Built on provenance audit v2 · every panel marked by how we'll source it · maximizing data we can stand behind · HMIS connection treated as committed · production latest frozen at c7d14bcf
WHIT
ForKids Housing Crisis Hotline
HMIS06-15
Vonage06-24
AI06-24
ImpactBoard
ExecutiveCOO
Operationalfull detail
Data completionprovenance
Reportingfunder
live — pulled & proven nowready to source — data we hold / HMIS-confirmed; fills on connection or re-runconditional — needs the linkage build or enumeration confirmReporting window · hotline Sep 2025–Jun 2026 · HMIS enrolled through Jun 2026 · Dashboard v2 · 2026-06-25
Reach & response · what we can prove today
Families across Hampton Roads reach for help — and WHIT is measuring the demand in full.
63,020
people in crisis reached
LIVE · Sep 2025–Jun 2026
89,182
calls handled
LIVE · Sep 2025–Jun 2026
26,162
repeat contacts
LIVE · derived · Sep 2025–Jun 2026
23,118
carried by AI after-hours
LIVE · Sep 2025–Jun 2026
Outcomes — did housing hold — light up with the linkage build below. The reach story is live today.
HUD System Performance Measures · enrolled population
The scoreboard funders score us on HMIS · enrolled population · through Jun 2026
These are HUD system performance measures over the enrolled population (measured through June 2026) — distinct from the 63,020 hotline callers (≈3% overlap). Placement and income are computed from HMIS today; returns awaits the certified 6/12/24-month cohort window; diversion needs an intake flag the hotline does not yet capture.
The need
Housing-crisis demand over time ready · HMIS
Contacts over timeHMIS history · shape illustrative until pull
Read → answered volume is derivable from Vonage now; the abandoned layer needs the Akixi field confirmed in enumeration. Every lost call is a family that didn't reach help.
Read → the 4+ worklist is sourceable now; return-within-7d needs the linkage; feed health is live, WHIT Worker error rate sources from Cloudflare.
Throughput
Speed and call timing conditional · Akixi confirm
Service levelAkixi depth
<30s
share answered confirm
p90
longest waits confirm
wait
mean wait confirm
Call-time breakdownring / talk / hold / wrap
Ringconfirm
Talkconfirm
Holdconfirm
Wrapconfirm
Read → these are Akixi telephony-depth fields — represented so the view is complete, flagged to confirm the field exists in enumeration.
Demand timing
When the load hits ready · re-run
MonTueWedThuFriSatSun
12a — 6a — 12p — 6p — 11p · derivable from the timestamps we hold
By day of weekre-run
After-hoursre-run
31%
overnight/weekend
Peakre-run
Tue–Thu
11a–3p
Read → the strongest "ready" band — every call carries a proven timestamp, so timing fills with real values on the next re-run, no HMIS dependency.
AI receptionist
What the automated layer does conditional · confirm
Outcomes
Resolvedconfirm
Escalatedconfirm
Messageconfirm
Why people call
Evictionconfirm
Homelessconfirm
Rent helpconfirm
Urgency / sentiment
Urgentconfirm
Distressedconfirm
Calmconfirm
Read → outcome, intent, and sentiment are AI-derived fields we're assuming exist — enumeration confirms or cuts them.
Caller patterns & identity
Repeat demand, reach, and duplicates
Call-back frequencyre-run
1 callre-run
2–3re-run
4+re-run
Duplicates as datalive · Sep ’25–Jun ’26
27,809
collapsed dupes
26,162
repeat contacts (derived)
Reachlive · Sep ’25–Jun ’26
2,336
out of state
3.7%
outside HR (derived)
Cross-source identity spinelive · proven
HMIS UCID
→
Vonage ref ID
→
AI / MapComm ref ID
→
One account-wide ID space
Verified by UTC-instant + sequential-contiguity proof — the spine that makes dedup work and the future caller-linkage possible.
Read → the dupe/reach/identity figures are live; the frequency split is one re-run from real.
Geography
Where the calls come from ready · re-run
By localitypost-normalization
Norfolkre-run
Virginia Beachre-run
Chesapeakere-run
Portsmouthre-run
Suffolkre-run
Hamptonre-run
Newport Newsre-run
Service-area compositionthe trust signal
Hampton Roadsre-run
Out of area · VAre-run
Out of state2,336
Unspecifiedre-run
Read → the trusted base is the HMIS Region field; the AI-extracted city shows only after the normalization delta buckets out-of-area and junk values. Out-of-state (2,336) is live; the rest fills on the re-run.
Call composition · HMIS
Who is calling and why ready · HMIS
Housing statusconfirmed table
Call purposeconfirm grain
EvictionHMIS
HomelessHMIS
UtilityHMIS
Household sizeconfirm grain
1HMIS
2–3HMIS
4+HMIS
Ageconfirm grain
18–24HMIS
25–44HMIS
45+HMIS
Read → housing status is a confirmed HMIS field; purpose / household / age are HMIS-plausible — enumeration confirms the grain. All fill on the connection.
Read → these tables already exist in forkids_dw (stale at 2022–24). The connection makes them current — strongest "ready to source" band.
Source freshness
The monitoring layer is the feature live
HMIS / WellSky06-15
Vonage / Akixi06-24
AI Receptionist06-24
10 days
HMIS staleness the Sentinel caught automatically
Why → every render-test passed while HMIS sat 10 days stale; the Sentinel flagged it anyway.
Source volume over time
Each feed's contribution ready · re-run
Records per month by sourceHMIS · Vonage · AI
Why → shows when a source lags or surges — derivable from records we already hold.
Record lineage
Raw to published, with the counts live · Sep 2025–Jun 2026
Raw
116,991
93,873 HMIS + 23,118 AI
→
Deduped
89,182
27,809 collapsed
→
Enriched
23,118
GPT-5-nano · US-pinned
→
Published
63,020
unique callers · sha-verified
Why → resolves "116,991 vs 89,182" — two counts, two bases, both labeled.
Drill to record
The call-level detail behind every chart build · query layer
Records exist and are published. An interactive filter / query / export view is a build — not in this version. Sample below shows the structure.
UCID
Timestamp (UTC)
Source
Purpose
Region
Status
UC-####
2026-06-24 02:14:07
AI
Eviction
Norfolk
At risk
UC-####
2026-06-24 01:58:33
Vonage
Rent help
Chesapeake
Doubled-up
UC-####
2026-06-23 23:41:12
HMIS
Homeless
Portsmouth
Homeless
Why → the records are live and published; an interactive drill / query / export layer is a separate build, scoped for v2, not committed in this dashboard.
Field completeness
How populated each field is confirm · enumeration · through Jun 2026
Call date/time~100%
Regionconfirm
Housing statusconfirm
Demographicsconfirm
Housing outcomelinkage
Why → real fill rates land from David's enumeration; the outcome bar stays low until the linkage build.
System reliability
Service health
Feed delivery · 30 daysready · re-run
received no data
Why → feed delivery health is live — each vendor’s last 30 days of receipts, straight from the re-run.
Integrity
Provenance & protection live
Verified publish
Live = exactly what passed proof. c7d14bcf…
Residency
US-pinned; raw PII never leaves region.
PII
Scrubbed pre-publish; sources git-ignored.
Rollback
Snapshot 2d5b3cc…; one-step restore.
Why → the SOC-2 / procurement story a reviewer can check, not take on faith.
HUD System Performance Measures
What we can file to the CoC
What this is: the CoC (Continuum of Care) is the HUD-required body that coordinates every homelessness org in the region. HUD scores the whole system on seven measures pulled from HMIS — and ties funding to them. This tab maps WHIT's data to those measures.
Measure
What it shows funders
Sourcing
SPM 1 · Length of time homeless
Crisis-to-resolution speed
linkage
SPM 2 · Returns to homelessness
Did housing hold
HMIS · cohort
SPM 4 · Income / employment
Earnings gains
HMIS · partial
SPM 5 · First-time homeless
New entries prevented
HMIS · partial now
SPM 6 · Prevention & placement
Diverted from shelter
intake flag
SPM 7 · Permanent housing
Placed & retained
HMIS
SPM 3 · Persons homeless (PIT)
System context
contextual
Build · not in this version. Report generation and export (CoC report, board deck, CSV) are a build, not yet committed. The SPM table above is the spec for what they would produce.
Funder packet snapshot
The reals that anchor the report today live · Sep 2025–Jun 2026
