Nine calls, eighteen cents, and nothing to book.

How to read this

Two phrases carry the piece, so they are defined before the numbers.

• Never reached a person: the carrier reported no answer, or the line played a voicemail greeting and closed. Either way, no human joined the call.
• Connected: the carrier reported the call as connecting at all, whatever was on the other end.

The method is one job, one ZIP code and one frozen search phrase: ZIP 75093 (Plano), phrase solar panel cleaning service. The search returns up to twenty companies and a batch dials up to twenty, so nothing caps at nine. Nine were selected and the other eleven were never called.

The nine calls, one by one

Each company gets a code, P1 to P7, because two of them were dialled twice and that duplication is part of the story. The listed hours are the directory’s own, and they say nothing about whether anyone was actually at the phone.

The three codes listed as open around the clock, P2, P3 and P4, were dialled four times between them and produced nothing at all.

The funnel

The same nine calls, narrowed four times. Every row is the sum of rows in the table above, and the “Calls” column says which, so nothing here is an orphan number.

One of nine calls reached a human being. That is the whole finding, and it is a harder number than the no-answer rate, because it counts a voicemail greeting as the dead end it is.

Two of the seven were called twice

Nine dials reached seven companies. The job’s provider list contained two businesses twice, under two different rows with the same name, so the dialler treated them as two prospects and called each of them again in a later batch. Neither repeat call produced anything the first had not already told us.

The second P6 call is the only call in the job that reached a human, so the repeat was not wasted. Two of nine dials going to a business already dialled is still a bug worth naming.

The one person who answered

The only conversation in nine calls lasted twenty-eight seconds and ended with one sentence.

“We only service Southern California.”

That is the entire yield of the job: not a price, not a date, not a maybe. The company is in the directory for Plano, and it does not work in Texas. A search that returns a business three states away from the job is a search problem, not a willingness problem.

The arithmetic

The job cost eighteen cents because almost nobody spoke, which is a strange thing to call a success. What it bought was certainty: these seven businesses were not going to answer that morning.

Talk time is the sum of the Length column in the first table: 0:00 six times, then 0:07, 0:04 and 0:28, which is 39 seconds. No call lasted long enough for the two-minute silence cap to matter, and the average is that total divided by nine.

What we would change next time

• De-duplicate the list before dialling. Two of nine dials went to a business that had already been called, because the directory returned it twice.
• Call all twenty, not nine. Eleven companies were never dialled, so the no-answer rate describes a selection rather than the whole list.
• Run the same list at a second hour. Mid-morning on a Monday is when a small crew is already out on a job. The same nine numbers at four in the afternoon could read completely differently.
• Separate a silent no-answer from a voicemail greeting. Both arrive as the same status today, and they mean different things. One is a phone nobody is near. The other is a business that chose to screen.

Why publish the whole funnel

A tool that reported “three companies answered” would be flattering itself. Two of those three were machines. The useful number is one, and the useful artefact is the table that shows why: which lines never picked up, which screened the call, and which business is in the wrong state entirely.