Nobody can reach the number
Start with one question: does a call record exist for the attempt? A call the carrier never saw, a call the carrier saw and Connect refused before any row existed, and a call that was answered and declined are three different faults with three different fixes. The checklist below runs in the order the call itself travels: claim, trunk, routing, worker, hours, switch.
Find out how far the call got#
Reading the carrier's own call log beside Connect's call list is the fastest audit available on a phone line, and it separates the three cases in seconds. A completed call on the carrier's side with no row on Connect's is a fault every time — that comparison is how several real defects were found rather than reported.
| Carrier log | Connect | What it means |
|---|---|---|
| Nothing | Nothing | The call never reached the carrier: the number, the claim or the trunk |
| A completed call | No record | It reached the carrier and was refused before a record existed: verification, scope, or no worker |
| A completed call | A record ending as line off or after hours | The line answered and declined — the caller heard the closed-line message |
The third row is not a line that fails to ring. It rang, it was answered, and the business's own closed-line message was spoken in its own name before the call ended. If that is what your callers are describing, the hours or the switch is the answer and the rest of this page is not.
The checklist, in the order the call travels#
- 1. Claim
- The number must exist as a route belonging to this workspace, with the right provider and inbound calling permitted. A line that is not claimed by a workspace has no owner to answer for it, and everything downstream reads as empty.
- 2. Trunk
- On the realtime engine the number has to be linked at the carrier to the inbound trunk, and the trunk's origination address has to be the SIP host in bare host-and-port form, with no scheme in front of it. This link is made at the carrier, not in Connect, and a number that is provisioned but not linked never leaves the carrier.
- 3. Routing
- The carrier must post to the application Connect answers on, and a dispatch rule must put the worker into the room the call creates. A number attached to no application registers perfectly and can never place or receive a call.
- 4. Worker
- A worker has to have checked in recently. When none has for ninety seconds, the ring is refused before a record exists, deliberately, so that the retry has something to come back to. The Phone screen used to say 'ready' in exactly this state, which is why line health now names an absent worker as one of the four things it watches.
- 5. Hours
- Outside business hours the call is answered and declined with the closed-line message, recorded with an after-hours outcome. The message names the business and deliberately does not read the hours out.
- 6. Switch
- A switched-off line does the same thing with a line-off outcome. This used to be a bare refusal that dropped the room, so the caller heard the call connect and die with nothing recorded anywhere — the refusal is now spoken and the row is written.
Two refusals that happen before any record exists#
Both are deliberate, and knowing them stops a long search for a record that was never going to be written. The first is the no-worker refusal above, which is retried by the drain rather than being final. The second is a malformed or unusable model identifier, which fails the call rather than letting it ring and die halfway through — an early test call rang the owner and then died mid-session for exactly that reason.
There is also a capacity refusal. When every live worker is over the busy threshold — 0.85 one-minute load per core, the same number the worker gives the media layer, so admission control and the gate agree — the ring is refused as not ready. That is every line busy, not a broken number.
What Connect completed#
- Where the call reached the line and was declined, it was answered, the refusal was spoken in the business's own name, and the row exists with an outcome of line off or after hours.
- Where the ring was refused for want of a worker, the refusal is recorded as a not-ready state that the drain retries rather than a failure that is forgotten.
- Line health continues to report the four things that drain a line by themselves: no worker checked in, no capacity, repeated calls that never rang, and sessions the model refused.
- Every handler that touches the call now runs inside the owning workspace, so a workspace-scoped credential is read as what it is rather than as empty.
What Connect did not complete#
- No conversation happened, so there is no transcript, no summary, no lead and no follow-up from the attempt.
- Where the refusal happened before a record existed, there is nothing in the call list to open — the carrier's log is the only account of the attempt.
- Connect did not link the number to a trunk, attach it to an application, or change the hours or the switch on your behalf. Those are decisions with an owner.
- Nothing was said to the caller in the cases where the call never reached the line at all.
What you can do#
Call the number yourself from a phone outside the business and note exactly what you hear: ringing then nothing, a message, or an immediate failure tone.
Result A spoken message means the line answered and declined; a failure tone before ringing means the call never reached the line.
Open the call list and look for a record in the last few minutes.
Result A record with an outcome tells you which of the six steps stopped it. No record sends you to the carrier's log.
Check the hours and the on-off switch on the Phone screen before anything technical.
Result These are the two that change most often and explain most reports, and neither needs anyone technical to fix.
If nothing at all appears on either side, verify the number's claim and its trunk link at the carrier.
Result A number that is not linked never leaves the carrier's network, which is why neither system has anything to show.
What an administrator can do#
- Check that a worker has checked in, and that the fleet has capacity, before touching the number's configuration.
- Confirm the carrier's callback verification is being checked inside the owning workspace — a scope mistake here refuses real calls before pickup and is covered by carrier verification.
- Confirm the trunk's origination address is the bare host and port with no scheme, which is the form that works and the one most often written wrongly.
- Compare the two logs for the whole day rather than for the reported call — the pattern separates a configuration fault from a moment of saturation.
When to escalate#
Escalate when the carrier logs completed calls that leave no record at all, when a line reports itself healthy while calls reach nothing, or when a number that was reachable stops being so with no change behind it. The first of those has been a real defect more than once and is worth treating as one on sight rather than as a reporting delay.
Questions#
The caller heard a message saying we are closed, but we are open.
The line answered and declined, so this is the hours or the switch rather than the number. The closed-line message is deterministic and names the business without reading out hours, which is why it can sound plausible even when the hours behind it are wrong.
Is a refused ring the same as a missed call?
No. A call that rang and was not picked up is one outcome; a call that never rang because the dial, the dispatch or the worker failed is another, and neither produces a lead. Keeping them apart is what stops a run of infrastructure failures from reading as an unpopular business.
Why refuse the ring instead of answering and apologising?
Because a refusal before a record exists can be retried, while a call answered into nothing cannot. The trade is deliberate: the caller can redial in seconds, whereas an answered call that goes nowhere is a lost conversation with a record that says it happened.