed into ninety minutes of clock time. One person cannot do that and also seat a dining room. That arithmetic is the whole staffing question for the phone, and most schedules are written without anyone performing it.
Owners tend to staff the phone by feel, which usually means staffing it by whoever is standing nearest. Doing the math instead takes about twenty minutes and gives you a number you can defend to a partner, a landlord, or yourself at the end of a bad month.
Talk minutes are the unit, not call counts
A call count tells you how busy the phone sounds. Talk minutes tell you how many humans it takes.
Pull your busiest hour from your phone system or your POS timestamps. Say it is 40 calls. Then measure average handle time, which is the whole call from pickup to hang-up, plus whatever the person does afterward before they are available again. For an order call that includes reading the ticket back, entering it, and sometimes walking to the printer.
Time twenty real calls with a stopwatch during an actual rush. Do not estimate. Nearly every owner who does this comes back with a number higher than they expected, because the dead air while a caller decides between two sizes feels like nothing to the person listening and is thirty seconds on the clock.
Forty calls at 2.8 minutes is 112 talk minutes inside a 60-minute hour. That is 1.87 people fully occupied by the phone and nothing else. The methodology for getting the raw call counts is in how many calls does your restaurant miss, and it matters that you use offered calls rather than answered calls, because the ones that rang out are still demand.
The staffing number is not the talk-minute number
Here is where most operators stop, and where the plan starts to fail.
If you staff 1.87 people for 1.87 people worth of talking, you have built a system with 100 percent planned occupancy. Every call would have to arrive exactly on schedule for that to work. Calls do not do that. They cluster around the moment a family finishes deciding, around the end of a shift downtown, around whatever your neighborhood's habits are.
The practical planning rule from queueing behavior is that waiting time climbs steeply as occupancy approaches full. Keeping planned occupancy around 70 percent leaves enough slack for the clusters. So divide: 1.87 divided by 0.7 is 2.7 people on the phone for that hour.
Round up. You need three.
That is the number that surprises people. A restaurant taking 40 calls in its busiest hour needs the equivalent of three dedicated phone positions to answer them all promptly, and almost no independent restaurant has ever scheduled that. Which is why the phone rings out at 6:30 in nearly every restaurant in the country.
What that position actually costs
Now price it. Take your fully loaded hourly rate for a host or counter person, meaning wage plus payroll taxes plus whatever your workers' comp and benefits load runs. Multiply by the number of peak hours per week where the math says you are short.
If you are short two positions for three hours a night, six nights a week, that is 36 hours a week. At a fully loaded rate somewhere in the high teens to low twenties per hour, you are looking at a real monthly number in the low thousands. Run it with your own rate rather than mine.
Then compare that against what you are currently doing, which is nothing, and against what the alternatives cost. The comparison against hiring is worked through in voice AI versus hiring another host, and the general labor arithmetic sits in cutting restaurant labor costs with AI phone answering.
The honest framing is that you have three choices: pay for the coverage, buy software that answers every line simultaneously, or keep losing the calls. Most operators are choosing the third one without knowing they made a choice.
Where this math misleads you
Three places, and they all push in the same direction.
The first is that a host answering the phone is not a phone position. They are a divided position, and their real availability to the phone during a seating rush might be 40 percent of their minutes. If you count them as one full phone person, your plan is wrong before it starts. That dynamic is the subject of reducing host stand phone interruptions.
The second is that handle time is not constant. A Saturday catering inquiry runs eight minutes. A regular ordering the usual runs forty seconds. If your call mix shifts seasonally, your staffing number shifts with it.
The third is that suppressed demand hides in your data. If your regulars have learned that you never answer between 6 and 7, they stopped calling then. Your call counts are lower than your actual demand, and the staffing math built on them understates the problem. You typically only find out when coverage improves and volume goes up without any marketing.
Running the number for your own week
Pick one week. For each hour of each day, record calls offered and multiply by your measured handle time. Divide by 60, divide by 0.7, and you have an hour-by-hour requirement in people. Lay that against your actual schedule.
Most operators find two or three specific windows where the gap is large and the rest of the week is fine. That is a more useful result than an average, because it tells you exactly where to spend. Averaging across the week produces a number that describes no hour you actually operate.
The gap windows are also the ones where the missed calls are worth the most, since a Friday 6:30 caller is more likely to be a full-ticket order than a Tuesday 3pm caller asking about parking. What that unanswered call is worth is worked out in the real cost of a missed restaurant phone call.
The decision this leaves you with
Do the arithmetic before you shop for anything. If your peak hour needs three phone positions and you have one distracted host, no vendor pitch and no new headset changes that. Either you add people, you add something that answers in parallel, or you accept the loss and stop calling it a mystery.
Run the numbers this week, write the hour-by-hour requirement on one page, and put it next to your schedule. If the two documents disagree by more than half a person in any window you care about, you have found your next operational decision, and you have the math to back whatever you choose. Pricing for the software path starts at $250 a month, which is a useful benchmark to hold your labor number against.