The usual answer divides monthly calls by calls-per-rep. That method cannot see the half-hours where the money actually leaks, which is why operations that pass it still miss a quarter of their calls.
Ask how many customer service reps a home-service operation needs and you will usually get a ratio. Two thousand calls a month, a rep handles a thousand, therefore two reps. Sometimes it is dressed up with a utilization factor. It is always arithmetic on a monthly total.
The arithmetic is not wrong so much as it is answering a different question. It tells you how many rep-hours it takes to handle your call volume. It says nothing about whether anyone was there when the calls arrived, and that is the only version of the question that costs money.
Averages are where capacity loss goes to hide. A month is roughly 500 open half-hours. Your calls are not spread across them; they pile into the first hour of Monday, the hour after the first cold snap, the window right after your radio spot runs. A monthly average answer time of nineteen seconds is entirely compatible with a Monday at 8:30 where a fifth of callers gave up waiting.
Published figures put the average home-service operation somewhere between a quarter and a third of inbound calls missed, rising in peak season while headcount stays flat. Whatever the exact figure is for your business, note the shape of it: the miss rate goes up precisely when demand does. That is a staffing-placement problem, not a staffing-quantity problem, and a monthly ratio cannot distinguish them.
Staffing a queue has been a solved problem in telephony for a century. It needs three inputs, and only one of them is call volume.
Here is the part that makes the ratio method fail rather than merely simplify. The relationship between staffing and abandonment is not a line. It is a curve that is nearly flat while you have slack and then falls off sharply, which means the same additional rep is worth wildly different amounts depending on where you already sit.
Modelled at two erlangs of offered load in a half hour, roughly twelve calls at five minutes each, with callers whose average patience is three and a half minutes:
| Reps on the phones | Callers who hang up | Answered within 30s |
|---|---|---|
| 2 | 29% | 49% |
| 3 | 12% | 74% |
| 4 | 4% | 89% |
| 5 | 1% | 96% |
Illustrative. This is the queueing model evaluated at stated inputs, not measured data from a client. The point is the shape, not the digits.
Read the third row against the second. One more rep in that half hour removes two thirds of the abandonment. Read the fifth against the fourth: the same rep removes almost nothing, because there was almost nothing left to remove. A ratio method cannot tell you which of those two rows you are standing on. It gives the same answer either way.
This is also why "we added someone and nothing changed" and "we added someone and it transformed" are both common and both true. They are different rows.
There are two defensible answers and they usually disagree, which is worth knowing before someone tells you there is one.
The service-target answer is the cheapest roster that keeps abandonment under your ceiling in every interval. It is what workforce-management tools in other industries compute. It is a real answer, and it is often more expensive than it needs to be, because the last few points of service level are bought at a steep price.
The economic answer is the roster where the next rep-hour stops paying for itself: keep adding while the contribution margin recovered exceeds the loaded cost of the hour, and stop when it does not. That requires pricing both sides: what a rep-hour costs you fully loaded, and what a prevented hang-up is actually worth once you discount for the callers who would have rung back anyway, the calls that were never a job, and the ones that would not have booked.
Both are computable from records you already have. Neither is computable from a monthly total.
You do not need us to start. Three things, in order of how much they will tell you:
Queueing models assume things about your business that are not exactly true: that calls arrive independently, that handle times follow a particular distribution, that a half hour is long enough to settle. On real home-service data those assumptions hold well enough to be useful and not so well that the absolute numbers should be quoted to the dollar without checking them against what actually happened. The right use of the model is comparative: this interval versus that one, this roster versus that roster , where the errors largely cancel.
Anyone who hands you a staffing number without telling you that is selling you something. Our version of the arithmetic is published, including the thresholds, so you can check it rather than take it.
Short, specific, and about arithmetic you can check: how phone operations are measured, where the standard numbers mislead, and what the data actually supports. No case studies we have not earned, and no sales sequence.
We use it to send the journal and nothing else. Unsubscribe from any issue. See the privacy notice.