We called 10 “24-hour” locksmiths at 9:36 PM. Four answered.
Seven of the ten listings said "Open 24 hours". Six providers reached a human at 9:36 PM on a Friday; only four would come out, and one named a price.
It was 9:36 on a Friday evening in Atlanta, Georgia, and a homeowner was locked out of the front door with the keys gone. It is not the hardest hour to need a locksmith. That is 3 AM. But it is late enough that every office is shut, and the only number that matters is the one that answers.
- 20
- locksmiths found
- 10
- called9:36 to 9:47 PM
- 4
- confirmed they could comeof the ten
- 1
- named a price$149 · ~20 min
How to read this
Three phrases carry the whole piece, so they are defined here rather than fifteen hundred words in.
- Connected: the carrier reported the call as connecting at all.
- Reached a human voice: somebody other than a carrier recording spoke.
- Confirmed they could come: the provider said so on the call. This is the product’s own availability field, and a stricter test than a human picking up.
The method is one job, one ZIP code and one frozen search phrase. The ZIP is 30310 (Atlanta); the phrase is residential locksmith lockout service. The search returns up to twenty providers and a batch dials up to twenty, so nothing caps at ten: ten were selected and dialled, and the other ten were never called.
Open, according to whom?
Every one of the ten locksmiths showed as open in the directory at the moment we searched, and seven of them listed their hours as “Open 24 hours.”
But a listing’s hours describe a schedule. They are not a switchboard. For someone standing outside a locked door, the only thing that matters is whether a human being picks up. That is a different question.
The ten calls, one by one
The “listed hours” column is what the directory claimed before we dialled. Everything else is what happened on the line. No provider is named: the recordings contain the homeowner’s address and callback number, so the call numbers are the identifiers here.
| # | Listed hours | What happened | Length | Confirmed | Price |
|---|---|---|---|---|---|
| 1 | Open 24 hours | Answered; quoted a price and an arrival time | 1:24 | Yes | $149 |
| 2 | 7 AM to 11 PM | Answered; took details and promised a callback | 2:13 | No | None |
| 3 | 7 AM to 11 PM | Carrier returned no answer (a four-second announcement was still captured) | 0:04 | No | None |
| 4 | Open 24 hours | Automated greeting, then the line dropped | 0:17 | No | None |
| 5 | 6 AM to 10 PM | Answered, asked who was calling, then ended | 0:38 | No | None |
| 6 | Open 24 hours | No conversation at all before the call ended | 0:16 | No | None |
| 7 | Open 24 hours | Rang to the end; nobody picked up | 0:36 | No | None |
| 8 | Open 24 hours | Answered after a hold queue; a technician would call back with the price | 2:24 | Yes | None |
| 9 | Open 24 hours | Answered via an automated menu; refused to price it without seeing the door | 2:24 | Yes | None |
| 10 | Open 24 hours | Answered; took the ZIP, address, name and number, then promised a callback | 1:21 | Yes | None |
Two things in that table are worth flagging before the summary. Call 3 is the single call the carrier reported as not connecting, and it is the one row the funnel below leaves out of “connected.” And calls 8, 9 and 10 are three different businesses, with three different transcripts and three separate recordings; two of them happened to run to exactly 2:24, which is a coincidence rather than a duplicated row.
The funnel
The same ten calls, narrowed five 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.
| Step | Calls | Count | Share |
|---|---|---|---|
| Locksmiths found | n/a | 20 | n/a |
| Called | 1 to 10 | 10 | 100% |
| Connected by the carrier | all but 3 | 9 | 90% |
| Reached a human voice | 1, 2, 5, 8, 9, 10 | 6 | 60% |
| Confirmed they could come out | 1, 8, 9, 10 | 4 | 40% |
| Named a price | 1 | 1 | 10% |
The 90% is the number an ordinary call log would print as “answered,” and it is the least useful of the six: it counts a four-second carrier voicemail (call 3) the same as a two-minute conversation.
The six that never confirmed
These are not one failure repeated six times. Every one of them failed differently, and the length of the call is the tell.
| # | Failure | Length | What the recording shows |
|---|---|---|---|
| 2 | Took details, promised a callback | 2:13 | The longest of the six. A human ran the intake properly, and quoted nothing. |
| 3 | Voicemail, never set up | 0:04 | A carrier recording: the mailbox has not been set up. |
| 4 | Recording, then disconnect | 0:17 | An automated greeting; the line dropped mid-sentence. |
| 5 | “Who’s this?” | 0:38 | A person answered, questioned the caller, and hung up. |
| 6 | No conversation at all | 0:16 | Neither side said anything before the call ended. |
| 7 | Rang, nobody picked up | 0:36 | It rang to the end. No greeting, no voicemail, no answer. |
The four yeses, and what each was worth
Four providers said yes. Only one of them said a number. The other three agreed to help and handed the pricing to a technician who would call the homeowner back. That is a real answer, and not a quote.
| # | What they said | Price | Arrival |
|---|---|---|---|
| 1 | Answered, and quoted a price and an arrival time on the spot | $149 | ~20 min |
| 8 | A hold queue, then a technician callback with the price | None | None |
| 9 | An automated menu, then a refusal to price without seeing the door | None | None |
| 10 | “I’ll have my technician give her a call right back.” | None | None |
“I should be able to get there in about twenty minutes … it’s gonna be, like, one forty nine for everything.”
That is the whole of what a locked-out homeowner takes away from ten calls: four yeses, a hundred and forty-nine dollars, and a wait of about twenty minutes, from one provider out of twenty.
The claim nobody could back up
Seven of the ten listings claimed round-the-clock hours. Four of those seven confirmed they could come out. All three of the ordinary-hours listings failed to confirm, the opposite of what the claim is supposed to buy you.
| Listing claim | Called | Confirmed | Gave a price |
|---|---|---|---|
| “Open 24 hours” | 7 | 4 | 1 |
| Ordinary daytime hours | 3 | 0 | 0 |
Ten calls is far too few to call that a pattern, and this is not evidence that “Open 24 hours” is a lie. It is the hypothesis for the next run: call all twenty at 2 AM, when the claim is actually under load, and see whether the split widens or disappears.
The arithmetic
The whole job cost less than a single call-out fee: twenty providers found, ten called, eleven and a half minutes of talking.
| Metric | Value |
|---|---|
| Providers found | 20 |
| Calls placed | 10 |
| Talk time | 11 min 37 s |
| Longest call | 2:24 (calls 8 and 9) |
| Shortest call | 0:04 (call 3) |
| Charged past the last spoken word | 0 s |
| Total charged | $3.47 |
| Average per call | $0.35 |
Talk time is the sum of the Length column in the first table: 1:24 + 2:13 + 0:04 + 0:17 + 0:38 + 0:16 + 0:36 + 2:24 + 2:24 + 1:21, which is 11 min 37 s. No call ran into the two-minute silence cap, so every second on the ledger is a second somebody was actually speaking, and the average is that total divided by ten.
What we would change next time
- Run the same ZIP at 2 AM.9:36 PM is early evening, not the worst case. If the “Open 24 hours” claim breaks anywhere, it breaks at 2 AM.
- Call all twenty, not ten. Half the list was never dialled, so the top of the funnel is a selection rather than the whole population, and the ten that were skipped could move any of these numbers.
- Hold the price question constant. One call got a price and an arrival time; the rest were promised a callback. Same ask, same moment, every call.
- Log the carrier’s connect status per call so the funnel and the call table reconcile by construction rather than by a footnote.
Why publish the whole funnel
A tool that reported “four locksmiths available” would be hiding the most useful thing it learned: what happened to the other six. Boocki finds the providers, makes the calls, and writes down what each one said, including the ones that never picked up.