Where does '85% never call back' come from?

Eighty-five percent of callers who get no answer never call back.

You've read that sentence. It's in the ads, it's in the vendor emails, it opens half the blog posts ever written about answering your phone. I sell software in this category, so I have every reason to want it to be true, and I've quoted numbers like it to people like you.

So I went and read the sources. Not the posts citing the sources. The sources.

There isn't one.

Not a thin study, not an old one, not a vendor's own study with a thumb on the scale. Nothing. The most-repeated number in this industry has no original research under it anywhere I could find: no sample, no instrument, no trail that leads anywhere except to another page repeating it.

Now be careful about what that means, because I'm being careful about it too. I'm not saying anybody made it up, and I'm not pointing at a company. I can't, and that inability is the finding: these numbers have no owner. They got handed around until repetition started doing the job a citation is supposed to do.

Here's what's underneath the ones you've seen most.

The number with no owner

That figure gets quoted at you specifically because it's the perfect closing line for selling a plumber a phone system. It says the loss is total, and it says the loss is instant. No callback, no second chance, no voicemail to work through later. One ring-out, one job gone.

And the call it describes is real. You know that call. A supply line lets go behind a finished basement wall, the homeowner gets your voicemail, and they're dialing the next name on the list before you're back out from under somebody's vanity with a basin wrench in your hand. That happens. It probably happened to you this month.

What I can't tell you is that it happens 85% of the time, because nobody measured that. The figure and the phenomenon are two different things — and the figure has been riding on the phenomenon's credibility for years. You believe it when you read it because you've lived the call, not because anyone ever showed you data. That's a good instinct doing a job evidence is supposed to do.

Three more that came apart when I pulled on them

• "67 to 90% of callers won't leave a voicemail." Footnoted, over and over, to a 2016 BIA/Kelsey Call Commerce report. The report is real. The figure doesn't appear to be in it. Somewhere back down the chain a genuine citation got stapled to a number it never contained, and everyone downstream inherited the staple.

• "$75 billion lost." This one comes from a real survey, and a good one: NewVoiceMedia, 2018. It's about customers leaving companies after bad service experiences, broadly. It isn't about missed calls, and it never measured one. (The gap matters more to you than it sounds, and what follows is my reading of it rather than anything the survey concluded. Its own list of what drives people off a company includes being put on hold and not getting through to a person, which sits nearer to a ringing phone than I'd like it to. Even so, a customer who writes a company off has usually had more than one bad afternoon to sit with it. The homeowner watching water come through a kitchen ceiling from a failed supply line upstairs is deciding in about forty seconds, and I wouldn't put the first survey in front of you as evidence about the second.)

• "Missed calls cost the average contractor $X,000 a year." Every version of this is arithmetic. Somebody multiplied a call volume by a close rate by an average job value, the same thing you'd do on the back of an invoice, then published the product without publishing the three numbers that made it. That isn't research. It's a calculation in a lab coat. I know exactly how it happens, because I shipped one — so I published the whole thing: our missed-call math, with every input and the arithmetic shown, including the part where two of our own posts disagree by $390 and exactly which rounding step did it.

The two I'd still quote, with the strings attached

Two survived, and neither survived clean. If I use them, these caveats travel with them in the same sentence, every single time. That's the deal.

411 Locals, 2016: 62% of small-business calls go unanswered. Eighty-five businesses, across fifty-eight industries. Self-published by a marketing agency that sells to small businesses. Ten years old now. All three of those are true at once, and I'm not going to say the 62 without saying the 85.

CallRail, 2025: 14% of home-services calls are missed, across 1.1 million leads. CallRail sells the software that counts the calls, so it isn't a disinterested party. What it does do is disclose what it looked at and how much of it there was. Then it does the other thing too, and this is the part that stopped me: the same 2025 release that gives me that 14% turns around and repeats the 85% with no sample under it either. A disclosed sample on one number and the category's folk tale on the next, in one document, from the outfit I'm holding up to you as the careful one. That isn't one company being sloppy. That's how far down the 85 has soaked in, and it's why I don't get to write this post as though the problem belongs to somebody else.

Then set those two side by side: 62 and 14. They can't both be describing the same thing, and they aren't. Different populations, different decades, different definitions of a miss.

Which brings up the part I'd bet money on showing up in the next statistic somebody sends you. CallRail measured home services. Home services includes you. It does not include the accountant two doors down from your supply house, and the moment anybody writes "14% of small-business calls," that number has been stretched past what was measured. It's still a real figure. The citation still resolves. The sentence is false anyway, and nothing about it looks wrong, which is what makes that one nastier than a made-up number.

Even the honest version isn't quite about you. "Home services" pools the emergency trades in with the ones where nobody's floor is wet, so your miss rate on the Saturday a water heater lets go is not the category average, and never was.

I found one of these on my own site

I went and read my own pages before writing this, which is the least a person can do before writing this, and I found one.

The heating and cooling page said missed calls cost the average HVAC contractor about $45,600 a year. No source line under it. Not a range, not a hedge: a specific figure with a tilde in front of it, and the tilde is the tell, because a tilde makes a number look like the rounding of something that got measured.

Three things made that worse than a missing citation. It sat in the page's meta description, so it was the sentence showing up under the blue link. The page's own calculator, a few inches below it on the same screen, produced a materially different figure from the sliders it shipped with. And I was drafting this post at the time.

It's gone. Cut, and not replaced. The not-replacing took more discipline than the cutting, because there's a sourced figure I could have dropped straight into the hole, and swapping an unsourced number for one I hadn't read myself is the same mistake with better manners.

The plumbing page kept its number, and why it kept it is the whole standard in one line: that number carries a source note in the file the page gets built from, naming whose data it is and saying plainly that the data is a vendor's own and not an independent study. Now the ugly half of that. The note is not on the page yet, where you could actually see it. It's in the file, the file is what builds the page, and what the page prints is the number without the name. For a post like this one that is exactly the wrong way round, and I've put in the request to get it showing. The test was never "is this number impressive." The test is "does it have a name on it, and does the name hold up."

You're exactly who that meta description was aimed at, by the way. The trade page next door to yours, an owner thumbing down a page of results, reading the two grey lines under each name and deciding off those two lines which one gets tapped. That's where a number with no owner does its work, and that's why it mattered more than the page's traffic did.

What to do with a number that has no name on it

Click the citation. That's the whole method, and it costs about ninety seconds. If the link lands on another blog post, you haven't learned anything yet, so keep clicking. Stop when you hit something with an author, a year and a sample size — or when you run out of links. In this category, you mostly run out of links.

I sell into this category, and the standard above is the one I want applied to me. Run it on the plumbing page: find the number, then go looking for whose it is. Today you won't find that on the page, for the reason I gave you a few paragraphs back, and holding it against me until it's showing is fair. Then run the same pass on the next vendor page you land on and see whether a name turns up there at all. If a number of mine turns out to have nobody behind it, that's a defect, and I'd like to hear it from you before I hear it from somebody selling against me.

And if you want a number about your own shop instead of the category's, the missed-call calculator is three sliders: calls you miss in a week, how often you'd close them, what a job is worth. All three inputs are yours, all three sit on the screen, and you can drag any of them until the output stops flattering me. A number you can move is a number you can argue with.

The two figures in this post that survived have names attached to them. The one at the top doesn't, and it's the one you've read the most times. Those two facts sitting next to each other is the post.