Diligence Index

RX 05.3 · INDEX v3 · 07 NOV 2025 · WINDOW MONTHS 1–34 · JAN '23 — OCT '25 · PARTNER-EQUITY DSO · SOUTHEAST · 6 REGIONS · 42 LOCATIONS · SPECIMEN

WALKED TAB BY TAB WITH THE CFO AND THE BUYER'S QUALITY-OF-EARNINGS TEAM · 45 MIN · AT DATA-ROOM OPEN · REFRESHED MONTHLY WHILE THE PROCESS RUNS

CONFIDENTIAL · RX 05FORM 884-D

Attribution history, cohort data, capacity models — in the data room before anyone asks for them. This document is mostly a table of contents.

READ THIS BEFORE THE INDEX — IT IS THE WHOLE COMMERCIAL ARGUMENT

Almost nothing in this package was made for this transaction. Every item in tabs 01 through 06 was produced on its own cadence during the engagement — a baseline reading taken at Day 0, thirty-two monthly board packs signed by the controller in the month they closed, eleven quarterly recommitment memos, twenty-eight months of loop minutes with a named owner on each. They were made by people who did not know a buyer would read them. That is what makes them worth reading.

The cost of assembling this index was thirty-one hours. It is thirty-one hours because the work was already done, in small monthly amounts, for three years. A group that did not instrument from Day 0 pays for its data room in a single terrified quarter, retroactively, from memory — and the retroactive version is the one that gets marked down, because a diligence team can tell the difference between a record and a reconstruction in about ten minutes. Nine of the seventeen items in this index could not have been created after the fact at any price (sheet 3).

ITEMS INDEXED

7 complete for the full window · 10 partial, each gap named · 0 absent

WINDOW

34 months

JAN '23 — OCT '25 · 34 closed periods

ASSEMBLY COST OF THIS PACKAGE

31 hours

18 analyst · 7 principal · 6 tracking · nothing in tabs 01–06 was created for it

COULD NOT BE CREATED AFTER THE FACT

9 of 17

at any price, by anyone — see sheet 3

COVERAGE BY TAB — AND WHY THE HEADLINE NUMBER IS NOT 100%

THE 83.4% HEADLINE IS AN UNWEIGHTED MEAN OF SEVENTEEN COVERAGE FIGURES WITH DIFFERENT BASES — MONTHS, SITES, PLANNED REFRESHES, SIGNED PACKS. IT IS A SUMMARY, NOT A MEASUREMENT. READ THE ROWS. AN INDEX CLAIMING 100% COVERAGE IS NOT BELIEVABLE AND WOULD BE TAKEN APART IN AN AFTERNOON — a three-year engagement across nine partners and six regional P&Ls does not produce complete data, and saying it did is the fastest way to make a buyer doubt the 83.4% that is real.

TAB 01 · METHOD & DEFINITIONS — THE TAB EVERY OTHER TAB DEPENDS ON

ITEM	FORM / SYSTEM	PERIOD	COVER.	RE-RUNNABLE?	NOTE
Baseline reading — Day-0 vitals, method and confidence flag on every number	247-BPMS · GL · CALL LOGS	Day 0 · month 1	100%	YES — source exports retained	The reading the client keeps either way. Everything downstream is measured against it, unrestated.
Diagnosis — constraint, ranked changes, go/no-go	247-DDOCUMENT	month 2	100%	N/A — JUDGMENT	An opinion written before the outcome, and dated. Its value in a data room is that it can be graded.
Definitions register — frozen metric definitions, cohort rosters, site open-date register, staffing normalisation rule	TAB 01CLIENT-OWNED SINCE MONTH 6	months 1–34	100%	YES — it is the method	Time-stamped at freeze. This is what lets a stranger re-cut any cohort in this package any way they like.

TAB 02 · DEMAND & CAPACITY — THE WEAKEST TAB IN THE PACKAGE, AND IT IS SECOND SO NOBODY HAS TO FIND IT

Demand curve vs capacity curve, per market — including the markets scored zero-spend, with reasons	118-MSEARCH · PMS · DRIVE-TIME	6 refreshes planned per region	67%	PARTIAL — 4 OF 6 REGIONS	12 of 18 planned refreshes delivered. Two regions never supplied chair-hour data and could not be compelled to.
Capacity models — chair-hours available, utilised, open, by pod	118-MPMS SCHEDULING	months 8–34	67%	PARTIAL — 4 OF 6 REGIONS	Same two regions. A third of the group has no instrumented capacity side, so any group-wide latent-capacity figure is stated for four regions only and labelled as such.

TAB 02 IS PLACED SECOND DELIBERATELY. IT IS THE THINNEST TAB IN THE PACKAGE AND THE REASON IS GOVERNANCE, NOT NEGLECT — IN A NINE-PARTNER EQUITY STRUCTURE, TWO REGIONS DECLINED TO SUPPLY SCHEDULING DATA AND NO ONE IN THE BUILDING COULD MAKE THEM. WE WOULD RATHER A BUYER READ THAT ON PAGE ONE THAN DISCOVER IT ON DAY FORTY.

Index — acquisition, intake, finance

TABS 03–05 · ELEVEN ITEMS · EVERY ONE PRODUCED ON ITS OWN CADENCE DURING THE ENGAGEMENT · SPECIMEN

CONFIDENTIAL · RX 05FORM 884-D

Coverage percentages are stated against a named base — months, sites, packs, or planned refreshes. Where the base differs, it is written in the row rather than averaged away.

TAB 03 · ACQUISITION & ATTRIBUTION

ITEM	FORM / SYSTEM	PERIOD	COVER.	RE-RUNNABLE?	NOTE
Attribution history — spend, attributed new patients, loaded cost per new patient, by channel, by month	610-M p.3CALL TRACKING · GA4 · PMS	months 1–34	73%	YES FROM MONTH 13 · NO BEFORE MONTH 7	Channel coverage weighted: 2 of 12 channels at Day 0, 7 of 12 by month 12, 11 of 12 from month 13. No channel-level claim before month 13 is supportable, and none is made.
Tracking configuration change log — every container, tag, call-tracking pool and conversion definition, dated	TAB 03GTM · GA4 · CALL VENDOR	months 4–34	91%	YES	31 of 34 months. Before month 4 no change log existed, so the original configuration cannot be reconstructed — only described.
Two-market pre-rollout test — 11 weeks, before anything was rolled outward	TAB 03PMS · CALL TRACKING	months 4–6	100%	NO — THE WINDOW CLOSED	Complete for its own window and not reproducible: the test ended. It can be re-read, not re-run. It establishes a pre-rollout read on two markets, not a group-wide counterfactual.

TAB 04 · INTAKE — THE DIFFERENTIATING TAB, AND THE ONE WITH OUR OWN ERROR IN IT

Call recordings and scored transcripts — answered, offered, booked, against the one-page rubric	247-KCALL VENDOR	months 5–34	88%	NO FOR MONTHS 1–4	30 of 34 months. Months 1–4 were destroyed by the vendor's 90-day default retention before we changed the setting. That was our error and it is stated here, not in a footnote. Nobody can re-score that quarter, including us.
Desk readouts — answer rate and booking rate, per desk, per daypart	247-KCALL VENDOR · PMS	months 6–34 · 16 sites	38%	YES — FOR THE 16 SITES	16 of 42 sites. This is an adoption number, not a data defect — three of six regions bought into the programme. Run weekly by the client's own desk managers; ADMEN attends monthly and on exception.
Desk → media loop minutes — what the desk heard, what changed in the media, who owned it	247-LDOCUMENT SERIES	months 7–34	82%	YES — AND THE PARTICIPANTS ARE INTERVIEWABLE	28 of 34 months, each with a named client owner and an attendance log. The loop was built at month 7; there is nothing before that because there was nothing before that.

TAB 05 · FINANCE CADENCE — THE TAB A QUALITY-OF-EARNINGS TEAM OPENS FIRST

Monthly board packs — vitals vs plan vs baseline, channel economics, capacity, assessment & plan, definitions appendix	610-MPMS · GL	months 3–34	94%	YES — TIE TO TRIAL BALANCE	32 of 34. Months 1–2 predate the definitions freeze and are excluded rather than restated. Every pack contains at least one regression printed at the same weight as the wins.
P&L tie-outs — attributed collections reconciled to GL 4000, signed by the client's controller	610-M stripGL · CONTROLLER	32 packs	100%	YES — PICK ANY MONTH	32 of 32. Signed before publication, by a client employee, in the month it closed. Variance to PMS under 1%; standing rule, the PMS wins.
Monthly readout — the one-page stakeholder version, with its own open-rate instrumentation	610-RDOCUMENT SERIES	months 15–34	59%	YES	20 of 34. It replaced a weekly series at month 15 — see gap G-04, sheet 3. Open rate is reported in the pack including the three months it fell below 50%.
Quarterly recommitment memos — continue / conclude, each with a pre-filled case for ending the engagement	610-QDOCUMENT SERIES	11 quarters	100%	YES	11 of 11. An agency that writes the argument for its own dismissal four times a year, in writing, and is retained anyway, is describing a relationship rather than a lock-in.

EVERY ITEM ON THIS SHEET WAS PRODUCED IN THE MONTH IT DESCRIBES, BY PEOPLE WHO DID NOT KNOW A BUYER WOULD READ IT. THAT IS THE ONLY PROPERTY OF THIS PACKAGE THAT CANNOT BE PURCHASED LATER, AND IT IS THE PROPERTY A DILIGENCE TEAM IS ACTUALLY TESTING FOR WHEN IT ASKS FOR SOURCE FILES.

Outcomes — and what this package does not have

TAB 06 · TAB 07 GAPS & LIMITATIONS · NINE ITEMS THAT CANNOT BE RECONSTRUCTED · SPECIMEN

CONFIDENTIAL · RX 05FORM 884-D

Tab 07 is the reason the other six tabs are believable. It is written by us, before anyone asks, and it is the first tab we walk.

TAB 06 · OUTCOMES & COHORTS

ITEM	FORM / SYSTEM	PERIOD	COVER.	RE-RUNNABLE?	NOTE
Post-op vitals — baseline vs retest, same instruments, same locations, same definitions	247-PPMS · GL · CALL LOGS	retest at month 34	100%	YES	Mirrors Form 247-B row for row, with a delta column and the misses left in. Definitions frozen at baseline and never re-cut to flatter the retest.
De novo cohort files — per-site ramp curves for the 12 sites opened inside the window	TAB 06PMS	opened months 5–29	58%	PARTIAL — 7 OF 12 SITES	7 of 12 sites carry 12+ months at index date; five are shorter and are marked short. These curves let a buyer read the blended decline as a ramp. They do not oblige the buyer to.

TAB 07 · GAPS & LIMITATIONS — WRITTEN BY US, IN ADVANCE, AND WALKED FIRST

G-01

Attribution before month 13

Two of twelve channels were traceable at Day 0; seven by month 12; eleven from month 13. No channel-level claim before month 13 is supportable and none is made anywhere in this package. The untracked spend of the first year is gone and cannot be re-attributed by anyone.

G-02

Call recordings, months 1–4 — our error

The call vendor's default retention was 90 days and we did not change it until month 5. Four months of recordings were destroyed. The leak figure quoted in the growth narrative rests on the sit-week sample and on months 5 onward. We could have described this as a vendor limitation. It was our omission.

G-03

Two regions never supplied capacity data

Nine partners, six regional P&Ls, no mandate. Two regions declined to supply chair-hour scheduling data for the full engagement. The demand/capacity map and every capacity model cover four of six regions, and every latent-capacity figure in this package is stated for four regions only.

G-04

The weekly readout series is archived, not indexed

A weekly readout ran months 1–14. It is in the archive and available on request, but it is deliberately not indexed as evidence: at weekly sample sizes the per-desk rates it reports cannot carry a decision, and coaching a person on noise damages the instrument and the person. It was replaced at month 15 by the monthly readout (610-R). We will not hand a diligence team a series we would not defend.

G-05

Payer-mix coding break at month 9

Three regions coded payer class inconsistently until a unification at month 9. The pre-month-9 mix series is not comparable to what follows. We state the break rather than restating history — a restated series is a series nobody can check.

G-06

Broken-appointment definitions before month 6

The baseline reading flagged this instrument LOW confidence at Day 0 and it stayed low until definitions were unified at month 6. The pre-month-6 series is excluded from every comparison in this package, including the ones where including it would have helped.

G-07

Five de novo sites have short curves

Five of the twelve sites opened inside the window carry fewer than twelve months of data at the index date. Their ramp curves are short, marked short, and should not be averaged with the seven mature ones. n = 12 is small before it is split.

G-08

No holdout, therefore no incrementality

No post-rollout holdout was ever run, and in a partner-equity structure no region would withhold spend from its own P&L to improve someone else's measurement. There is no incrementality evidence in this package and none is claimed. Attribution here is last non-direct touch — a rule for assigning credit, not a method for establishing a counterfactual (Form 884-N, C-05 and declined claim D-03).

G-09

No post-close evidence exists

KPI retention through the first two post-close quarters is the signal that would prove the operating system survives a change of ownership. It cannot exist before a close. Anyone claiming it does is describing a different transaction.

WHAT COULD NOT HAVE BEEN BUILT AFTER THE FACT — NINE OF SEVENTEEN

A Day-0 reading cannot be taken on day 1,030. A diagnosis written before the outcome cannot be written after it. Definitions cannot be frozen backwards, a controller cannot sign a reconciliation for a month that closed two years ago, destroyed call recordings cannot be re-scored, and a memo arguing for ending an engagement is worth nothing if it is written once the engagement has ended well. Nine of the seventeen items here are of that kind. The other eight could be rebuilt at cost — the market maps, the capacity models, the cohort files, an unsigned recompute of the P&L tie-outs.

This is the sentence a buyer's advisers will recognise and a seller's competitors will not enjoy: pre-exit cost this group thirty-one hours because the work was already done. An uninstrumented group of the same size spends a quarter reconstructing what it can, discovers it cannot reconstruct the rest, and presents a narrative supported by memory — which is the version that gets marked down, re-traded, or walked away from. You cannot buy three years of contemporaneous records in the ninety days before a process. That's the whole product.

— specimen —
PARTNER, ADMEN · INDEX OWNER
— specimen —
CLIENT CFO · CONFIRMS TABS 01–07 MATCH THE DATA ROOM
— specimen —
WALKED WITH QUALITY-OF-EARNINGS TEAM · 45 MIN · DATE