Session-level analysis of every call on the City of Midland RingCentral system: where residents reach a live person, where attempts fail, how calls are routed, and what to fix first.
legssessionsLoading current data…—inbound queues
Dashboard filters
Overview scorecards use inbound caller-experience sessions, not the full export denominator. Date, time, and detail filters narrow the inbound scorecard rows.
Loading the current call-volume and caller-experience summary.
The one finding to act on
Loading the current managed-queue and direct-line comparison.
Into a managed queueLoading…
Direct to a desk
Leg vs session: what we are counting
One call can ring several desks. Sessions, not legs, are what a resident lives.
call legs collapse into sessions.
Leg pickup punishes well-routed queues for internal re-rings. Final answer (did any leg reach a person) is the honest measure.
Every headline in this dashboard is session-level final answer unless labeled otherwise.
Where calls go
Inbound from the public dominates
Monthly volume and inbound miss rate (leg level)
Filtered by the active date and detail controls when scoped data is available.
Inbound trend
Filtered date window, leg-level inbound volume, missed calls, and miss rate.
01 Two systems: queues work, desks do not
The highest-leverage finding in the dataset, measured at the session level. The direct-dial layer behaves like unmanaged public demand.
Final answer by path
Did the caller reach a live person at all
Where the failures actually are
sessions went into queues. Only reached nobody ( final answer).
sessions were dialed to a single desk. reached nobody.
of all no-answer sessions are direct-dial in the active scope.
The fix: move published numbers off desks onto queues with backup, overflow, and voicemail-to-ticket. Cheap, fast, measurable.
02 The constituent view
Stripping internal and clearly City-origin calls barely moves the headline, which confirms the public-facing failure is real.
Loading the constituent-origin comparison.
03 Two defects hide inside good answer rates
Managed-queue performance varies materially across the current API archive. Use the table below to isolate high-volume queues with low final answer, and evaluate after-hours demand separately. Per-call wait times are not available from the RingCentral API feed; ring-time trends live on the live dashboard.
Lowest final-answer queues
The four lowest final-answer results among queues with at least 300 sessions in the active scope. This table does not test transfer recovery.
04 Department scorecard (session level)
Department ownership uses the current curated queue-owner map. Click a column to sort.
05Queue inventory
Search and sort the full inventory on final answer, no-answer sessions, cross-queue burden, and leg-level misses. API call-log data does not provide reliable wait duration.
06 Individuals and direct lines
Named missed-call and callback profile for direct lines and individual recipients. Citizen-origin calls stay citizen-origin even when they are routed through internal service paths.
Individual callback means the same listed person later placed a connected return call to the same caller number.
Callback aging curve
Cumulative callback rate at each business-day mark. Every point uses the same cohort with a complete five-business-day follow-up window.
Owner signal
Department ownership is strongest where the line carries a consistent queue or org signal.
07 Citizen-origin call centers
Queue-level missed-call and recovery profile for sessions that began outside the City. If a citizen call was routed internally, it remains citizen-origin.
Recovery is a same-caller, same-queue connected interaction within the business-day window. Rates are cumulative and are not added across deadlines; recovered by 5BD plus not recovered by 5BD equals 100%. It is a recovery proxy, not proof of a formal outbound callback.
08 Internal-origin call centers
Queue-level missed-call and recovery profile for sessions that began from City staff or internal lines. Citizen calls routed through staff are excluded from this internal-origin view.
This separates staff-origin handoff demand from constituent access. Recovery deadlines use one fixed five-business-day-eligible cohort, so the cumulative rates never decrease; recovered by 5BD plus not recovered by 5BD equals 100%.
09 Callback and recovery backlog
Prioritized worklist from the current recovery bundle: unrecovered 5-business-day misses first, then missed sessions and ownership confidence.
This backlog intentionally mixes named lines and queues because the resident does not care which routing object failed. The action owner should validate line ownership before assigning individual performance accountability.
10 Transfers and handoffs
Cross-queue paths are the practical proxy for transfers and handoff burden.
Top queue-to-queue transfers
Bar width is share of visible transfers. Department and search filters apply where queue ownership is known.
Call center to direct line routing
Date, department, status, min-session, and search filters apply. Direct routing means a queued/source-labeled call landed on a named recipient or extension instead of that source's queue number.
The Records finding, corrected
Use the filtered transfer and recovery tables above to distinguish citizen routing burden from internal handoff reliability.
11 When the city is reachable
Quality breaks at three predictable moments (lunch, the 5 PM cliff, Sunday) and there is one anomaly to investigate.
Inbound by hour: volume and final answer
Bars are call volume (legs). The line is session final answer. Watch the 5 PM collapse.
Final answer by day of week
Weekdays tight. Sunday is the weak point.
Low-pickup day anomalies
The two weakest inbound days in the current data window. Confirm operational causes before treating them as typical.
Day-by-hour heatmap of inbound miss rate (leg level)
Darker red = more missed. Hover any cell. The hot 5 PM column and the Sunday row tell the story.
Miss rate:0-5%5-15%15-25%25-40%40%+
12 Callers, noise, and recovery
Two things leadership will ask: who keeps calling, and how much of the miss rate is a real, unrecovered failure.
Recovery: the honest harm number
A missed call is only a true failure if the caller never gets through. Did each missed caller reach a person within 24 hours?
The data labels its own noise
Loading the verified dialer profile.
Several numbers are tagged SUSPECTED ROBOCALL (an Exeter Finance line at ~88% missed). Filter from service metrics.
Known Texas811 forwarding traffic is legitimate utility-locate demand, not automatically a service defect.
You can build a defensible machine-vs-human filter off these labels today.
Repeat callers and short-call noise
distinct inbound numbers. called 10+ times; called 20+ times.
High repeat rates signal low first-contact resolution: people call back because the first call did not resolve.
of calls () lasted under 10 seconds. Review these separately for likely misdials, robocalls, abandoned rings, and automated traffic.
Isolate short-call abandonment before publishing a formal service level. The direct-dial gap holds either way.
13 The Jacky telemetry gap
The most important forward-looking finding, and it is specific to you. Jacky is already live in this dataset, and the city cannot yet prove whether it is winning.
The ambiguity, and why it is the point
Reliable pre-handle wait is unavailable in the API feed. The call log also cannot tell whether a no-answer Jacky session was resolved through automation or abandoned.
If deflectedJacky resolved the question without a human. This is the value you want to prove, and it should not count as a miss.
If abandonedThe caller was stuck and gave up. This is failure. Same telephony row, opposite meaning.
RingCentral has no disposition codes, so there is no verified resolved-versus-abandoned denominator. Closing this gap in Jacky 3.0 requires logging intent, terminal state, resolution status, and escalation reason per call. Until then, Jacky no-answer sessions are unresolved signals, not confirmed failures.
How to read missed calls to Jacky
A missed Jacky leg means RingCentral recorded no handle time on that specific leg. It does not tell who disconnected, whether the caller abandoned, or whether the session later recovered.
Jacky legs were marked missed. Treat these as "not answered by Jacky in RingCentral," not as confirmed caller abandonment.
sessions that touched Jacky never reached any human. These are unresolved sessions that need terminal-state context.
of those no-answer sessions actually ended at Jacky. This is the closest RingCentral proxy for "caller hung up before Jacky answered," but it is still not proof.
If the session continued to Viviana Neptune or Adriana H Campos after Jacky, the final dead end is downstream from Jacky.
If the session reached a human after a missed Jacky leg, it should not be counted as a lost caller.
To prove abandonment vs deflection, Jacky needs terminal-state logging: resolved, transferred, escalated, caller hung up, system disconnect, or unable to help.
Jacky quality cross-check
Jacky QA telemetry is a separate source from RingCentral and is not embedded as a fixed snapshot here. Verify the live Jacky dashboard before publishing QA counts.
Do not publish pass/fail QA until manual review outcomes are populated and governed.
Audio/session watch items are triage signals, not confirmed failed calls.
Separate operational events such as idle sessions, stopped streams, widget closes, refreshes, and disconnects.
Track unknown intent as classifier coverage, not automatic service failure.
Map known demand to the queue rollout using live intent counts.
Risk candidates require review before they become findings; report severity and disposition together.
Publish outcome mix only from the governed live telemetry source.
Correction: the live all-time view strengthens the telemetry and classifier-coverage case, but it does not prove a phone-channel failure crisis. Normal hangups, abandonments, disconnects, transfers, escalations, and unresolved sessions need terminal-state classification and manual review before Voice Reliability counts can be used as failure counts.
What the no-answer sessions actually look like
Dialer (432) 224-0994 dead-end chains
14 The deflection opportunity
The call data is the clearest financial argument for the concierge you already built.
Six routine queues, mostly scriptable
First-wave deflection targets. Call-log volume is shown below; reliable wait duration is unavailable from the API feed.
Why this is the case study
missed callers never reached a person within a day.
— inbound legs into six routine queues suited to concierge triage.
after-hours sessions (about a third unanswered) that no staffing model captures.
Deflection can reduce demand in queues with persistent no-answer burden, including Health and Code routes.
Rollout order
Wave 1: Utility Billing and Animal Services. Highest volume, most scriptable.
Wave 2: Solid Waste and Golf.
Wave 3: Code and Health, where deflection can relieve persistent access burden.
Always-on: 24/7 callback capture for the after-hours and weekend gap.
Instrument deflection and post-deflection final answer so the value is provable.
15 Phased action plan
With owners. Quick wins need no new staff. Projects are scoped for the LSS portfolio. Strategic items are platform-level. Two items are flagged as highest priority by FMEA risk score.
Quick wins
0 to 30 days · ITSD-led configuration
Kill or reroute the ghost extensions
Use the live direct-line inventory to validate lines with sustained volume and near-zero answer, then retire or route them to an owned queue.
Owner: Toree Finley / ITSD
Convert public reception desks to managed routes
Move published main numbers off single desks and into hunt groups or queues with named backup coverage.
Owner: ITSD + front-desk supervisor
Stand up the daily no-answer defect list
A daily list of public-facing no-answer sessions for supervisor review.
Owner: OPI analyst (Nabeel / Noel)
Validate and govern the top direct-line backlog
Start with the highest unrecovered citizen-origin 5BD gaps: Jamie Bowers, Viviana Neptune, Jeremy Mills, Sophie Smith, and Juan Davila. Confirm owner, public-facing status, backup route, voicemail owner, and callback expectation before assigning accountability.
Owner: ITSD + department supervisors, OPI tracking
Start Jacky live QA triage
Use the live Jacky telemetry source to review safety/escalation candidates first, then run an unknown-intent coverage sprint before reporting pass/fail.
Owner: OPI / ITSD / Jacky product owner
Filter robocalls and the verified (432) 224-0994 dialer; confirm low-pickup anomalies
Remove machine traffic from service metrics and verify special-cause days before they contaminate trends.
Owner: OPI analyst + ITSD
Process projects
30 to 90 days · LSS / DMAIC
TRU-to-Records routing project (revised)
Use the current transfer-path summary to test whether the defect is routing burden rather than dropped calls. Give Records a clean path or TRU first-contact authority. Strong Green Belt charter.
Owner: DC Greg Chatwell with OPI facilitation
Records internal handoff SLA
Use the live internal-origin Records row to define when staff should transfer, warm-transfer, ticket, or use a callback queue instead of blind routing.
Owner: DC Greg Chatwell + Records supervisor
Development Services internal responsiveness sprint
Jeremy Mills, Sophie Smith, Alexis Torres, and Jeff Pinkstaff dominate internal-origin individual misses with near-zero callback recovery. Treat this as an internal escalation workflow and coverage-design problem.
Owner: Elizabeth Triggs with OPI facilitation
Lowest-final-answer queue improvement project
Use the live defect table (at least 300 sessions) to charter the highest-volume access defects; validate transfer recovery separately.
Owner: affected department directors, OPI facilitating
Best-practice transfer + lunch/5 PM coverage + taxonomy cleanup
Document Utility Billing and TRU as the SOP standard; staggered lunches and a 4:30-5:30 closer rotation; resolve the "Utiliies" typo queue and Code Admin sprawl.
Owner: OPI + department supervisors
Strategic
90 days and beyond · platform
Jacky 3.0 telemetry · FMEA risk 384
Add terminal-state classification, manual review outcomes, and disposition logging (intent, resolution status, escalation reason, normal hangup vs disconnect) so deflection, abandonment, pass/fail QA, and real audio failure are measurable. This is the gate to the xAI case study.
Owner: OPI / ITSD, you leading
Direct-extension governance program · FMEA risk 405
A public-facing number register (number, owner, hours, backup route, voicemail owner, callback SLA) for every published line. The single biggest structural fix.
Owner: City Manager sponsorship, ITSD + departments executing
Jacky deflection rollout + 24/7 callback capture
Roll out queue by queue from Utility Billing outward, using the live after-hours and weekend volume as the opportunity baseline.
Owner: OPI / ITSD
Callback SLA and live monthly dashboard
Refresh this analysis monthly as a standing control chart for the City Manager and directors, with callback recovery tracked explicitly.
