The problem, in one sentence
A twelve-second recording tells every existing client to press 0, which routes about 297 people a day through two staff, whose transfer targets ring for five seconds and drop into mailboxes that are deleted after being emailed. Clients then call back, which is why 76% of all call volume is repeat callers.
We then make it worse from our own side. 41.8% of the calls we place to clients show them the main menu number, and 76% show them some shared number rather than the person calling. When they ring back they are put through the same menu and asked to guess an option, which is why so many end up choosing intake or the operator.
Nothing in that chain is a people problem. Reception answers a higher share of what reaches them than any other group in the firm. They have nowhere reliable to send anyone, and the number we put on the client's phone does not lead back to us.
What we are proposing
The menu
| Key | The caller hears | Goes to | Calls/day |
|---|---|---|---|
| 1 | For the status of your case, your hearing, or a decision | Front Line Queue | ~90 |
| 2 | For medical records, forms or paperwork | Front Line Queue | ~30 |
| 3 | If you are a doctor's office, records company or government agency | Front Line Queue, no identity check | ~97 |
| 4 | If you are returning our call and are not sure why | Front Line Queue, flagged as a callback | ~49 |
| 0 | For anything else, or to speak with an operator | Operator Queue | remainder |
| * | Dial by name (already enabled) | The person, then their team queue | ~49 |
All four options land on the same team. The keys exist so the agent knows why the person is calling before they pick up, and so we can measure the mix and split later if the volume justifies it. Simplicity now, evidence for the Twilio design later.
Key 4 earns its place on the numbers. "Returning our call" is 135 calls a fortnight, 61% of them ending in voicemail, and only 1% of them could be routed by a menu today because the caller genuinely does not know who rang them or why.
A key that says exactly that turns the least routable category in the system into the most routable one. The agent sees a callback flag, opens the record, and can tell the client in seconds who called and about what. It is the cheapest fix on this page.
No new-claim option, and the evidence says it costs us nothing. We checked the callers against CV1 directly.
| Who we checked | Numbers | Already a CV1 client | Not in CV1 |
|---|---|---|---|
| Everyone who pressed the intake option | 178 | 84.8% | 15.2% |
| Callers a transcript review judged to be a new claim | 91 | 83.5% | 16.5% |
The second row is the important one. Those callers sounded like new business, and five in six were already clients. Genuinely unknown numbers arriving on the main lines work out at roughly two to three calls a business day, and that is an upper bound, since "not in CV1" also covers a family member ringing from their own phone and numbers CV1 never captured.
Today those two or three real prospects press 1, land in extension 5011, and get a mailbox 100% of the time. Zero of 229 calls to that group were answered in thirteen days. The new-claim option is currently the worst-performing path for new business in the firm. Removing it and letting those callers reach the front line is strictly better for the exact people it was built to serve.
When we call a client, we hand them the wrong number back
Outbound caller ID is set per extension. For three quarters of our calls it is not the number of the person calling, it is a shared number that leads somewhere else.
| What the client sees on their phone | Calls | Share | What that number actually is |
|---|---|---|---|
| 812-232-9595 | 7,499 | 41.8% | The main menu |
| 812-234-9469 | 3,179 | 17.7% | Client Concierge Queue |
| 812-232-9599 | 1,561 | 8.7% | Decision Support group |
| 812-232-3246 | 1,244 | 6.9% | Case Status group |
| The caller's own direct line | 4,285 | 23.9% | rings the person who called |
So a case worker rings a client, the client rings back, and instead of reaching that person they are dropped into a phone menu and asked to guess an option. That is the mechanism behind callers choosing intake or the operator. They are not confused. They were handed a number that does not go where they were told it goes.
How this figure was derived, because it is the one number here that is not read straight from the call record. Vonage records the originating extension on an outbound call, not the caller ID shown to the client, so this is the extension configuration joined to call volume. It has been spot-checked live against an Initial Application extension, which matched.
If any team's phones present something different from their configured value, the share showing the main number would be higher than 41.8%, not lower, since the alternative shared numbers are the three below that answer nothing. Treat 41.8% as a floor.
What happens when they do ring back
| Number we presented | Outbound calls | People rung | Rang back in 72h | Reached a person |
|---|---|---|---|---|
| 812-232-9595 main menu | 7,499 | 3,745 | 537 | 38.2% |
| 812-234-9469 Concierge | 3,179 | 1,272 | 6 | 0% |
| 812-232-9599 Dec Support | 1,561 | 1,061 | 50 | 0% |
| 812-232-3246 Case Status | 1,244 | 549 | 5 | 0% |
3,179 outbound calls from the Concierge number produced six callbacks. 1,244 from Case Status produced five. Almost nobody rings them back. It is not that clients decline to call back in general, because 537 rang the main number back over the same period.
And of those 537, only 38.2% reached a person and 55.5% gave up, worse than that number's own average, because they are being asked to navigate a menu to get back to somebody who had just called them.
The two that point at call groups answer nothing
Separating out calls where the caller withheld their number, which our anti-spam setting rejects account-wide, this is what the three shared numbers actually did between 1 June and 14 August:
| Number | Routes to | Identifiable callers | Answered |
|---|---|---|---|
| 812-232-9599 | call group 6526 | 414 | 0 |
| 812-232-3246 | call group 6522 | 40 | 0 |
| 812-234-9469 | call queue 6518 | 10 | 0, but see below |
454 identifiable calls to the two group-backed numbers, none of them answered, 318 of those during business hours. That is the same five-second blast failure described later in this proposal, reached from a different direction.
The third number is a different case and we should not overstate it. 812-234-9469 points at the Client Concierge queue, and only 10 identifiable calls reached it in ten weeks, which is far too few to draw a conclusion from. A live test on 14 August dialled it and entered the queue normally. The number works.
The real question it raises is why 3,179 outbound calls presented that number and only 10 people rang it back. Either clients are not seeing it, or they are calling the main number instead because that is the one they know.
Separately, we are putting one of these numbers in front of clients in writing. A CasePulse portal message sent on 13 August, opened by the client three minutes later:
"Hello [client], this is Newlin Disability. We need to speak with you regarding your Disability case. Please call us at (812) 234 - 9469. Thank you."
That number is fine. But the same template logic applied to a Case Status or Decision Support sender would name 812-232-3246 or 812-232-9599, and those two have not connected a call in ten weeks. Worth checking what the portal and the outbound voicemail script tell people to dial, per team, before anything else on this list.
One more thing the call record turned up
Our system rejects calls from withheld or anonymous numbers. That is 1,292 legs across ten weeks, 0.7% of inbound, from 123 different callers, and it applies to every number including the main line. It is a reasonable anti-spam default. It is worth a conscious decision rather than an inherited one, given some clients call from hospitals, shelters and switchboards that withhold by default.
The rule that fixes the most calls for the least work
Every direct extension's no-answer destination changes from that person's voicemail to their team's queue.
Asking for someone by name is the second largest reason people call, about 49 a day, and today 77% of those end in voicemail. The caller already knew exactly who they needed. Some current examples, each a live number a client can dial:
| Direct line | Calls in 13 days | Reached a person | Voicemail |
|---|---|---|---|
| 812-231-6144 | 66 | 0% | 100% |
| 812-231-6151 | 60 | 0% | 100% |
| 812-231-6165 | 44 | 0% | 100% |
| 812-231-6156 | 38 | 0% | 100% |
These are not understaffed. They ring one person who is on another call, then stop. Pointing the overflow at the team queue turns roughly 47 dead calls a day into answered ones, and costs nothing but a dropdown per extension.
This rule and the caller ID fix only work together, and that is why they are one step.
Setting a case worker's outbound caller ID to their own direct line is the obviously right thing to do, and it is wrong today, because their direct line goes to voicemail 100% of the time. Fix the failover first and it becomes safe. Do both and the chain finally closes:
we ring the client from a number that is ours, they ring it back, they reach us, and if we are on another call they land in our team's queue where a supervisor can see them waiting. No menu, no guessing, no mailbox.
For 54 staff this rule cannot apply, and that is a separate fix
Client Concierge, Case Status and Decision Support staff have no direct number at all. All 32 Concierge extensions share 812-234-9469, all 14 Case Status share 812-232-3246, all 8 Decision Support share 812-232-9599, and not one of them has a line of their own to fall back to.
For those teams the shared number is genuinely the right thing to show a client, because the client should reach the team, not one person. The problem is only that all three of those numbers currently reject every call. Point them at their queues and the same outcome is reached from the other direction: the client rings the number on their phone and lands in a queue somebody can see.
What changes for the people answering
1. Verify identity the same way, every time
Of 1,464 calls from clients or their families, 36% were handled with no identity check at all and another 35% with a first name only. We discuss case status and hold protected health information on those calls.
The script already exists in our own best calls and is used 29% of the time: name, then date of birth. Standardising it across one front-line team is achievable in a week. Standardising it across 128 people is not, which is the strongest argument for the front line having its own team rather than being a rota.
2. De-escalate rather than route to an attorney
Attorney requests run about 17.5 a day. Reviewing the recordings, most are clients who want reassurance and an answer, not counsel. The front line is trained to answer the underlying question and to escalate only where it is genuinely legal advice.
3. Warm transfer, never a cold hand-off
One transfer maximum. The front line stays on the line until the specialist queue answers. Because every destination is a queue rather than a group, the caller holds where a supervisor can see them instead of disappearing into a mailbox.
Queues, not groups. This is the whole fix.
We already run both kinds of object. They perform completely differently, on the same staff, over the same thirteen days.
| Destination type | Calls | Reached a person | Ended in voicemail |
|---|---|---|---|
| Call Queues, callers hold in line | 5,137 | 69% | 30% |
| Call Groups, ring once then drop | 536 | 4% | 81% |
Every destination failing today is a group: 5011, 6524,
6525, 6526, 6528, 6529, 5063.
Every destination working today is a queue: 5000, 6518, 9000,
5045.
How the two actually route a call
The rule: if a caller can land in it, it must be a queue. Groups stay only as internal ring lists that no outside caller can reach. This one distinction turns accountability from an attitude question into a number on a screen.
The catch, stated plainly
A queue only delivers to agents who are logged in. That is its strength, because it never rings a phone on Do Not Disturb, an empty extension, or somebody already on a call, all of which the groups do today. It is also a new daily habit, and if nobody logs in the queue holds callers with no one to answer, which is worse than a group.
This is already live and worth seeing: the Live View this week showed Operator Queue with 2 agents logged in carrying 3,226 calls a fortnight, and Client Concierge with 9.
"Logged in" is not the same as "available", and the Live View only shows the first one. This is the trap.
A live test on 14 August put a caller into the Client Concierge queue. The screen read 5 agents logged in, 1 caller waiting, 51 second average wait, and the caller was still holding. All five were logged in. None were free.
A supervisor glancing at that column sees five agents and concludes the queue is covered. So the numbers that actually matter on that screen are Callers Waiting and Average Wait Time, and the shift standard has to be written about being ready to take a call, not about having logged in at nine o'clock.
What every new queue needs decided before it goes live
| Setting | Our recommendation | Why |
|---|---|---|
| Ring time per agent | 20 seconds | 5 seconds is exactly why the groups fail |
| Maximum wait in queue | 3 to 5 minutes | callers give up at a median of 47 seconds anyway |
| Overflow when that is hit | another queue, never voicemail | otherwise we have rebuilt the problem with extra steps |
| Who is ready, and when | named per shift, measured on Callers Waiting and Average Wait, not on Agents Logged In | five agents logged in and none available looks identical to full cover on that screen |
Two practical points for the discussion.
A group cannot be converted into a queue. We create the queue, add the members, then repoint whatever fed the group.
Licensing is the one open cost. Vonage licenses call queues separately from user seats and we have four today. Going to eight or nine needs confirming with our Vonage representative. It is the only item in this plan that may carry a price, and it is worth asking before we commit to a date.
Build sheet, in order
- Today: point 812-232-3246 and 812-232-9599 at queues. Between them they took 454 identifiable calls since June and answered none, because both land on call groups that ring for five seconds. At the same time, check what the CasePulse portal message and the outbound voicemail script tell clients to dial, team by team, since those templates carry the sender's number and two of the three do not currently connect. This one is urgent because it is already reaching clients in writing.
- Check the front door's no-input setting, then fix it. Three other
receptionists we inspected are each set to wait 2 seconds and then play a message and
disconnect. We have not yet opened that screen on
400000itself, so confirm it before quoting it. If it matches, change the action to forward to extension 5000 and widen the window to 5 seconds. 476 business-hours callers a fortnight reach that greeting and go no further, whatever the cause. One dropdown, safe regardless of everything else. - Create the Front Line Queue and staff it. Then create or convert Case Status, Medical Records, Hearing and Decision Support to queues. Ring 20 seconds, not 5. Untick "delete voicemail after sending email" everywhere it appears.
- Repoint every direct extension's no-answer destination to its team queue, then set outbound caller ID to that person's own direct line. These are one step, in that order. Today 76% of our outbound calls hand the client a shared number, and 41.8% hand them the main menu. This is the highest return per minute of work in the whole plan.
- Record the new greeting and assign keys 1, 2, 3 and 0. Only now, once the destinations are real, does the menu change go live. Doing this earlier just routes people into broken groups faster.
- Change the outbound voicemail script. The 135 "returning our call" cases a fortnight cannot be fixed by any menu, because the caller does not know who rang them. The message we leave has to name the team and the key to press.
- Decide what the shared numbers are for. Step 1 makes 812-234-9469, 812-232-3246 and 812-232-9599 answer. The open question is whether the 54 staff on them keep presenting them, which is reasonable once they work, or whether those teams get direct lines of their own. They have none today, which is why the failover rule in the section above cannot reach them.
How we will know it worked
Baselines are already captured, so this is measurable within a week of go-live rather than argued about.
| Measure | Today | Target |
|---|---|---|
| Callers who reach a person | 47.8% | 70%+ |
| Reached a person and still ended in voicemail | 25.8% | under 5% |
| Voicemails with no callback in 72 hours | 56% | under 20% |
| Client calls with name and date of birth verified | 28.8% | 95% |
| Calls handled by one person, no transfer | 23.1% | 45%+ |
| Repeat callers as a share of volume | 76.1% | falls as the rest improve |
What this does not fix
- After 5pm. Both after-hours options are mailboxes and the answer rate is 4%. We are not adding evening coverage, so the honest change is a closed greeting that states the hours and stops promising a callback the system cannot keep.
- Staffing the front line. This proposal assumes a team exists to answer roughly 220 calls a day at first touch. It is a real commitment, not a configuration change.
- The 47% figure. How much a front line can close is a judgment made by reading the recordings, not a measurement. Every other number in this document is counted. It is worth a supervisor listening to thirty of those calls before we staff against it.
- Twilio. All of this is interim and deliberately simple. It should make the migration easier by shrinking the routing map we have to carry across, not compete with it.