Field notes · Phones
7 September 2026
Nobody can tell you your missed-call rate. Including us.
If a vendor has told you what percentage of your missed calls they will recover, they made that number up. Not maliciously, in most cases. They modelled it, the model got rounded into a slide, and the slide went on a sales call. But the measurement they would need to produce it never reached them, because the phone network does not send it.
What the carrier actually forwards
Conditional call forwarding is the mechanism behind almost every AI receptionist. Your line forwards to a second number when it is busy, when nobody picks up, or when you are closed. It is a good mechanism. It is also nearly silent about itself.
When that forwarded call arrives, the receiving system is told a call arrived. It is not told which of the three conditions fired. Busy and no-answer are indistinguishable on arrival. The field that would identify the original line is passed inconsistently, and plenty of carriers omit it. So the receiving end knows a call came, and knows almost nothing about why.
That matters because “calls you missed” and “calls that were forwarded” are not the same set, and the gap between them is exactly where a recovery percentage would have to come from.
What that rules out
It rules out any honest claim of the following shape:
- i“We recover sixty per cent of your missed calls.” Recovered against what denominator? The denominator is the thing nobody has.
- ii“Your practice misses forty calls a week.” Before a system is answering them, a missed call leaves no record anywhere. That is what makes it a missed call.
- iii“Most of your misses are during business hours.” Or after hours. Either claim needs the split the carrier declined to send.
A vendor can still be useful while being unable to say any of that. What should worry you is a vendor who says it anyway, because it tells you they have not looked closely at their own plumbing.
What can be measured, honestly
Two things survive contact with reality, and both are checkable in your own records.
The first is the timestamp. A forwarded call carries when it arrived, and you know your own hours. Comparing the two gives a real split: after hours, or during business hours. That is a genuine measurement rather than an inference, and it is usually the more useful cut anyway, because the fix is different on each side. After-hours volume is a coverage problem. Business-hours volume is a staffing or routing problem.
The second is the list. Every call that gets answered produces a name, a number, a reason for calling, and a recording. That list is not a rate and does not pretend to be one. It is simply the people who reached someone instead of a voicemail box, and what each of them wanted.
A list of names is a weaker-sounding promise than a percentage. It is also the only one that can be checked line by line, which is why we make that one and not the other.
What to ask
If you are evaluating anyone in this category, the question that separates them quickly is not about accuracy or languages or how the voice sounds. It is this: show me how you derive that number from what the carrier sends you.
The answers divide into three. Some will show you a model and say it is a model, which is fine and honest. Some will discover on the call that they cannot derive it, which is useful for both of you. And some will keep talking. The third group is the one to be careful with, because the same instinct that produced the number will be operating on everything else they tell you.
We do not quote a recovery rate. We hand over the list, the recordings, and the hour-by-hour pattern from your own system, and you decide what it was worth.
© 2026 KET Digital · Written by the people who run the systems.