What happens to the call you don't answer at 9pm

The call comes in at ten past nine. It's a Tuesday in February, the furnace has quit, there's a baby in the house, and the person ringing you is not in a patient mood. They get your voicemail greeting. They leave a message, or they don't, and either way the next thing they do is press back and ring the next name on the list.

By the time you play it back with your coffee, that job belongs to somebody else. You don't get told. There's no missed-job report. The only trace is a voicemail from a stranger who has since stopped needing you.

That's the job I want to talk about, because it's the one I get asked about most and the one where the obvious fixes are worse than they look.

First, what a missed call actually is

It isn't one thing, and treating it as one thing is why the fixes disappoint.

Some of it is genuinely urgent and genuinely yours — no heat, no hot water, water coming through a ceiling, a tenant with no power and a landlord who will hear about it either way. That call is worth money tonight and worth more than money tomorrow, because the person who answers it gets the next four years of that customer's boiler.

Some of it is urgent to the caller and routine to you. It can wait until seven without anyone suffering, but it cannot wait without somebody being told it can wait.

Some of it isn't work at all. It's a customer from March asking about an invoice. It's a supplier. It's someone who found your number on a job board and wants a price for something you don't do.

And some of it is the same person ringing three times in twenty minutes because nobody rang back, which looks like three calls in the log and is one job.

You already half-know this mix. What you probably don't know is the proportions, and the proportions decide whether any of what follows is worth doing. Two weeks of writing down every out-of-hours call — when it came, what it was, whether you rang back, whether it became work — costs you nothing and settles it. If it turns out you get four calls a month and three of them are invoice questions, the fix is a line in your voicemail greeting telling people when the office opens, and you've just saved yourself a project.

Why the answering service disappoints

The traditional answer is to pay people to pick up. It works, in the narrow sense that the phone gets answered by a human who is polite.

The trouble is what happens next. Whoever answers doesn't know your trade, your patch, or your book. They take a name, a number and a sentence, and that sentence has been through someone who doesn't know that “it's making a noise like a kettle” and “there's water under it” are different levels of trouble. What lands on your phone is a message that a message was left. You still have to ring the customer to find out what's wrong, and you're now doing that at seven in the morning with less information than the caller gave forty minutes after they rang.

They're also priced per call or per minute, which quietly means you're paying the same for the invoice question as for the no-heat call, and paying it every time the same person rings three times.

None of that makes them useless. If the value of you personally ringing back is high and your volume is low, a human answering the phone is hard to beat. But the thing people actually want from one — arriving at 6am already knowing which of last night's calls needed them — is not what it delivers.

Why the chatbot on the website disappoints

The other thing people get sold is a chat widget on the homepage.

Consider where your caller is at ten past nine. They are in a cold house, on a phone, probably standing in front of a boiler. They are not browsing. Even if they do land on your site, a box that asks them to type their problem into a form is asking a stressed person to do the slow version of the thing they were already doing fast.

There's a second problem with the ones sold as always-on customer service. They answer. That's the pitch — it deals with the enquiry so you don't have to. But answering means committing you: to a time, to a price, to whether this is covered under the work you did in November. A system that tells somebody you'll be there by nine has made you a promise you didn't make, about a morning it can't see, and you'll find out when the customer rings at ten past nine wanting to know where you are.

Your customers are on the phone. The work is on the phone. A tool that only helps the ones who happened to be at a laptop is solving a version of your problem that mostly doesn't happen.

What triage can honestly decide

Here's the narrower thing, and it's the one I build.

The line rolls over after six. The caller says what's wrong in their own words, the way they always have. Nothing answers them and nothing promises them anything — they get a greeting that says you'll come back to them, and they leave a message like they always would.

What changes is what's waiting for you. Instead of nine voicemails to play back at half speed while you're trying to eat breakfast, there's a written list. Each one has the number, what the caller seems to need, how bad it sounds, and — this is the part I care about most — what it couldn't make out.

That last field is the whole difference between a system you can use and one you can't. Real voicemails have a compressor running behind them. People say their postcode once, fast, in an accent the system hasn't heard much of. Half of them start talking before the beep. A summary that quietly guesses at a house number is worse than no summary, because after two weeks you stop checking it. A summary that says “address unclear, sounded like Marsh or March Road” is one you can act on, because it tells you where to spend your attention.

So the honest scope is: sort, summarise, flag what's uncertain, put the urgent ones at the top. That's it. Nothing is booked. Nothing is quoted. Nothing is answered. The decision is still entirely yours, and it takes you four minutes instead of forty.

That sounds unambitious and it is, deliberately. It's also the version where a mistake costs you a re-read on a Tuesday morning rather than a customer standing in a driveway waiting for a van that was never coming.

What has to stay on a person

I'd keep four things off any system like this, and I'd keep them off permanently, not until it gets better.

Committing to a time. Only you know what today looks like, who called in sick, and which job on the board is going to overrun.

Anything about money. Prices, whether this is under warranty, whether the work in November should have covered it. Get that wrong once in writing and you're arguing about it for a fortnight.

Anything with safety in it. If somebody says gas, or smoke, or a smell, that goes to the top of the list with a note and a human rings them. It does not get handled.

And the one where somebody is angry. A complaint answered by a machine becomes a worse complaint, reliably, and about a different thing.

You also need a decision made in advance about what happens when the whole thing is down, because one evening it will be. The phone provider changes something, or the service it depends on has a bad night. There has to be a route back to plain voicemail that takes one phone call to switch on, and somebody other than me has to know what it is.

What this doesn't fix

It doesn't answer your phone. Somebody who wants a human voice at 9pm still doesn't get one, and for the customers who care about that, this isn't the answer — hiring is.

It doesn't get the job booked overnight. If the person who rings back first genuinely wins in your patch, then being sorted at 6am beats being sorted at 8am and it doesn't beat being answered at 9pm.

And it doesn't stay fixed by itself. Voicemail systems change. The way people talk to them changes. A new estate goes up with street names nothing has heard before. It shows up as the thing getting quietly worse, which is why I price my own work as a build plus a monthly rather than a one-off — an automation nobody is watching is a thing that worked in the week it was installed.

Have a look at the actual output

I built a demo of exactly this, so you can judge it instead of taking my description of it. It takes a sample voicemail, works out what the caller needs and how urgent it is, and says what it couldn't make out: triage a sample voicemail. The data in it is invented and it takes about a minute.

Read what comes out and ask yourself one question — if that had been sitting on your phone at 6am, would you have rung a different person first than you actually did?

If the answer is yes, book a call and tell me what your out-of-hours volume really looks like. If it turns out you get four calls a month and three of them are about invoices, I'll tell you that too, and you can keep your money.