Staffing is the largest cost of a call centre by a wide margin. The software is a rounding error next to the people. So it is worth doing the arithmetic rather than guessing.
You need two numbers, and you probably already have both.
The two numbers
Calls per day. Count them for a week. If you are on mobile phones, the call log gives you this. Do not estimate — people always estimate low, because the calls they remember are the ones they answered.
Average handling time. How long a call takes, start to finish, including the note written afterwards. For most Kenyan support desks this lands between three and six minutes. Time twenty calls and take the middle.
The simple version
Multiply them, and you have minutes of work per day.
120 calls × 4 minutes = 480 minutes = 8 hours of talking.
That looks like one agent for an eight-hour day. It is not, for two reasons.
Nobody talks for eight hours. Breaks, lunch, writing things up, the conversation that goes nowhere. A realistic figure is about 70% of a shift on calls, and that is a busy desk. So 8 hours of talking needs about 11.5 hours of shift.
Calls do not arrive evenly. This is the part that catches people out, and it is the whole reason queues exist.
The part that actually decides it
Calls arrive in a shape. Almost every Kenyan support desk looks roughly like this: quiet at eight, busy from ten, a dip over lunch, a second peak around four, falling away after five.
Your staffing is not set by the daily total. It is set by the busiest hour.
If 120 calls a day means 25 of them land between ten and eleven, then at 4 minutes each that hour contains 100 minutes of work. One agent cannot do 100 minutes of work in 60 minutes. Two can, just. Three can do it without anyone waiting long.
So the calculation is:
- Calls in your busiest hour
- Times average handling time
- Divided by 60
- Plus roughly half an agent of headroom
For our example: 25 × 4 ÷ 60 = 1.7, plus headroom, so two to three agents at peak.
If you do not know your busiest hour, assume it holds about 15 to 20% of the day’s calls. Then measure properly once you have a system that reports it.
Why three is the honest floor
Below three people, a queue has nothing to queue for.
With one agent, a queue is just a polite way of saying engaged. With two, one person at lunch halves your capacity. With three, somebody is almost always free, which is the entire thing you are buying.
That is why our smallest plan is three seats. Not as a sales tactic — below that, the software solves less than you would hope, and a good voicemail plus a callback habit would serve you nearly as well.
What changes the answer
Longer calls. A technical support desk at 12 minutes a call needs three times the people of a 4-minute order desk at the same volume.
Spiky months. A SACCO at month end, a school in admissions season, a retailer at Christmas. Staff for the normal week and have a plan for the spike — overtime, a borrowed colleague, or a callback offer — rather than carrying peak headcount all year. There is an article on the month-end case.
Outbound mixed in. If the same people also make calls, outbound work fills the quiet hours but competes with inbound at the peak. Decide which wins, explicitly, or the agents will decide for you.
A good menu. A menu that answers the three commonest questions without a person removes them from the arithmetic entirely. This is usually cheaper than hiring. More on writing one.
What to do with the answer
Two warnings.
Do not staff for the worst day of the year. You will carry idle people for eleven months. Staff for a normal busy hour and plan for the exceptions separately.
Do not staff for the average. Average staffing means half your days have a queue nobody can clear. The average is the least useful number in this exercise.
Aim at the busy hour of a normal week, add headroom, then watch the abandoned-call figure for a month and adjust. The numbers worth watching are here.