How many calls is my business missing? Count them first

Tuesday, 12:40pm. Dana runs the office for a two-van plumbing company. She's eating a sandwich in her car because that's the only quiet twenty minutes in her day. While she's out there the office line rings four times. One caller leaves a voicemail. One rings again at 2pm and gets through. Two never appear again.

By Friday nobody remembers any of it happened. There's no ticket, no note, no row in the CRM. The two callers who vanished left no trace at all, and a problem that leaves no trace never gets fixed.

So before you buy anything to fix it, get the number. Not an industry number. Yours.

The figure on the vendor blog is not your figure

Search for missed calls and you'll meet the same two statistics everywhere: that 62% of business calls go unanswered, and that the average small business loses around $126,000 a year because of it. Go looking for the study behind either one and you land on another vendor blog citing a third. We stopped repeating them a while ago, for the obvious reason.

There is real data. Invoca analysed over 60 million phone calls for its benchmark report and puts the share of business calls answered by a person at 56% across industries, with big swings by sector. That's a genuine measurement of 60 million calls, and not one of them was yours.

Your phone system logged every one of them already

Every call that hit your number left a record, whether anyone picked up or not. It's sitting in a call log you probably have never exported.

If you're on a VoIP platform, the export is a menu item. RingCentral keeps call history in the account for up to 12 months and will hand it to you as a CSV. Dialpad, Nextiva, 8x8 and Grasshopper all do the equivalent. Ask for 90 days rather than 30. One month is too short to tell a real pattern from a bad week.

If your business runs on mobiles, the call detail is on the carrier bill. Log in to the account rather than reading the PDF summary, because the itemised view is the one with timestamps and durations.

Then there's the awkward setup, which is also the most common one in trades: a landline number that forwards to two mobile phones. The forwarding leg looks answered at the switch the instant it's handed off, so the landline log will tell you everything went through. It didn't.

You need the logs from the handsets that actually rang, matched up by timestamp. Budget an hour. This is also the setup where the leak usually turns out to be biggest, because nobody has ever been able to see it.

Decide what "missed" means before you open the spreadsheet

A call log has more outcomes than answered and missed, and the difference matters more than people expect. Sort every row into four buckets.

The caller spoke to a person. The caller reached voicemail and left a message. Or they reached voicemail and hung up without leaving one. Or, worst of all, they rang out or hit a busy tone and never reached voicemail at all.

The first two are fine. You have the caller either way. The last two are the leak, and the fourth is the worst of them, because that caller got no acknowledgement of any kind and has no reason to believe you exist.

One trap to check for: plenty of systems mark a call "answered" the moment an auto-attendant picks up. If your log shows a 99% answer rate and a pile of eleven-second calls, your greeting is answering the phone and nobody else is. Filter out anything under about twenty seconds and look again.

Deduplicate, or you'll double your own leak

Every missed call calculator on the internet skips this step. Skipping it is what makes the number big.

Collapse the log by phone number, not by call. Somebody who rang three times in eleven minutes is one lost customer, not three. Somebody who missed you at 11:02 and got through at 11:40 is not lost at all, they're a slightly annoyed customer you already have. Take them out.

Then strip the numbers that were never leads. Your own technicians calling in. Suppliers. The merchant you order from twice a week. Spam, which in most logs you can spot by the pattern of one ring and never again.

A quick rule gets you most of the way: if the number already exists in your CRM or your outbound call history, it's a known contact, not a lost lead. Worth checking whether your CRM actually holds that history in a usable form, because a lot of integrations turn out to be thinner than they look.

What survives is the real unit of measurement: unique phone numbers that reached nobody, and never reached you afterwards. In our experience that final list runs somewhere between a third and a half of the raw missed-call count. Anyone quoting you the raw count is quoting you a number they know is inflated.

Sort the survivors by hour and by weekday

Now make two simple tallies: lost callers by hour of the day, and lost callers by day of the week. A spreadsheet pivot does it in about four minutes.

Almost every log we've looked at spikes in three places. Early morning, before anyone is at a desk but after people have woken up to a cold house. The lunch hour, which is Dana in her car. And late afternoon, roughly 4:30 to 6pm, when the office has shut and the callers are people who waited until they got home from their own job to ring you.

The weekday tally usually has a shape too. Monday morning after a weekend of people deciding to finally do something about it. The first genuinely cold week of autumn, when heating systems fail all at once and the calls do not spread out politely across the month.

This is the payoff. An average tells you that you're missing calls. The hourly tally tells you when, and only the second one can be staffed, forwarded or automated.

Price it with your own close rate

Skip the industry revenue calculators. You have better inputs than they do.

Take the new-customer calls you did answer in the period, and count how many became booked jobs. That ratio is your close rate on inbound calls, measured on your own phones. Multiply it by your average first-job value. That's the expected value of one answered new-customer call.

Now apply it to the deduplicated lost list, and then discount it. Some share of those callers would have rung back on their own. Nobody can tell you that share from the outside, and no vendor who tries is measuring your business. Until you've checked, be pessimistic: assume half of them would have come back anyway, and halve the figure.

You end up with one defensible sentence. "We lost 34 first-time callers in 90 days, mostly between 4:30 and 6pm, worth roughly this much in booked work." Every term in it came off your own phones, which means you can defend it to a partner, a bank or yourself.

Ring twenty of them back and ask

Almost nobody does this. It's the only step that turns an estimate into something you know.

Take twenty numbers off the lost list and ring them. The script is one line. You called us on the 14th and we missed you, did you get it sorted?

Three kinds of answer come back. Some got the work done by somebody else, which confirms the loss was real and gives you a competitor name for free. Some never got round to it, which means the list you just built is a callback list rather than a statistic. And some were never customers, which tells you the filtering needs another pass.

An hour, give or take. It usually ends the internal argument about whether any of this matters, because it's harder to dismiss a person who says yes, I got someone else in, than a cell in a spreadsheet.

When the honest answer is that you don't have a problem

Sometimes the audit comes back small. It happens more than we'd like.

A business rings us convinced it's bleeding calls. We run the log, dedupe it properly, and the ninety days contain six unique first-time callers who never got through. Five of the six called back the next day. At that volume nothing we build pays for itself, and we say so. We've talked ourselves out of work this way, and we'd rather do that than install a system that quietly costs more than the leak it plugs.

Two other outcomes are worth naming, because both of them are cheaper than hiring us.

If the whole leak sits in one hour, you have a rota problem, not a coverage problem. Move who covers lunch and the number goes to nearly zero for free.

If the leak is entirely after 6pm and your work isn't urgent, a voicemail greeting that promises a callback by 9am, plus somebody who actually makes those callbacks, will recover a good share of it. That costs nothing but discipline.

It won't hold up for emergency trades. A caller with water coming through a ceiling is dialling the next number before your beep finishes. The promise also gets weaker the longer the gap runs, because the window for responding to a new lead is shorter than most teams assume. For plenty of businesses it's still enough.

The audit also has a real blind spot you should know about. A call log records that a number rang and that nobody answered. It cannot tell you what the caller wanted. Every number you keep on the list is an assumption that it was a customer, and every number you strip is an assumption that it wasn't. That's why the twenty callbacks matter, and it's why we'd rather quote off a measured list than off a percentage.

What to do this week

Block ninety minutes. Export 90 days of call detail from wherever it lives, bucket the outcomes, collapse by number, strip the known contacts, then pivot by hour. You'll have your figure before the time is up.

If it comes back large, and it's concentrated in hours when nobody can realistically be at a desk, that's the case where something has to pick up the phone for you.

Either way, send us the pivot table. Tell us what you found, paste the hourly breakdown into the message, and we'll tell you which of the three answers it is. Sometimes it's the lunch rota.