Eleven Houston calls. Ten counted as completed. Five reached a person.

How to read this

Five phrases carry the piece, and the first two are the ones a reader is most likely to swap.

• Recorded as completed: the status on the call row in our own database. It is not a statement that a person spoke. A voicemail box that took a message comes back as completed, because a message was left.
• Reached a person: a human at the provider’s end spoke, read from that call’s own summary. This is the number that describes the job, and it is smaller than the one above by half.
• Confirmed they could come: the product’s availability field was set to true because the provider said so during the call. It is a stricter test than sounding helpful.
• Named a figure: a provider said a number out loud. It is not a price for the job, and this post does not average it against anything.
• Arrival window: a provider gave a time they could be there. A provider who says yes without one has confirmed nothing a homeowner can plan around.

The claim being tested

The search phrase for this job was emergency plumber toilet unclogging, and the job was marked as soon as possible. That is the most demanding request the product takes: a homeowner with a blocked toilet who wants somebody today, searching with the word emergency in the query. What follows is what eleven emergency plumbing listings on the west side of Houston actually answered at 8:11 on a Wednesday morning.

Nineteen listings came back from the search. Eleven were dialled, which is a choice rather than a limit: this job did not dial the whole list. The eight that were not called are part of the denominator here and appear in the funnel and the arithmetic, because “eleven called” means nothing without “nineteen found”.

The eleven calls, one by one

Everything below reconciles to this table. No provider is named: the recordings carry business names, the homeowner’s street address and a callback number, and one call summary names the homeowner, so the call numbers are the identifiers here. The “Confirmed” column is the product’s own availability field, not our reading of the transcript: Yes is a provider who said they could come, No would be one who declined, and None is a call that never produced an answer either way.

Two things to note before the summary. There is no Noin the Confirmed column: nobody refused the job. Three yeses out of eleven calls is not three providers standing against eight refusals, and the difference matters when reading the funnel below. And call 8’s price cell reads No figure rather than None, because that provider did say something about money, which is not the same as saying nothing.

The ten calls our own records called completed

This is the finding. Ten of the eleven calls carry a status of completed in the Boocki database. Five of those ten reached a person. The other five did not, and the split is not random: it is what our own status field means. A voicemail box that accepted a message is recorded as completed, because the call had a result and something was delivered on it.

Ten decomposes exactly: five calls where a person spoke, three voicemail boxes that took a message, and two lines that connected and then produced nothing. Call 7 sat open for thirty-four seconds, and the only thing its own record has to say about it is one sentence.

No conversation happened.

That sentence is the whole record of a call that our status field counts as completed. It is not a data error and we are not going to relabel it: the field answers “did this call reach a resolution we can act on”, and for a voicemail the answer is yes. But it is the wrong column to build a headline from, and a board that reported “ten of eleven answered” would be wrong by a factor of two.

The funnel

The same eleven calls, narrowed seven 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.

Read the third row against the fourth. Both describe the same eleven calls, they differ by five, and the five are calls 4, 7, 9, 10 and 11: three voicemail boxes, one line that opened on nothing, and one that dropped during the greeting. If this page had a single number on it, it would be the fourth row, because the third one is a statement about our database rather than about Houston.

The six calls that produced nothing

Six of the eleven calls ended without a conversation. Three of them were voicemail boxes, which our records count as completed calls, because a message was left and our own rule treats a message as delivered once it runs past five seconds.

Three voicemail boxes in eleven calls at 8:11 on a weekday morning is the number that surprised us. It is the hour the story of this job turns on: the job was created at 10:50 the night before, when every one of these offices was closed, and by the time the calls went out the advantage had moved back to the provider. None of the six costs more than half a minute of talk, and together they account for one minute and thirty-six seconds of the job’s twelve minutes and twenty seconds.

The five that reached a person

Five calls reached a human being. Three of those five confirmed they could come, which is the availability field set to true, and two of the three then named a time they could arrive. The other two humans were willing and produced nothing a homeowner could act on.

Call 2 is the one worth reading twice. A person answered, understood the job, and then refused to proceed until the homeowner themselves gave permission, which is a reasonable thing for a business to insist on and is a wall that no script gets past. It is also the shape of refusal that a summary written carelessly would record as a failure to book, when what actually happened is that a third-party caller reached the limits of what a third-party caller can do.

The one figure anybody named

One provider of the eleven said a number out loud: a $45 dispatch fee, then $220 to clear the toilet with a hand auger. That is not a price for the job. It is a fee plus a price for one method, and the same provider would have priced a different method differently. It is the only figure in this post, and averaging it against anything, including the one window that came with it, would invent a market rate out of a sample of one.

Call 8 is the more interesting answer even though it contains no number. That provider offered an afternoon window and said there is no charge for sending the technician, with pricing to be decided on site. A homeowner reading that has learned something real about how that business works, and has learned nothing they can compare against the $220. Both of those calls are answers. Only one of them is a figure.

The arithmetic

The job cost less than the dispatch fee it was quoted, which is the arithmetic of an emergency search that mostly hit voicemail boxes: eleven calls, twelve minutes and twenty seconds of talk, and one provider willing to name a number.

The eleven durations sum to 740 seconds, which is the 12 min 20 s in the table: 2:36 + 2:28 + 0:54 + 0:07 + 0:32 + 0:03 + 0:34 + 4:14 + 0:10 + 0:08 + 0:34. The $3.69 is the ledger row this job settled to, not a sum of per-minute estimates, and $0.34 is that divided by eleven while $0.74 is the same money divided by the five calls that reached a person.

What we would change next time

• Dial the whole list. Nineteen listings came back and eleven were called, so eight providers were never asked. The denominator of this post should have been nineteen.
• Report the human count, not the status count. The two differ by a factor of two on this job, and only one of them is a statement about Houston. A summary that leads with calls marked completed is measuring our own bookkeeping.
• Ask for a figure in the same breath as the window. Three providers confirmed they could come and one named a number, which is the second half of the question a homeowner is really asking.
• Start when the job is created, or say why not.This job was created at 10:50 PM and dialled at 8:11 AM, nine hours and twenty-one minutes later, and three of the eleven lines were voicemail boxes by then. Whether that wait is correct is a product decision; publishing the gap is not optional, because the reader is otherwise comparing this job’s 8:11 AM against a different job’s 2 AM as though the two were the same experiment.

Why publish the whole funnel

A summary that said “ten of eleven calls answered” would be technically true about our database and completely wrong about Houston. Five people picked up the phone, three of them would come, one said a number, and the most useful line in the whole job is the one that reads “no conversation happened”. The value of calling eleven plumbers at once is not the single appointment or the single price at the end of it. It is knowing how many lines were actually open, how few of them would commit, and how much of the answer was hiding in a status column that was never meant to describe a person.