The eight reports

Filtered:GB-A1001ALARM — far off rhythm× clear the alert filterCritical + Watch× clear the ranking filterBoth filters are in the address bar, so this exact view can be linked to.Clear both filters

DescriptionValueDir

On their own rhythm29of 37 · 78.4%▲

People on rhythm (numerator)29

People with enough history to read (denominator)37

DescriptionValueDir

Promise kept2of 3 · 66.7%▲

Open cases still inside the promised window (numerator)2

Open cases attached to someone here (denominator)3

DescriptionValueDir

Roster this report can read37of 40 · 92.5%—

People with two or more attendance events (numerator)37

People on the roster (denominator)40

Group — shepherdPeopleFlaggedHeaviest alertOn rhythmDir12 weeksPercentile

Priya Nair71ALARM — far off rhythm4of 6 · 66.7%▲256/8

Groups with at least one person matching GB-A1001 under the Critical + Watch ranking filter — 1 of 8. Every rate above shows the two numbers it came from.

Confidential — prepared for Grace Fellowship volunteers. Do not forward.

End of the report. The Key, which explains every mark on this page, follows on the next sheet(s).

Key — every mark on this page

AlertIDRankingFieldWhat it means

ALARM — far off rhythmGB-A1001Criticallast attendanceAttending at 3.5× their own baseline gap or worse.

Promise brokenGB-A1007Criticalpromise clockAn open case passed the first-contact window this house promised it.

URGENT — off rhythmGB-A1002Criticallast attendanceAttending at 2.5×–3.5× their own baseline gap.

WATCH — driftingGB-A1003Watchlast attendanceAttending at 1.75×–2.5× their own baseline gap.

Promise clock runningGB-A1008Watchpromise clockAn open case is still inside its first-contact window, and nobody has called yet.

No shepherd assignedGB-A1006Notewho is assignedNobody owns this person by name, so nothing on this page has an owner either.

No attendance on fileGB-A1004Noteattendance historyOn the roster with zero attendance events — no rhythm can be read, and none is guessed.

One visit on fileGB-A1005Noteattendance historyExactly one attendance event — a baseline needs two, so no rhythm can be read yet.

Alert families — the badges on a person’s row, and the sections on their page

RhythmGB-A1001, GB-A1002, GB-A1003How this person's attendance compares to their own usual gap — never to anybody else's.

PromiseGB-A1007, GB-A1008The first-contact window this house promised on an open case attached to this person.

RosterGB-A1006What the roster itself is missing about this person, such as nobody being assigned to them.

CoverageGB-A1004, GB-A1005Whether there is enough history to read a rhythm at all. Not a judgement about the person.

A badge on a person’s row is that family’s icon and how many of its alerts that person carries. Families with none are left off the row — the strip at the top of the page already shows every zero, and repeating four zeroes on every row would bury the counts that mean something. On a person’s own page every family gets a section, including the empty ones.

Ranking — how hard one line pulls

CriticalAct today. Someone is far off their own rhythm, or a promise this house made has run out.

WatchAct this week. Drifting, or a promise clock still running with nobody on it yet.

NoteFor information. Nothing here has a clock on it — it is context the other lines are missing.

A ranking describes the ALERT, not the person. A person’s tier — WATCH, URGENT, ALARM — comes from their attendance and is the same arithmetic The Shoulder Tap uses, unchanged. Someone can sit at WATCH and still carry one Critical line, because the promise clock on their open case ran out while their attendance was only drifting. The ranking is what tells a volunteer which line to act on first.

The other marks

ExpandchevronA row with this at its right-hand end opens in place, without leaving the page or losing the reader’s spot in the list. Nothing is hidden behind it that changes a number — what opens is the evidence for the number already on the row. Turned the other way, at the top of the page, the same mark is the way back up.

ChangedflagThis signal is new, or moved tier, since 7 days before the report’s as-of date. Derived by recomputing the whole report at that earlier moment from the same history — never read back from a stored snapshot of a previous run.

▲Direction▲ ▼ —Which way it moved over the same window: ▲ worse, ▼ better, — no change.

What clears ittickThe single smallest action that makes this line go away, solved from the same arithmetic that raised it — the rhythm tiers in one case, the promise clock in the other. It is computed for each person, not written down as general advice, so the numbers in it are theirs.

No single actionstruckThe same question was asked and the answer was nothing. A promised window that has already closed is the case this report actually has: contacting someone today is still right, and it still does not un-close it. The line says so rather than offering an action that would not work.

Tile ruleunderlineThe thin rule under a scorecard tile is that alert’s severity, and it is the only colour on the tile: rust for alarm and urgent, amber for watch, plain for information. The number is the thing that changes, so the number is the thing set large.

12 weekssparkline12 weekly readings of the same rate, oldest on the left, each one a full recomputation from the same history. The scale is fixed at 0–100% so a small wobble cannot be drawn as a collapse. A tinted band is a week with NO reading — nobody in that group had enough history to compute one — and the line breaks across it rather than joining up through it. How many weeks are blank is printed beside the line.

Percentile0–100Where one group’s rate sits among the groups whose rate can be read at all, in quarters: Top quarter (75+) · Upper half (50+) · Lower half (25+) · Bottom quarter (0+). The rank and the size of the field are printed with it, because a percentile among six groups is a different claim from a percentile among six hundred. A house with one group gets no percentile at all — there is nothing to be a percentile of.

Own wordsquoteAn intake narrative or a volunteer’s note, printed exactly as it was written, with its date and — for a note — who wrote it. Nothing in a quotation is summarised, rewritten, shortened or given a mood by this build. Longer text folds at about 600 characters behind a “show the rest” link, and both halves are on the page.

Words withheldplaceholder“(words not authorized for display)” stands where a person’s words would have been when the intake record does not carry permission to display them — including when the consent question was never asked at all. The words stay in the record and are not shown; the absence is shown instead, because a blank space would read as somebody having nothing to say.

0Honest zero0A tile reading 0 means nobody carries that alert right now, and a section reading (0) means nothing in that family is flagged for this person. Zero tiles and zero sections are never hidden, and an empty table is always replaced by a sentence that says what the zero means and when it was true.

%Number / percenttoggleEvery rate on this report shows the count, the percentage and the denominator, in both settings. The control above the table only decides which of the two is set large; it cannot hide any of the three, and the setting travels in the address bar with the filters so a shared link opens on the page the sender was reading.

Service track chips

IDBNSCCNHSEMATService track chips — ID (ID walk) · BN (Benevolence) · SC (Sobriety coaching) · CN (Coaching, Counseling & Recovery) · HS (Housing) · EM (Employment) · AT (Attendance & follow-ups). Shown after a client’s name for every track currently in process; nothing shown means nothing is open right now.

Generating report…

Every signal on this report is recomputed twice — once for today and once for the start of the report window — so nothing here is read back from a stored snapshot. That is what takes the moment.

See what a clear week looks like

Demonstration data — every person shown is fictional. Grace Fellowship.

Who came back, and how?

11 of 73 people have had care reopened.

The win-back ledger, step by step.

They're back!

(The cycle)?Example — seeded data

The win-back cohorts, and where each person sits in the outreach cycle — from drifted, to reviewed, to reached, to home.

Report window: A single seeded cycle — no window

House

Grace Fellowship

Report window

A single seeded cycle — no window

Generated on Oct 2, 2026, 5:18 AM CT

This report reads a static seeded record with no history behind it, so there is no window to compare against and no “Changed” column.

The headline

11 of 73 people across 3 cohorts have had care reopened as of Oct 2, 2026, 5:18 AM CT — the stage the whole cycle exists to produce.

Generating report…

Key

Half or more of the cohort lost stepping into this stage
The biggest drop in a cohort — the one step to fix first
Counted, never named — see the Key
—No “changed” on this report; there is no history to earn one
A count of 0 is shown, never hidden.
Every mark and every stage ↓

Rates asNumber Percent

Every tile is measured against the same 73 people in a cohort — the denominator, on the page.

Twelve-week shape — care reopened, of everyone in a cohort

no weekly history

Blank on purpose. This report reads one static seeded cycle with no dates behind it, so there is no earlier week to recompute this — or any other stage’s — rate at. A fabricated shape would be the one thing this suite’s honesty law spends the most care avoiding; the honest answer is that this report has no twelve-week shape to draw until the Check-In Rail lands and gives it real receipts to compute weeks from.

StageOf everyone in a cohortThe step into itWhat it means

1In cohort73of 73 · 100.0%the denominator — nothing steps into itEveryone this cohort was drawn to cover — the denominator every other stage is measured against.

Who is at this stage — every division this build actually holds

Drifted (3–12 months)40of 40 · 100.0%

Gone (1–3 years)22of 22 · 100.0%

Long-gone (3+ years)11of 11 · 100.0%

That is as far as this report can go. It knows how many people are at in cohort and which cohorts they are in; it does not know WHO, and it will not guess. The Check-In Rail that would record a particular person moving from one stage to the next builds inside Lost Sheep P1 and has not landed — until it does, these are counts on a seeded record, not people this build can open.

2Reviewed62of 73 · 84.9%73 → 62−11 · −15%A person looked at the name and decided whether outreach was the right thing.

Who is at this stage — every division this build actually holds

Drifted (3–12 months)34of 40 · 85.0%

Gone (1–3 years)19of 22 · 86.4%

Long-gone (3+ years)9of 11 · 81.8%

That is as far as this report can go. It knows how many people are at reviewed and which cohorts they are in; it does not know WHO, and it will not guess. The Check-In Rail that would record a particular person moving from one stage to the next builds inside Lost Sheep P1 and has not landed — until it does, these are counts on a seeded record, not people this build can open.

3Approved52of 73 · 71.2%62 → 52−10 · −16%Outreach was agreed. Nobody is contacted before somebody says so.

Who is at this stage — every division this build actually holds

Drifted (3–12 months)30of 40 · 75.0%

Gone (1–3 years)15of 22 · 68.2%

Long-gone (3+ years)7of 11 · 63.6%

That is as far as this report can go. It knows how many people are at approved and which cohorts they are in; it does not know WHO, and it will not guess. The Check-In Rail that would record a particular person moving from one stage to the next builds inside Lost Sheep P1 and has not landed — until it does, these are counts on a seeded record, not people this build can open.

4Reached40of 73 · 54.8%52 → 40−12 · −23%A message or a call went out. Reaching is not replying.

Who is at this stage — every division this build actually holds

Drifted (3–12 months)26of 40 · 65.0%

Gone (1–3 years)11of 22 · 50.0%

Long-gone (3+ years)3of 11 · 27.3%

That is as far as this report can go. It knows how many people are at reached and which cohorts they are in; it does not know WHO, and it will not guess. The Check-In Rail that would record a particular person moving from one stage to the next builds inside Lost Sheep P1 and has not landed — until it does, these are counts on a seeded record, not people this build can open.

5Replied19of 73 · 26.0%40 → 19−21 · −53%The person answered. This is the first stage that needs something from them rather than from us.

Who is at this stage — every division this build actually holds

Drifted (3–12 months)14of 40 · 35.0%

Gone (1–3 years)4of 22 · 18.2%

Long-gone (3+ years)1of 11 · 9.1%

That is as far as this report can go. It knows how many people are at replied and which cohorts they are in; it does not know WHO, and it will not guess. The Check-In Rail that would record a particular person moving from one stage to the next builds inside Lost Sheep P1 and has not landed — until it does, these are counts on a seeded record, not people this build can open.

6Returned12of 73 · 16.4%19 → 12−7 · −37%They came back — attended again.

Who is at this stage — every division this build actually holds

Drifted (3–12 months)9of 40 · 22.5%

Gone (1–3 years)2of 22 · 9.1%

Long-gone (3+ years)1of 11 · 9.1%

That is as far as this report can go. It knows how many people are at returned and which cohorts they are in; it does not know WHO, and it will not guess. The Check-In Rail that would record a particular person moving from one stage to the next builds inside Lost Sheep P1 and has not landed — until it does, these are counts on a seeded record, not people this build can open.

7Care opened11of 73 · 15.1%12 → 11−1 · −8%Somebody is walking with them now. This is the stage the whole cycle exists to produce.

Who is at this stage — every division this build actually holds

Drifted (3–12 months)8of 40 · 20.0%

Gone (1–3 years)2of 22 · 9.1%

Long-gone (3+ years)1of 11 · 9.1%

That is as far as this report can go. It knows how many people are at care opened and which cohorts they are in; it does not know WHO, and it will not guess. The Check-In Rail that would record a particular person moving from one stage to the next builds inside Lost Sheep P1 and has not landed — until it does, these are counts on a seeded record, not people this build can open.

House-wide, the biggest drop is Reached → Replied: 21 people, of the 40 who got that far. Each cohort’s own worst step is named in its row below, with the reason a person wrote down.

DescriptionValue

Care reopened, of everyone in a cohort11of 73 · 15.1%

People walking with somebody now (numerator)11

People in a cohort (denominator)73

DescriptionValue

Replied, of everyone reached19of 40 · 47.5%

People who answered (numerator)19

People a message or a call went out to (denominator)40

The three cohorts, and where each one leaks

CohortIn cohortReturnedCare opened

Drifted (3–12 months)409of 408of 40

Step by step — the arithmetic, checkable

In cohort Reviewed40 34−15%

Reviewed Approved34 30−12%

Approved Reached30 26−13%

Reached Replied26 14−46%

Replied Returned14 9−36%

Returned Care opened9 8−11%

Biggest drop: Reached → Replied — Reached → replied is the steepest drop: most first outreach here is a single text. A follow-up phone call has closed roughly half of this gap in past cycles.

Gone (1–3 years)222of 222of 22

Step by step — the arithmetic, checkable

In cohort Reviewed22 19−14%

Reviewed Approved19 15−21%

Approved Reached15 11−27%

Reached Replied11 4−64%

Replied Returned4 2fewer than 5 — shown as a count

Returned Care opened2 2fewer than 5 — shown as a count

Biggest drop: Reached → Replied — Reached → replied again, and worse: contact info this old is often stale. About a third of the phone numbers on file for this cohort bounced; the deacons are cross-checking against this year's directory.

Long-gone (3+ years)111of 111of 11

Step by step — the arithmetic, checkable

In cohort Reviewed11 9−18%

Reviewed Approved9 7−22%

Approved Reached7 3−57%

Reached Replied3 1fewer than 5 — shown as a count

Replied Returned1 1fewer than 5 — shown as a count

Returned Care opened1 1fewer than 5 — shown as a count

Biggest drop: Approved → Reached — Approved → reached: this cohort's addresses and numbers are the oldest on file. A mailed card has outperformed a call here, and only two of eleven have a working number on record.

The highlighted step in each cohort is the one with the largest count-based drop; a rate on fewer than 5 people is shown as a count instead — small groups deserve names, not percentages. Each cohort’s cause line is human-authored context that is not derivable from the counts, and it says so rather than reading as a finding.

Confidential — prepared for Grace Fellowship volunteers. Do not forward.

End of the report. The Key, which explains every mark on this page, follows on the next sheet(s).

Key — every mark on this page

StageWhat it means

1In cohortEveryone this cohort was drawn to cover — the denominator every other stage is measured against.

2ReviewedA person looked at the name and decided whether outreach was the right thing.

3ApprovedOutreach was agreed. Nobody is contacted before somebody says so.

4ReachedA message or a call went out. Reaching is not replying.

5RepliedThe person answered. This is the first stage that needs something from them rather than from us.

6ReturnedThey came back — attended again.

7Care openedSomebody is walking with them now. This is the stage the whole cycle exists to produce.

The order is fixed and each stage is a subset of the one before it, so a count can never go up as you read down. Every stage is shown against the SAME denominator — everyone in a cohort — as well as against the stage immediately before it, because “half of the people we reached replied” and “a fifth of the cohort replied” are two different facts and a reader is owed both.

The other marks

Tile ruleunderlineThere is no drift tier on this report, so the thin rule under each stage tile is not a person’s own severity — it is how much of the cohort was lost stepping INTO that stage, read off the same step arithmetic the row below it already shows in numbers: rust at half or more lost, amber at a quarter to a half, plain under a quarter or too few people for a rate at all. The cut points are a first pass, not yet a PO ruling the way the five-person rate floor is.

Twelve weeksblank sparklineThe same twelve-week Sparkline the rest of the suite draws real weekly history with, and it is blank on this report — on purpose, not by omission. This report reads one static seeded cycle with no dates behind it, so there is no earlier week to recompute any rate at, the same reason this page carries no “Changed” column. A drawn shape would be invented; the blank line is the honest answer.

Biggest droprustThe one step in a cohort that loses the most PEOPLE — a count, not a rate, so a steep percentage on four people cannot outrank a shallow one on forty. Ties go to the earlier step, because the earliest leak is the one worth plugging first.

The steparrowThe move from one stage to the next, printed as both counts and, where the group is big enough, the share lost between them. A rate on fewer than 5 people is shown as a bare count instead — small groups deserve names, not percentages, and “−50%” on four people is a sentence about two people wearing a statistic.

Counted, never namedpeopleThis report can tell you how many people are at a stage and which cohorts they are in. It cannot tell you WHO, and it does not pretend to: the Check-In Rail that would record a particular person moving from reached to replied builds inside Lost Sheep P1 and has not landed. Expanding a stage gives you every division this build actually holds, and then says where it stops. The Found Ledger names the people who came back; it is a separate seeded record and its numbers do not join to these.

0Honest zero0A stage reading 0 means nobody has reached it, and no stage is ever hidden for being empty. All seven are always listed, in order, against the same denominator.

—No changed flagabsentEvery other report in this suite carries a “Changed” column earned by recomputing itself at the start of the report window. THIS ONE DOES NOT, and the absence is the honest answer rather than an oversight: it reads a static seeded record with no history behind it, so there is no earlier moment to recompute at. A flag derived from nothing would be exactly the lie the rest of the suite spends two passes per request avoiding. It arrives when the Check-In Rail does.

Generating report…

Reading the seeded cycle and dividing every stage by the cohort it came from.

Demonstration data — every cohort shown is fictional. Grace Fellowship.

Whose paperwork is stuck, and where?

2 of 5 open document chains are stuck past the expected wait.

The document chain, and where it's stuck.

The ID Walk Board

?(The document chain)

Every open ID-document chain — birth certificate, Social Security card, Texas ID — where each one is, and which are stuck waiting on an agency.

Report window: Jul 10 – Oct 2, 2026

House

Grace Fellowship

Report window

Jul 10 – Oct 2, 2026

Generated on Oct 2, 2026, 5:18 AM CT

Days waiting are counted to the moment this report was generated; the weekly readings and the completed count cover these 12 weeks.

The headline

2 of 5 open document chains are stuck past the expected wait as of Oct 2, 2026, 5:18 AM CT; 1 chain was completed in the last 12 weeks.

Generating report…

Key

alarm / urgent watch for information
The link each chain is on now
A count of 0 is shown, never hidden.
Every mark and every expected wait ↓

2 of 5 open document chains are stuck past the expected wait.

5Open chains

2Stuck past the expected wait

1Completed, last 12 weeks

A document arrived, by week1/12 blankWeek to Oct 2: 0 of 5 open

Every open document chain, stuck first. Read each row left to right: who, the three links of the chain with the one it is on now ringed, how long that link has waited, where it stands, and who shepherds it.
Person	The chain	Days waiting	Where it stands	Shepherd
Annette KowalczykIDGB-1006 · 1 of 3 in hand	Birth certificate — In hand Social Security card — Past expected wait, the link this chain is on now Texas ID card — Not yet Social Security card (replacement) requested 23 days ago — 9 days past the expected wait of about 14.	23 dayssince it was requestedexpected in ~14	Past expected wait	Priya Nair
Jerome TateIDGB-1007 · 0 of 3 in hand	Birth certificate — Past expected wait, the link this chain is on now Social Security card — Not yet Texas ID card — Not yet Birth certificate requested 19 days ago — 4 days past the expected wait of about 15.	19 dayssince it was requestedexpected in ~15	Past expected wait	Tomás Rivera
Faye LindholmIDGB-1008 · 2 of 3 in hand	Birth certificate — In hand Social Security card — In hand Texas ID card — Requested, the link this chain is on now Texas ID card requested 9 days ago — expected in about 21 days.	9 dayssince it was requestedexpected in ~21	Requested	Elena Cho
Carlos Mendez-RuizIDGB-1009 · 0 of 3 in hand	Birth certificate — Requested, the link this chain is on now Social Security card — Not yet Texas ID card — Not yet Birth certificate requested 5 days ago — expected in about 15 days.	5 dayssince it was requestedexpected in ~15	Requested	Dana Whitfield
Odessa GrimesIDGB-1010 · 1 of 3 in hand	Birth certificate — In hand Social Security card — Ready to request, the link this chain is on now Texas ID card — Not yet Birth certificate arrived 16 days ago; Social Security card (replacement) has not been requested yet.	16 dayssince the last document arrivednot requested yet	Ready to request	Priya Nair

Completed in the last 12 weeks — 1 chain

Lorenzo Pruitt GB-1005all 3 in hand by Sep 20

Confidential — prepared for Grace Fellowship volunteers. Do not forward.

End of the report. The Key, which explains every mark on this page, follows on the next sheet(s).

Key — every mark on this page

Where a link isWhat it means

Past expected waitRequested, and it has waited at least as long as this link's expected wait without arriving. Time for a human to check on it.

RequestedRequested, and still inside this link's expected wait. Nothing to chase yet.

Ready to requestEverything before it is in hand and nobody has requested it yet. The chain names no expected wait for this, so it is never called stuck — the days it has sat ready are still printed.

Not yetCannot start until the link before it is in hand.

In handThe document has arrived, and a volunteer recorded it in hand.

The chain, and what it expects

Link, in orderAgencyExpected waitFee

Birth certificateTexas Vital Statistics (DSHS) — or the birth state's registrar~15 days~$23 (estimate — confirm at first walk)

Social Security card (replacement)Social Security Administration (form SS-5)~14 daysFree

Texas ID cardTexas DPS (Identification Certificate)~21 days~$16 (estimate — confirm at first walk; fee-waiver eligibility unverified)

These waits and fees are ESTIMATES written into the engine before the first real walk confirms them, and they are printed here so a reader can see exactly what “stuck” is measured against. A link is called stuck only when it has been requested and has waited at least this long.

The other marks

Current linkringThe earliest link in the chain that is not yet in hand — the one the chain is on now. Days waiting, and the status at the end of the row, are about this link.

→Chain orderarrowThe order the chain works in: each link waits on the one before it — the Social Security card on the birth certificate, the Texas ID on the Social Security card.

Row tonedotThe suite’s own severity, and nothing new: rust when a link in the chain is past its expected wait, amber when the current link is requested and still inside it, plain when the current link is ready and nobody has requested it yet.

9Days waitingnumberWhole days, counted to the moment the report was generated, from the line printed under the number: since the current link was requested; or, if it has not been, since the document before it arrived; or since the walk opened; or since the case opened.

12 weekssparklineOne reading per week for the last 12, oldest on the left: of the chains open at any moment that week, the share that had a document arrive inside it. Each is a full recomputation from the same receipts with later ones filtered out. The scale is fixed at 0–100%. A tinted band is a week when no chain was open — there was nothing to read, so the line breaks rather than drawing a zero, and the number of blank weeks is printed beside it.

Back to the reportschevronThe one icon on this page, at its top-left: the way back to the reports index. No row on the board opens or hides anything — every chain is shown whole.

0Honest zero0A count of 0 on this page means nobody is in that state right now, and it is printed, never hidden. A board with no open chain is replaced by a sentence saying so and as of when.

Service track chips

Generating report…

Reading every case that asked for ID documents and its receipts, then recomputing each chain at the end of each of the last twelve weeks.

Demonstration data — every person shown is fictional. Grace Fellowship.

What did the engine catch, and was it right?

2 of 14 engine decisions alerted a human; 1 marked a false alarm.

Every ALARM raised, and whether it deserved to be.

The Alert Log

(Every alarm, and whether it deserved to be)?

Every decision the engine recorded — what fired, at what level, who was alerted and whether it went, and what a human made of it.

Report window: Jul 10 – Oct 2, 2026

House

Grace Fellowship

Report window

Jul 10 – Oct 2, 2026

Generated on Oct 2, 2026, 5:18 AM CT

Decisions, alerts and outcomes are counted over these 12 weeks, to the moment this report was generated.

The headline

2 of 14 engine decisions in the last 12 weeks alerted a human; 1 was marked a false alarm, and every alert has a human response recorded, as of Oct 2, 2026, 5:18 AM CT.

Generating report…

Key

alarm / urgent watch for information
False alarma human said the alarm did not fit
A count of 0 is shown, never hidden.
Every level, every delivery and every outcome ↓

2 of 14 engine decisions alerted a human; 1 marked a false alarm.

14Engine decisions, last 12 weeks

1Alarm

2Urgent

1Watch

2Alerted a human

Alerted a human, by week5/12 blankWeek to Oct 2: 2 of 7

Of the 2 alertsSMS sent1No SMS rail1No contact configured0Send failed0False alarm1Acted1Acknowledged0Not yet0

Every decision the engine recorded in the last 12 weeks, alarm first and newest first. Read each row left to right: the level, when and which decision, who it was about, what fired, who was alerted, and what came of it.
Level	When	Person	What fired	Who was alerted	Outcome
Alarm	Sep 29, 5:19 PM CTTier 2 reply screen	Rosa DelgadoGB-0996	Crisis language (1 phrase) Alarm — Tier 2 reply screen for Rosa Delgado: Crisis language (1 phrase). SMS sent — held to the verified number, not to Dana Whitfield (recipient lock on). Marked a false alarm by Dana Whitfield 59 minutes after. Why — the policy's reason crisis language detected in a Tier 2 reply (1 phrase) — routed to a human immediately	SMS sentSMS sent — held to the verified number, not to Dana Whitfield (recipient lock on).	False alarm59 minutes after · Dana Whitfield
Urgent	Oct 2, 3:17 AM CTintake read	Marcus BoydSCHSGB-1002	The engine flagged URGENT Urgent — intake read for Marcus Boyd: The engine flagged URGENT. No one — this decision raised no alert. Acted on: a note signed by Susan Marsh 1 minute after. Why — the engine's own rationale the reasoning engine flagged URGENT — human review immediately	No alertNo one — this decision raised no alert.	Acted1 minute after · Susan Marsh
Urgent	Oct 2, 2:19 AM CTintake screen	Marcus BoydSCHSGB-1002	Declared crisis-today at intake Urgent — intake screen for Marcus Boyd: Declared crisis-today at intake. Not sent — no SMS rail is configured (meant for Dana Whitfield). Acted on: a note signed by Susan Marsh 59 minutes after. Why — the policy's reason seeker declared crisis-today — human review immediately	No SMS railNot sent — no SMS rail is configured (meant for Dana Whitfield).	Acted59 minutes after · Susan Marsh
Watch	Sep 29, 3:19 AM CTintake read	Grace OkaforBNGB-1001	Lane: Bills & essentials stabilization (Benevolence arm) (medium confidence) Watch — intake read for Grace Okafor: Lane: Bills & essentials stabilization (Benevolence arm) (medium confidence). No one — this decision raised no alert. Acted on: a note signed by Priya Nair 22 hours after. Why — the engine's own rationale Reduced hours plus a new caregiving cost may outlast one bill; a short stabilization plan is worth offering.	No alertNo one — this decision raised no alert.	Acted22 hours after · Priya Nair
For information	Sep 30, 11:19 PM CTintake screen	Delia ReyesBNGB-1003	Nothing — no crisis language, no crisis declared For information — intake screen for Delia Reyes: Nothing — no crisis language, no crisis declared. No one — this decision raised no alert. Why — the policy's reason no crisis language found and no crisis declared	No alertNo one — this decision raised no alert.	Nothing raised
For information	Sep 29, 2:19 AM CTintake screen	Grace OkaforBNGB-1001	Nothing — no crisis language, no crisis declared For information — intake screen for Grace Okafor: Nothing — no crisis language, no crisis declared. No one — this decision raised no alert. Why — the policy's reason no crisis language found and no crisis declared	No alertNo one — this decision raised no alert.	Nothing raised
For information	Sep 25, 5:19 AM CTintake screen	Carlos Mendez-RuizIDGB-1009	Nothing — no crisis language, no crisis declared For information — intake screen for Carlos Mendez-Ruiz: Nothing — no crisis language, no crisis declared. No one — this decision raised no alert. Why — the policy's reason no crisis language found and no crisis declared	No alertNo one — this decision raised no alert.	Nothing raised
For information	Sep 24, 7:19 AM CTintake screen	Rosa DelgadoGB-0996	Nothing — no crisis language, no crisis declared For information — intake screen for Rosa Delgado: Nothing — no crisis language, no crisis declared. No one — this decision raised no alert. Why — the policy's reason no crisis language found and no crisis declared	No alertNo one — this decision raised no alert.	Nothing raised
For information	Sep 12, 5:19 AM CTintake screen	Marisol QuinteroCNGB-1004	Nothing — no crisis language, no crisis declared For information — intake screen for Marisol Quintero: Nothing — no crisis language, no crisis declared. No one — this decision raised no alert. Why — the policy's reason no crisis language found and no crisis declared	No alertNo one — this decision raised no alert.	Nothing raised
For information	Sep 11, 5:19 AM CTintake screen	Jerome TateIDGB-1007	Nothing — no crisis language, no crisis declared For information — intake screen for Jerome Tate: Nothing — no crisis language, no crisis declared. No one — this decision raised no alert. Why — the policy's reason no crisis language found and no crisis declared	No alertNo one — this decision raised no alert.	Nothing raised
For information	Sep 1, 5:19 AM CTintake screen	Odessa GrimesIDGB-1010	Nothing — no crisis language, no crisis declared For information — intake screen for Odessa Grimes: Nothing — no crisis language, no crisis declared. No one — this decision raised no alert. Why — the policy's reason no crisis language found and no crisis declared	No alertNo one — this decision raised no alert.	Nothing raised
For information	Aug 21, 5:19 AM CTintake screen	Annette KowalczykIDGB-1006	Nothing — no crisis language, no crisis declared For information — intake screen for Annette Kowalczyk: Nothing — no crisis language, no crisis declared. No one — this decision raised no alert. Why — the policy's reason no crisis language found and no crisis declared	No alertNo one — this decision raised no alert.	Nothing raised
For information	Aug 6, 5:19 AM CTintake screen	Faye LindholmIDGB-1008	Nothing — no crisis language, no crisis declared For information — intake screen for Faye Lindholm: Nothing — no crisis language, no crisis declared. No one — this decision raised no alert. Why — the policy's reason no crisis language found and no crisis declared	No alertNo one — this decision raised no alert.	Nothing raised
For information	Jul 22, 5:19 AM CTintake screen	Lorenzo PruittGB-1005	Nothing — no crisis language, no crisis declared For information — intake screen for Lorenzo Pruitt: Nothing — no crisis language, no crisis declared. No one — this decision raised no alert. Why — the policy's reason no crisis language found and no crisis declared	No alertNo one — this decision raised no alert.	Nothing raised

Every one of the 2 alerts in the last 12 weeks has a human outcome recorded — 0 not yet, as of Oct 2, 2026, 5:18 AM CT.

Confidential — prepared for Grace Fellowship volunteers. Do not forward.

End of the report. The Key, which explains every mark on this page, follows on the next sheet(s).

Key — every mark on this page

LevelWhat it means

AlarmCrisis language was found — in an intake story or in a reply on the after-care journey. The engine called for a human immediately.

UrgentA human was called for without crisis language: the person declared crisis-today on their intake, or the reasoning engine flagged the story URGENT.

WatchNo human was called. The engine named a lane of care at medium or high confidence — a suggestion a volunteer carries in person, never an automated offer.

For informationThe engine read it and raised nothing — no crisis language, no crisis declared, and no lane named, or a lane named at low confidence.

Who was alerted, and whether it went

DeliveryWhat it means

SMS sentThe text went out. While the recipient lock is on, it goes only to the house's verified number, not to the designated contact — the row says which.

No SMS railA contact is configured but this house has no SMS rail, so nothing was sent. The crisis pin on the casework dashboard is the surface.

No contact configuredNo escalation contact is configured for this house, so nothing was sent. The crisis pin on the casework dashboard is the surface.

Send failedThe SMS rail refused or dropped the send. Nothing reached anyone by text.

Whether it deserved to be

OutcomeWhat it means

False alarmA volunteer looked and recorded that the alarm did not fit. Counted here whatever else happened on the case — it is the answer this log exists to keep.

ActedA volunteer contact or a signed volunteer note was recorded on the case after the decision.

AcknowledgedA volunteer recorded that a human has it, and no contact or note has been recorded since. Seen, not yet acted on.

Not yetNothing a human did about this decision is on the ledger yet.

The other marks

Row tonedotThe row’s level and nothing else: rust for alarm and urgent, amber for watch, plain for information. Rows are listed alarm first, then urgent, watch and information, newest first within each.

▸WhydisclosureOpens the one sentence of why that was written when the decision was made: the reasoning engine’s own rationale on a read, the crisis policy’s reason on a screen. Never the person’s story or reply, and never the crisis phrases that matched — a screen records how many matched, not which. A rationale that repeats 5 or more words of the person’s own story in a row is withheld, and so is the engine’s rationale for a person who did not consent to their story being shown; the row says which.

12 weekssparklineOne reading per week for the last 12, oldest on the left: of the decisions the engine recorded that week, the share that alerted a human. The scale is fixed at 0–100%. A tinted band is a week with no decision at all — there was nothing to read, so the line breaks rather than drawing a zero, and the number of blank weeks is printed beside it.

Back to the reportschevronThe one icon on this page, at its top-left: the way back to the reports index.

0Honest zero0A count of 0 on this page means nothing of that kind was recorded in the window, and it is printed, never hidden. A log with no decision on it is replaced by a sentence saying so and as of when.

Service track chips

Generating report…

Reading every case's decision receipts and the later receipts that say what a human made of them, then counting each of the last twelve weeks.

Demonstration data — every person shown is fictional. Grace Fellowship.

Who has quietly stopped coming?

8 of 40 people are off their own rhythm.

Who has fallen off their own rhythm, and how badly.

The Shoulder Tap

(Ranked)?Example — seeded data

Who has fallen off their own rhythm, and how badly — ranked by how far each person is from their own baseline, never by raw days off.

Report window: Sep 25 – Oct 2, 2026

House

Grace Fellowship

Report window

Sep 25 – Oct 2, 2026

Generated on Oct 2, 2026, 5:18 AM CT

“Changed” compares against a full recomputation at the start of this window.

The headline

8 of 40 people are off their own rhythm as of Oct 2, 2026, 5:18 AM CT — ranked worst-first below, by how far each is from their own baseline.

Generating report…

Key

New or moved tier in the last 7 days
▲ worse▼ better— no change
alarm / urgent watch for information
A count of 0 is shown, never hidden.
Every mark and every alert ID ↓

1ALARM — far off rhythm

3URGENT — off rhythm

4WATCH — drifting

29On their own rhythm

3Not enough history

TableOff rhythm only (8)Everyone (37)

Rates asNumber Percent

Tile counts are everyone on the roster in that tier, filters or no filters.

Seed datafreshest record Oct 237 of 40 have enough history— details

Where this data comes from

Seed dataExample — seeded demonstration data

Most recent record on file: Oct 2 — how fresh the underlying data is, not the moment this report ran.

NameRhythmRatioShepherdChanged

1Odette BranniganALARMevery 9d — been 33d3.7×Priya Nair▲

Why this name is here — and what would clear it

AlertFieldRankingChangedMessage

GB-A1001ALARM — far off rhythmlast attendanceCritical▲Odette Brannigan usually comes about every 9 days, and it has now been 33 days — 3.7 times that rhythm. Call today — this is the longest gap on the page, not a busy week. (GB-A1001)What clears it: One visit clears this outright — attendance on file for Odette Brannigan resets the gap from 33 days to zero, and nothing here reads as drifting under 15.8 days (1.75× a 9-day rhythm). There is no heavier tier above this one; left alone the gap just grows.

Attendance record — the primary receipt

May 23
Jun 1
Jun 10
Jun 19
Jun 28
Jul 7
Jul 16
Jul 25
Aug 3
Aug 12
Aug 21
Aug 30

Baseline is the median gap between these 12 events (not the average — one long gap should not flatter or ruin the reading): 9 days. Last seen Aug 30, 33 days before this report ran.

In Odette’s own words

Verbatim. Nothing here is summarised, rewritten or given a mood.

No intake record is attached to Odette Brannigan, so there are no words of their own on this page. A case is attached only when the full name on it matches exactly one name on the roster — see the foot of this page for how many open cases that rule could and could not place. This is an absence in the ledger, not a person with nothing to say.

Everything on Odette — the Roll-Up’s person page

2Winifred AchebeURGENTevery 30d — been 87d2.9×Renee Foster—

Why this name is here — and what would clear it

AlertFieldRankingChangedMessage

GB-A1002URGENT — off rhythmlast attendanceCritical—Winifred Achebe usually comes about every 30 days, and it has now been 87 days — 2.9 times that rhythm. Worth a call this week rather than a text. (GB-A1002)What clears it: One visit clears this outright — attendance on file for Winifred Achebe resets the gap from 87 days to zero, and nothing here reads as drifting under 52.5 days (1.75× a 30-day rhythm). Left alone it becomes ALARM in 17.7 days.

Attendance record — the primary receipt

Aug 11
Sep 10
Oct 10
Nov 9
Dec 9
Jan 8
Feb 7
Mar 9
Apr 8
May 8
Jun 7
Jul 7

Baseline is the median gap between these 12 events (not the average — one long gap should not flatter or ruin the reading): 30 days. Last seen Jul 7, 87 days before this report ran.

In Winifred’s own words

Verbatim. Nothing here is summarised, rewritten or given a mood.

No intake record is attached to Winifred Achebe, so there are no words of their own on this page. A case is attached only when the full name on it matches exactly one name on the roster — see the foot of this page for how many open cases that rule could and could not place. This is an absence in the ledger, not a person with nothing to say.

Everything on Winifred — the Roll-Up’s person page

3Margaret EllisonURGENTevery 8d — been 21d2.7×Dana Whitfield▲

Why this name is here — and what would clear it

AlertFieldRankingChangedMessage

GB-A1002URGENT — off rhythmlast attendanceCritical▲Margaret Ellison usually comes about every 8 days, and it has now been 21 days — 2.7 times that rhythm. Worth a call this week rather than a text. (GB-A1002)What clears it: One visit clears this outright — attendance on file for Margaret Ellison resets the gap from 21 days to zero, and nothing here reads as drifting under 14 days (1.75× a 8-day rhythm). Left alone it becomes ALARM in 6.7 days.

Attendance record — the primary receipt

Jun 15
Jun 23
Jul 1
Jul 9
Jul 17
Jul 25
Aug 2
Aug 10
Aug 18
Aug 26
Sep 3
Sep 11

Baseline is the median gap between these 12 events (not the average — one long gap should not flatter or ruin the reading): 8 days. Last seen Sep 11, 21 days before this report ran.

In Margaret’s own words

Verbatim. Nothing here is summarised, rewritten or given a mood.

No intake record is attached to Margaret Ellison, so there are no words of their own on this page. A case is attached only when the full name on it matches exactly one name on the roster — see the foot of this page for how many open cases that rule could and could not place. This is an absence in the ledger, not a person with nothing to say.

Everything on Margaret — the Roll-Up’s person page

4Walter KimURGENTevery 10d — been 25d2.5×Elena Cho▲

Why this name is here — and what would clear it

AlertFieldRankingChangedMessage

GB-A1002URGENT — off rhythmlast attendanceCritical▲Walter Kim usually comes about every 10 days, and it has now been 25 days — 2.5 times that rhythm. Worth a call this week rather than a text. (GB-A1002)What clears it: One visit clears this outright — attendance on file for Walter Kim resets the gap from 25 days to zero, and nothing here reads as drifting under 17.5 days (1.75× a 10-day rhythm). Left alone it becomes ALARM in 9.7 days.

Attendance record — the primary receipt

May 20
May 30
Jun 9
Jun 19
Jun 29
Jul 9
Jul 19
Jul 29
Aug 8
Aug 18
Aug 28
Sep 7

Baseline is the median gap between these 12 events (not the average — one long gap should not flatter or ruin the reading): 10 days. Last seen Sep 7, 25 days before this report ran.

In Walter’s own words

Verbatim. Nothing here is summarised, rewritten or given a mood.

No intake record is attached to Walter Kim, so there are no words of their own on this page. A case is attached only when the full name on it matches exactly one name on the roster — see the foot of this page for how many open cases that rule could and could not place. This is an absence in the ledger, not a person with nothing to say.

Everything on Walter — the Roll-Up’s person page

5Bernard OseiWATCHevery 14d — been 32d2.3×Tomás Rivera—

Why this name is here — and what would clear it

AlertFieldRankingChangedMessage

GB-A1003WATCH — driftinglast attendanceWatch—Bernard Osei usually comes about every 14 days, and it has now been 32 days — 2.3 times that rhythm. Say hello before it becomes a gap. (GB-A1003)What clears it: One visit clears this outright — attendance on file for Bernard Osei resets the gap from 32 days to zero, and nothing here reads as drifting under 24.5 days (1.75× a 14-day rhythm). Left alone it becomes URGENT in 2.7 days.

Attendance record — the primary receipt

Mar 30
Apr 13
Apr 27
May 11
May 25
Jun 8
Jun 22
Jul 6
Jul 20
Aug 3
Aug 17
Aug 31

Baseline is the median gap between these 12 events (not the average — one long gap should not flatter or ruin the reading): 14 days. Last seen Aug 31, 32 days before this report ran.

In Bernard’s own words

Verbatim. Nothing here is summarised, rewritten or given a mood.

No intake record is attached to Bernard Osei, so there are no words of their own on this page. A case is attached only when the full name on it matches exactly one name on the roster — see the foot of this page for how many open cases that rule could and could not place. This is an absence in the ledger, not a person with nothing to say.

Everything on Bernard — the Roll-Up’s person page

6Louise FontaineWATCHevery 7d — been 15d2.2×Priya Nair▲

Why this name is here — and what would clear it

AlertFieldRankingChangedMessage

GB-A1003WATCH — driftinglast attendanceWatch▲Louise Fontaine usually comes about every 7 days, and it has now been 15 days — 2.2 times that rhythm. Say hello before it becomes a gap. (GB-A1003)What clears it: One visit clears this outright — attendance on file for Louise Fontaine resets the gap from 15 days to zero, and nothing here reads as drifting under 12.3 days (1.75× a 7-day rhythm). Left alone it becomes URGENT in 2 days.

Attendance record — the primary receipt

Jul 1
Jul 8
Jul 15
Jul 22
Jul 29
Aug 5
Aug 12
Aug 19
Aug 26
Sep 2
Sep 9
Sep 16

Baseline is the median gap between these 12 events (not the average — one long gap should not flatter or ruin the reading): 7 days. Last seen Sep 16, 15 days before this report ran.

In Louise’s own words

Verbatim. Nothing here is summarised, rewritten or given a mood.

No intake record is attached to Louise Fontaine, so there are no words of their own on this page. A case is attached only when the full name on it matches exactly one name on the roster — see the foot of this page for how many open cases that rule could and could not place. This is an absence in the ledger, not a person with nothing to say.

Everything on Louise — the Roll-Up’s person page

7Hollis BramwellWATCHevery 7d — been 15d2.2×unassigned▲

Why this name is here — and what would clear it

AlertFieldRankingChangedMessage

GB-A1003WATCH — driftinglast attendanceWatch▲Hollis Bramwell usually comes about every 7 days, and it has now been 15 days — 2.2 times that rhythm. Say hello before it becomes a gap. (GB-A1003)What clears it: One visit clears this outright — attendance on file for Hollis Bramwell resets the gap from 15 days to zero, and nothing here reads as drifting under 12.3 days (1.75× a 7-day rhythm). Left alone it becomes URGENT in 2 days.

GB-A1006No shepherd assignedwho is assignedNote—Nobody is assigned to Hollis Bramwell — pick a shepherd, because every other line on this list is waiting on someone to own it. (GB-A1006)What clears it: One name clears this — put any volunteer in the shepherd field for Hollis Bramwell. This line reads the field being empty and nothing else, so filling it in is the whole of the fix.

Attendance record — the primary receipt

Jul 1
Jul 8
Jul 15
Jul 22
Jul 29
Aug 5
Aug 12
Aug 19
Aug 26
Sep 2
Sep 9
Sep 16

Baseline is the median gap between these 12 events (not the average — one long gap should not flatter or ruin the reading): 7 days. Last seen Sep 16, 15 days before this report ran.

In Hollis’s own words

Verbatim. Nothing here is summarised, rewritten or given a mood.

No intake record is attached to Hollis Bramwell, so there are no words of their own on this page. A case is attached only when the full name on it matches exactly one name on the roster — see the foot of this page for how many open cases that rule could and could not place. This is an absence in the ledger, not a person with nothing to say.

Everything on Hollis — the Roll-Up’s person page

8Carol JenningsWATCHevery 7d — been 13d1.8×Renee Foster▲

Why this name is here — and what would clear it

AlertFieldRankingChangedMessage

GB-A1003WATCH — driftinglast attendanceWatch▲Carol Jennings usually comes about every 7 days, and it has now been 13 days — 1.8 times that rhythm. Say hello before it becomes a gap. (GB-A1003)What clears it: One visit clears this outright — attendance on file for Carol Jennings resets the gap from 13 days to zero, and nothing here reads as drifting under 12.3 days (1.75× a 7-day rhythm). Left alone it becomes URGENT in 4.8 days.

Attendance record — the primary receipt

Jul 4
Jul 11
Jul 18
Jul 25
Aug 1
Aug 8
Aug 15
Aug 22
Aug 29
Sep 5
Sep 12
Sep 19

Baseline is the median gap between these 12 events (not the average — one long gap should not flatter or ruin the reading): 7 days. Last seen Sep 19, 13 days before this report ran.

In Carol’s own words

Verbatim. Nothing here is summarised, rewritten or given a mood.

No intake record is attached to Carol Jennings, so there are no words of their own on this page. A case is attached only when the full name on it matches exactly one name on the roster — see the foot of this page for how many open cases that rule could and could not place. This is an absence in the ledger, not a person with nothing to say.

Everything on Carol — the Roll-Up’s person page

8 rows — ranked worst-first by how far each person is from THEIR OWN baseline, never by raw days off. Every rate below shows the two numbers it came from.

DescriptionValueDir

On their own rhythm29of 37 · 78.4%▲

People on rhythm (numerator)29

People with enough history to read (denominator)37

DescriptionValueDir

Roster this report can read37of 40 · 92.5%—

People with two or more attendance events (numerator)37

People on the roster (denominator)40

Not enough history yet — 3 people of 40Expand ↓

Fewer than two attendance events on file for each person below — no baseline can be computed yet, so nobody here is scored or flagged as off rhythm. They stay on the page instead of disappearing from the math.

Isabel Concannon1 attendance event on file
Faye Delacroix-Nwosu1 attendance event on file
Marcus WhitfieldNo attendance recorded yet