A call centre does not grow smoothly. It works, then something breaks, then it works again at the next size.
The useful thing is that the breakages are predictable. Here is roughly what goes wrong, and when.
Three to five agents
What works. Everything, informally. One queue, one menu, everybody knows everything, the manager hears most of the calls because they are in the room.
What breaks first. Cover. One person at lunch is 20% of your capacity. One person ill is a bad day.
What to put in now, while it is easy.
- A menu, even a simple one
- A record of every call, with a note and an outcome
- The weekly report, looked at by somebody
That last one matters. The habit of looking at the numbers is far easier to build at five agents than at fifteen, and a desk that has never looked will not start.
Six to ten
What breaks. The single queue. Different calls now need different people — sales does not want support’s calls and is not good at them.
What to do. Split into two or three teams with their own queues and routing. Keep it simple: a team per obvious category, not a team per person.
What also breaks, quietly. The manager stops hearing the calls. At five they overheard everything; at ten they are in meetings. This is the point at which call recording stops being a compliance feature and starts being the only way anybody knows what is being said.
Start listening to three calls a week, deliberately. It takes twenty minutes and it is the only thing that stops quality drifting.
Eleven to twenty
What breaks. Supervision. One person cannot manage twenty agents and also do their own job. Somewhere around twelve to fifteen you need a supervisor whose actual role is the floor.
What also breaks. Training by osmosis. New agents learned by sitting next to good ones; now there are too many new ones and the good ones are busy. The first week needs designing.
And reporting changes shape. At five agents you look at totals. At twenty you need the breakdown — by team, by hour, by agent — because the average now hides everything interesting.
What to put in. Skills-based routing, so the Kiswahili call reaches someone who speaks it and the billing question reaches someone who knows. A wallboard the supervisor can see. Scorecards, used for coaching.
Twenty to thirty
What breaks. Scheduling. Up to twenty you can staff by common sense. Beyond it, you need to look at the arrival pattern properly and build shifts against it, or you will be overstaffed at three and drowning at ten.
What also breaks. Consistency. Thirty people answer the same question thirty ways. This is where a short written answer to the twenty commonest questions earns its keep — not a script, which makes people sound like machines, but an agreed answer.
And churn becomes a system rather than an event. At thirty agents somebody is always leaving. Recruitment stops being a project and becomes a continuous process, which it should be resourced as. The costs are here.
What should not change
Two things should survive every stage.
Every call leaves a record. The note and the outcome. This is what lets thirty people serve one customer coherently.
Somebody listens to calls every week. Numbers tell you where to look; only calls tell you what is happening. The desks that lose this at fifteen agents are the ones whose quality quietly collapses at twenty-five.
A note on the software
Everything described here — several queues, skills routing, wallboards, scorecards, the reporting breakdown — should be on the plan you already have, not sold to you again at each stage.
That is why every feature is on every plan with us: the three-seat desk gets the same system as the fifty-seat floor, and growing means adding seats rather than renegotiating. The plans are here.