Search this question and you get a wall of confident statistics. Some percentage of callers never call back. Some dollar figure per missed call. Some study, usually uncited, usually about an industry that is not yours.
I am not going to add one. Every number I found traced back to a vendor selling the fix, and the ones with a methodology attached measured a specific industry in a specific year. Yours is different, and you can work it out from data you already have.
The number you need is four numbers
Missed calls per month. This is on your phone bill or in your carrier's dashboard, and almost nobody looks. Ask for unanswered inbound, not total inbound. If you use a mobile, your call log has it. Count a month, not a week — one bad week is not a pattern.
What fraction of answered calls become work. If you answer a hundred and twenty of them turn into a booking, a quote, a sale, that is 17%. Use your own ratio, not an industry one. If you have no idea, this is the number worth spending an afternoon on, because everything else multiplies by it.
Average value of that work. Not lifetime value — the first job. Lifetime value is a projection and it makes the answer look better than it is.
What fraction of missed callers come back. This is the one people guess at. You can estimate it honestly: of the numbers that rang and got no answer, how many rang again within a day, or turn up in your customer list? Your call log has both halves.
Then:
monthly loss = missed calls
× (1 − fraction who come back)
× conversion rate on answered calls
× average job value
Why the fourth term is the whole argument
Most vendor arithmetic quietly sets "fraction who come back" to zero. That is the assumption doing all the work, and it is usually wrong.
If someone is ringing their regular plumber, they will ring again. If they are working down a list of six from a search result, they will not — they are ringing the next one while your voicemail is still playing. Same missed call, wildly different cost.
So the honest version of this question is not "what does a missed call cost". It is "how substitutable are you?" A business with a booked-out reputation loses very little to a missed call. A business competing on availability loses nearly the whole thing.
Work that out before you spend anything on answering the phone faster, because it decides whether the spend is worth it more than any rate card does.
The pattern that changes the answer
Pull the timestamps of the missed calls. They usually cluster, and where they cluster tells you what to buy.
Clustered outside opening hours — the cheapest fix is anything that answers at all, including a better voicemail message with a text-back. You may not need an agent.
Clustered during your busiest hour — you are not understaffed across the day, you are understaffed for ninety minutes. Overflow is the fix. A human service with a minimum is poor value for ninety minutes a day; a per-minute line costs nothing for the other twenty-two and a half hours.
Spread evenly — you are genuinely short of capacity and the question is a hire, not a tool.
The middle case is the one an AI line is actually built for, and it is the one that per-minute pricing suits: you pay for the ninety minutes and nothing for the rest.
What answering the call does not fix
Answering is not the same as converting. An agent that picks up on the first ring and then cannot answer the question has converted a missed call into a disappointed one, which is not obviously better.
So before costing out any tool, write down the last twenty things callers actually asked. If most of them have one answer that is the same for everybody — hours, location, price range, whether you cover an area, whether you do a thing at all — a machine can hold that line. If most of them need your records or your judgement, the fix is a person, and the arithmetic above is telling you to hire rather than to buy software.
Our agent knows only what you wrote down for it and what it can find on the web. It has no CRM, no calendar, no order lookup, and no way to identify a caller. That bounds it sharply and it is worth knowing before, not after.
Putting the two numbers next to each other
Once you have your monthly loss, compare it against what answering costs:
monthly AI cost = calls × ceil(avg minutes) × $0.23
+ carrier inbound rate × total minutes
+ number rental
23 cents a minute is our rate, derived rather than picked: 90,000 µUSD of cost per minute plus a 150% markup is 225,000 µUSD, rounded up to the cent. Carrier charges are yours and separate — we do not resell minutes.
For most small businesses the AI side of that comparison comes out surprisingly small, and the decision ends up hinging on the substitutability question rather than on cost. Which is the right place for it to hinge.
Status
Osokoro Phone is a private beta, approved manually, one business at a time. Inbound only, five-minute default call cap, no transcripts retained, and your own SIP number from your own carrier. Number provisioning is done by an operator rather than self-serve, and the rate above is proposed rather than a billing commitment.
Do the arithmetic first. If it says the loss is small, you have learned something more useful than a free trial would have told you.