---
title: "How We Rate Private equity firms"
canonical: "https://papertrailfinance.com/private-equity/methodology"
as_of: 2026-11-02
updated: 2026-11-02
type: methodology
publisher: "The Paper Trail"
firm_type: "private-equity"
sources:
  - name: "SEC's Investment Adviser Public Disclosure site"
    url: "https://adviserinfo.sec.gov/"
---

# How We Rate Private equity firms

The Paper Trail profiles 433 private equity firms and scores them on version 1.8 of a written rubric; 424 have a published score. This page is the rubric as the rating program applies it: each factor and its weight, the evidence it uses, when a factor is left out, and the gate a score must pass. Most recent data: Nov. 2, 2026.

## What weights does the rating program use?

These weights are read from the scored profiles themselves, so they match every score on the site (424 published). A factor without enough evidence for a firm is left out and the remaining weights are scaled up to 100%.

| Factor | Weight |
| --- | --- |
| Conflicts and related-party disclosure | 25% |
| Disclosure consistency and timeliness | 15% |
| Fund safeguards (audit, custody, administration) | 15% |
| Regulatory and disciplinary record | 15% |
| Filing transparency | 10% |
| Independent enforcement search | 10% |
| Registration history | 10% |


Private equity firms' main investor-facing risks are fee and expense allocation and conflicts between funds, portfolio companies and the sponsor. **Conflicts and related-party disclosure carries 25%.** Fee and expense *language* sits in the Part 2A brochure, which the IAPD brochure endpoint did not serve on Nov. 9, 2026 (it returned 404 for every brochure version id, including ones that worked on Oct. 14, 2026), so brochure-text checks are not part of the rubric; the registered firm's Item 8 / 7.A / 7.B.(1) answers are what the rubric scores, and a brochure-text fee/expense check is the first planned addition (change log). Fund safeguards carries 15%: PE funds rarely use prime brokers or administrators, so those components carry little weight and valuation independence is added.

## Which firms does this rubric cover?
Universe: firms headquartered in the 13 Census West states that file Form ADV and report at least one private equity fund in Schedule D section 7.B.(1) (built by a program from the SEC's monthly adviser files; an ERA is in only after 24 months on file and after the data-quality screen). **One firm, one record.** A firm that reports several fund types is scored once under its *primary type*: the type with the most funds in its Schedule D 7.B.(1) count (ties: private equity, hedge fund, real estate fund, venture capital). Its other types are listed in other firm types on the record. A program applies the rule; the primary-type universe for this type has **433 firms** (388 SEC-registered, 45 exempt reporting advisers).

Nothing here scores investment performance, returns, fees charged, or any individual person.

## Which factors are scored, and how much does each weigh?
Each factor is scored 0-10. Weights sum to 100.

| Factor | Key | Weight | Inputs (all from public filings) |
|---|---|---|---|
| Regulatory and disciplinary record | regulatory | 15% | Form ADV Item 11 answers and counts (SEC bulk), DRP pages (Form ADV PDF) |
| Independent enforcement search | outside | 10% | Official-domain enforcement search (SEC, CFTC, NFA, FINRA, DOJ), primary pages read by hand |
| Fund safeguards (audit, custody, administration) | safeguards | 15% | Schedule D 7.B.(1): audit, auditor independence/PCAOB status, opinions, administrator, custodian, prime broker, valuation |
| Conflicts and related-party disclosure | conflicts | 25% | Form ADV Items 8 and 7.A (SEC bulk, SEC-registered advisers), 7.B.(1) Q28 marketers |
| Disclosure consistency and timeliness | disclosure | 15% | Form ADV filing date, Schedule D Q21-22, Item 9.C(2), EDGAR Form D data sets |
| Filing transparency | transparency | 10% | Item 1.I website, IAPD brochure list, Schedule D completeness |
| Registration history | stability | 10% | SEC status effective date, yearly SEC snapshots 2016-2025 |

## How is each factor scored?
**Regulatory and disciplinary record (regulatory).** Form ADV Item 11 covers the firm and its *advisory affiliates*. Item 11 answered No for every sub-question: 10, from the firm's own filing. Any Yes: every Yes sub-question must be traced to a Disclosure Reporting Page (DRP) in the Form ADV PDF, and each DRP is read for who it was filed for, its status, its dates and its monetary amount. Start at 10; for every **final** action filed for the firm (or the firm and affiliates): -3.0 if resolved in the last 5 years of the as-of date, -1.5 if 5-10 years ago, -0.5 if older, plus -1.0 when a monetary sanction is >= $250,000; floor 0. **Pending or on-appeal matters are shown and never scored.** DRPs filed only for advisory affiliates (individuals, or sister entities the filing does not show the firm controlling) are listed on the record, labeled, and not scored (adviser /F2a), whatever their kind. **Only actions by a regulator, exchange, SRO or other government body count (v1.4):** a civil-judicial DRP brought by a private party (an investor, a shareholder, a counterparty, a class) is shown, labeled "private-party civil suit", and **never scored**, final or not; a final matter whose page records no outcome (no sanction, no resolution type, no amount) is shown and not scored. Customer arbitration claims are not Form ADV Item 11 events; if one appears in a primary record it is shown and never scored. If any Yes cannot be traced to a parsed DRP, or a firm-level DRP's status, date or firm/affiliate designation cannot be read, or a criminal DRP filed for the firm has not been read by hand, the factor is **N/A**.

**Independent enforcement search (outside) - an outside-record factor (v1.1).** Built on primary bulk records. Exactly these searches back it: (1) the SEC's full lists of **litigation releases** (12,009 rows) and **administrative proceedings** (18,749 rows), 1995 to 2026-11, crawled from sec.gov (robots rules allow it) and name-matched against every firm's legal name, primary business name, IAPD other business names and the legal names in the yearly SEC snapshots (rule: normalized name on word boundaries, >= 2 words or >= 9 characters); (2) the **CFTC press-release list** (5,470 titles; cftc.gov's robots rules allow it) with the same name matching; (3) **web search** run twice for every firm (official-domain search, open-web lead search) as a **lead source only**: an official-site page it surfaces is fetched from sec.gov/cftc.gov/finra.org/justice.gov and read; open-web results are listed unscored. Not used: **NFA BASIC** (nfa.futures.org robots rules: Disallow: /basicnet/ for all agents), **FINRA Disciplinary Actions Online and Arbitration Awards Online** (robots-disallowed; not touched). Every name hit is read by a reviewer against the **primary document** and confirmed or rejected as the same firm (same-name collisions are common); only a confirmed, firm-level matter counts. Start at 10; a **resolved, firm-level** matter confirmed in a primary record that the firm's Form ADV does **not** report is scored -3.0 / -1.5 / -0.5 (+1.0 for >= $250,000); a matter already in the Form ADV is scored once, in regulatory. Pending matters and unconfirmed leads are never scored. The factor is scored **only for a firm where the SEC-list search ran and every hit has been reviewed**; otherwise it is **N/A**.

**Fund safeguards (safeguards).** From Schedule D 7.B.(1), read fund by fund from the Form ADV PDF (answers are drawn as images in IAPD PDFs; a program reads them and a program structures them). The parse is **reconciled to the SEC bulk file before it is used**: the number of fund blocks must equal the Item 7.B.(1) count, and per-type counts must agree; otherwise the factor is N/A. Components are shares of the firm's funds (each 0-10 = 10 x share), combined with these weights and **renormalized over the components that can be determined**:

- **audit** (weight 30): Annual audit: share of funds whose financial statements are subject to an annual audit (Q23(a)(1))
- **quality** (weight 20): Audit quality: among audited funds that name an auditor, share where every auditor is an independent public accountant, registered with the PCAOB and subject to PCAOB inspection (Q23(d)-(f))
- **opinions** (weight 15): Opinions: among audited funds with a yes/no answer, share whose audit reports since the last annual update all carried unqualified opinions (Q23(h); 'report not yet received' is excluded, not penalized)
- **admin** (weight 10): Independent administrator: share of funds (with a determinable answer) that use an administrator that is not a related person (Q26(a), (d))
- **custody** (weight 10): Independent custody: share of funds (with a determinable answer) that use a custodian, none of them a related person (Q25(a), (e))
- **valuation** (weight 15): Independent valuation: average share of fund assets valued by a person who is not a related person

**Conflicts and related-party disclosure (conflicts).** SEC-registered advisers only (Item 8 is not filed by ERAs; for an ERA the factor is N/A, not a free 10). Start at 10; subtract: Item 8.A(1) principal trades 1.5; 8.A(3) other proprietary interest 1.0; 8.B(1) agency cross 1.0; 8.B(3) other sales interest 1.0; 8.D a related-person broker is used 1.5; 8.G(1) soft dollars 0.5 (+1.0 if not all section 28(e) eligible); 8.H(1) pays for referrals 1.0; 8.I receives referral compensation 1.5; Item 7.A related broker-dealer 1.0, insurance company 0.5, bank 0.5, trust company 0.5; and 1.0 x the share of funds that use a third-party marketer whose marketer is a related person (7.B.(1) Q28(b) over Q28(a) Yes). Shown but **not scored** because they are near-universal for fund managers: 8.A(2), 8.B(2), 8.C(3), 8.E, 8.F, 8.H(2), and performance-based fees (5.E). Floor 0.

**Disclosure consistency and timeliness (disclosure).** Needs a parsed Schedule D. Start at 10; -3.0 if the latest Form ADV is more than 15 months old (an annual updating amendment is due within 90 days of fiscal year-end); -3.0 x the share, of the funds that say they relied on Regulation D, that give no Form D file number (Q21-22); -3.0 x the share, of the funds that say they never relied on Regulation D, that have an EDGAR Form D under the identical normalized name (EDGAR Form D data sets 2019q1-2026q1 (2026q2 not yet published on Nov. 9, 2026: HTTP 404)); -2.0 if Item 9.C(2) says an independent accountant audits the pooled vehicles annually while Schedule D reports unaudited funds. A name with no Form D match is "none found", never a finding.

**Filing transparency (transparency).** Mean of the components that apply: website listed in Item 1.I (10/0); Part 2A brochure on file at IAPD (SEC-registered only, 10/0; an ERA does not file one, so the component is N/A for it, not 0); Schedule D completeness (share of funds with name, gross asset value, number of owners and, if audited, auditor all disclosed).

**Registration history (stability).** Mean of these: tenure since the SEC status took effect, in whole years by calendar anniversary (N/A under 3 years: too short to assess; 3-5 years 7.0; 5-10 years 8.5; 10+ years 10) and legal-name changes across the ten yearly SEC snapshots 2016-2025 (0: 10, 1: 8, 2: 6, 3+: 4; SEC-registered firms only, N/A for an ERA). Item 4.A "succeeding to the business of" and ownership are shown, not scored: ownership and key-person history is not in any bulk file (change log).

## What happens when a factor cannot be scored?
A factor with no usable input is excluded and the overall is the weighted mean of the scored factors (weights renormalized) x 10, rounded once, half up. The record names which factors were scored. Missing data is never treated as good or bad.

## How is confidence set?
Points: +1 for each of Regulatory, Independent search, Fund safeguards, Conflicts, Disclosure that is scored (max 5); +1 if Registration history is scored; +1 if the firm's facts were independently verified (max 7). **Verified (v1.2)** means: the firm's SEC bulk-file facts (identity, Item 11, 7.A, 8, 9.C, fund counts and gross assets, IAPD fields) were re-derived by a separate parser written by a separate reviewer from the CSVs and IAPD JSON alone, with no mismatch (all firms: 165,932 facts, 0 mismatches). The Schedule D radio answers (audit, custodian, administrator, fund type) were eye-checked by rendered-page reading on a seeded 5% sample of firms (78 firms, 1,440 fields: the audit, custodian, administrator and fund-type answers only; it did not cover valuation, marketers, prime brokers or Form D numbers) and for the 80 trial firms (3 funds each); for every other firm they are reconciled to the SEC fund counts and types but were not individually checked by eye. The wider blind reads of every scored input are in Measured accuracy below. High >= 7, Medium 5-6, Low <= 4. An exempt reporting adviser cannot reach High: its Conflicts factor is N/A by rule. **Thin record (v1.4):** a firm with little to observe cannot reach High however many factors score: confidence is capped at Medium when the firm's SEC status took effect under 5 years ago (the same window as the rubric's most recent deduction band), or when it reports a single private fund with under $25 million in gross assets (the Advisers Act floor for SEC registration, section 203A). Affected in this type: 102 firms moved from High to Medium when the rule was introduced (88 for under 5 years since the SEC status date, 6 for a single fund under $25 million, 8 for both). **How much of the method was scored (v1.5):** every record carries share of the rubric's weight that was scored (the share of the method's weight that was actually scored) and the factors left unscored, each with its reason (each missing factor with its weight and the plain reason). When a factor of weight 25 or more is not scored, or less than 75% of the weight is scored, confidence is Medium at most and the record says why in note on confidence; when less than 60% is scored, the rating is held ("not enough of the rubric could be scored"). Measured effect in this type (every firm scored with the rule on and off): 0 ratings held, 0 confidence labels moved from High to Medium, 42 rated firms carry a note on confidence (the confidence-points rule above already kept them below High, because a firm with Conflicts or another key factor missing cannot reach 7 points). Rated firms by share of weight scored: 100%: 345, 90%: 34, 85%: 3, 75%: 30, 65%: 12.

## When is a score published?
A rating is published only when **all** hold: at least 5 of the 7 factors scored; the independent search (the outside-record factor) scored; fund safeguards scored; confidence Medium or High. Otherwise the record is a **profile, not rated** (profile without a rating, the page says why): its facts, funds and outside record are shown and its overall score is left blank. A number exists only for a published rating. **A firm with a pending government action** (a firm-level DRP from a regulator or government body, or a reviewed primary-record matter, status Pending or On Appeal) is always a profile, not rated: the pending matter is shown, labeled pending, cited to its primary record and never scored.

## How are scores distributed, and what does a percentile mean?
Rated firms: 424; min 66, P10 84.0, P25 89.0, median 93.0, P75 96.0, P90 97.0, max 100; mean 91.47, standard deviation 5.83. Confidence of rated firms: High 243, Medium 181. Scores of this type are concentrated near the top (most firms file clean records), so a firm's position is better read from its within-type percentile (percentile within its type in the underlying records: since v1.6 the share of rank-eligible firms of this type with a lower overall, ties counted half; see Ranking and data gaps) than from the number alone. The rubric anchors are not changed by this addition.

## What exactly does each term mean?
Written after an independent re-implementation of this method from this page alone; each point says what the rating program does. Inputs are the facts in the firm's record (facts[].key), plus its verification block for the verification point.
- **Elapsed years.** Every period in years (Registration history tenure bands, the thin-record test, the Regulatory and Independent-search recency bands, "historical" matters older than 10 years) is whole years by calendar anniversary: N years have elapsed on the Nth anniversary of the start date (a 29 February start counts on 1 March), as of Nov. 9, 2026. The tenure start is st.effective date (the SEC status effective date in the SEC file); st.tenure years (days / 365.25, one decimal) is shown for reading only. A matter's date is its resolution date, else its initiation date.
- **Months.** "More than 15 months old" (Disclosure) is adv.age months > 15, where months = days between the Form ADV filing date and Nov. 9, 2026 / 30.4375.
- **Regulation D shares (Disclosure).** -3.0 x ds.regd nofile n / ds.regd yes n (of the funds that answered Q21 "Yes", the share with no Form D file number in Q22; no deduction when no fund answered Yes) and -3.0 x ds.formd for regd no n / ds.regd no n (of the funds that answered Q21 "No", the share with an EDGAR Form D under the identical normalized name; none when no fund answered No).
- **Item 9.C(2) check (Disclosure).** -2.0 when item9c2 is Yes and sg.audited n < sg.audit answered n: at least one fund that answered Q23(a)(1) answered No. Funds whose answer could not be read are not counted either way.
- **Related marketer share (Conflicts).** 1.0 x rp.marketer related n / rp.marketer funds n: of the funds that use a third-party marketer (Q28(a) Yes), the share whose marketer is a related person; no deduction when no fund uses a marketer.
- **Valuation component (Fund safeguards).** sg.valued nonrelated avg / 100: the simple mean, over the funds that answer Q27, of the percentage of the fund's assets valued by a person who is not a related person. The prime-broker component (prime broker or custodian share) belongs to the hedge-fund method only.
- **Brochure component (Filing transparency).** For an SEC-registered firm: 10 when tr.brochure count >= 1 or the SEC marks it exempt from Part 2 (tr.part2 exempt = "Y"), else 0; N/A for an exempt reporting adviser.
- **Tracing Item 11 (Regulatory).** Every Item 11 sub-question answered Yes (item11.yes items, e.g. "11H(1)(b)") must be listed on a parsed disclosure page with the same sub-question, compared without dots ("11.H(1)(b)" = "11H(1)(b)"); a page that lists only the parent ("11.H(1)") does not trace a sub-question answered Yes ("11H(1)(b)"), and an untraced Yes makes the factor N/A (not a deduction).
- **No recorded outcome.** Applies to Form ADV disclosure pages. A reviewed SEC or CFTC order or litigation release is itself the recorded outcome: a resolved firm-level matter found only in those lists is deducted on the -3.0 / -1.5 / -0.5 schedule whether or not it states an amount.
- **Pending government actions.** A firm-level disclosure page that is pending or on appeal and brought by a government body (regulatory, criminal, or a civil suit the method attributes to a government), and every reviewed search-list matter that is firm-level and pending and not merged into a page, whether or not the firm's Form ADV reports it.
- **Matters that are not the firm's own or are one matter.** A page filed only for individual advisory affiliates (concerns "individual"), and a page whose respondent is a fund or affiliate entity the firm manages (whether the respondent is an affiliate), are class affiliate: shown, never scored. A second page for the same case, and a pending action that a later order resolved, are class duplicate (counted once); the Independent search sentence counts such matters once. A firm whose SEC registration was revoked in a final SEC order is a profile, not rated. The arithmetic and the rubric version (1.8) are unchanged.
- **Mixed matters: an order against the firm and an individual (no arithmetic or version change).** The concerns vocabulary is firm, affiliate, individual, predecessor, not-this-firm, unreviewed, unknown, and one *mixed* form: a matter that names the firm and an individual keeps concerns "firm" and carries the designation the filing gave "firm-and-affiliate" with kind of affiliate "individual"; the page words it "the firm and an individual associated with the firm". The firm's own sanctions score as the firm's. The individual's sanctions (a separate fine, a bar, a suspension) are stated on the row as on "an individual" (never named), are never scored and are never summed into the firm's money. A money amount ordered **jointly and severally** with an individual is the firm's liability for the whole amount: it is counted once, in full, as the firm's money, and the row says "jointly and severally with an individual"; the individual's separate amounts in the same order are not added. The audit fails a row whose text says the order is against the firm and an individual but which lacks the mixed marker.
- **Thin record.** Uses the SEC bulk 7.B.(1) count (funds.count bulk) and the bulk total of private-fund gross assets (funds.gross asset value).
- **Verification point (Confidence).** +1 when the record's verification block has at least one fact checked and no mismatch still open.
- **Precision (v1.8).** Each factor score is rounded half up to one decimal (floor(x x 10 + 0.5 + 1e-9) / 10, the same formula as ) and stored and shown that way; the overall is the weighted mean of those one-decimal factor scores (scored factors, weights renormalised) x 10, rounded half up once (a value exactly at.5 rounds up). Anyone can recompute the overall from the factor table on the page. An amount in another currency than US dollars is never compared with the $250,000 line.

## How are firms ranked when data is missing?
eligibility for ranking is true for a rated firm with at least 85% of the method's weight scored whose place in the top decile does not depend on a data gap; percentile within its type (share of rank-eligible firms of this type with a lower overall, ties counted half) and rank within its type (1 + rank-eligible firms with a higher overall) are given for rank-eligible firms only, blank otherwise, with a plain ranking note. In this type 382 of 424 rated firms are rank-eligible.

Data-gap test: every rated firm's overall is recomputed with each N/A factor scored at the type median of that factor instead of renormalising the weights. 79 rated firms have at least one N/A factor. Rank correlation between the two overalls (0.9985); firms whose overall moves by 3 points or more (16); largest change: 6 points; median change: 0.0. Firms in the top decile only because of a data gap: none (these are not rank-eligible; their rating is unchanged).

## How sensitive are the scores to the weights?
Each factor's weight was moved by +5 and -5 points, the other weights scaled so the total stays 100, and every rated firm's overall recomputed with the rubric's own arithmetic. Spearman is the rank correlation with the published overalls; "moved >= 3" counts rated firms whose overall changes by 3 points or more; "top decile kept" is the share of the published top decile still in the top decile (scores tie heavily at the top, so the decile cut moves by whole points and this share is coarse; read it with the rank correlation).

| Factor (weight) | Shift | Spearman | Median change | Max change | Moved >= 3 | Top decile kept |
|---|---|---|---|---|---|---|
| Regulatory and disciplinary record (15%) | +5 | 0.995 | 0 | 3 | 3 | 100% |
| Regulatory and disciplinary record (15%) | -5 | 0.994 | 1 | 3 | 1 | 96% |
| Independent enforcement search (10%) | +5 | 0.994 | 0 | 3 | 1 | 100% |
| Independent enforcement search (10%) | -5 | 0.994 | 1 | 3 | 1 | 94% |
| Fund safeguards (audit, custody, administration) (15%) | +5 | 0.988 | 1 | 7 | 27 | 100% |
| Fund safeguards (audit, custody, administration) (15%) | -5 | 0.984 | 1 | 7 | 27 | 79% |
| Conflicts and related-party disclosure (25%) | +5 | 0.990 | 0 | 3 | 6 | 97% |
| Conflicts and related-party disclosure (25%) | -5 | 0.988 | 0 | 4 | 5 | 93% |
| Disclosure consistency and timeliness (15%) | +5 | 0.989 | 0 | 3 | 1 | 94% |
| Disclosure consistency and timeliness (15%) | -5 | 0.989 | 0 | 2 | 0 | 97% |
| Filing transparency (10%) | +5 | 0.994 | 0 | 3 | 2 | 100% |
| Filing transparency (10%) | -5 | 0.992 | 1 | 3 | 2 | 94% |
| Registration history (10%) | +5 | 0.991 | 0 | 2 | 0 | 90% |
| Registration history (10%) | -5 | 0.990 | 0 | 2 | 0 | 98% |

Across all 14 shifts the lowest rank correlation is 0.984 (Fund safeguards (audit, custody, administration), -5 points) and no rated firm's overall moves by more than 7 points (424 rated firms).


**Plain finding.** The top of this ranking depends on the weights: with the safeguards weight -5 points, only **79%** of the published top decile stays in the top decile. The rubric is not changed; each rank-eligible firm in the top tenth carries ranking caution saying so (it is added whenever the lowest share kept is under 80%).

## How complete are the inputs?
Input availability (share of the primary-type universe):

| Input | Firms | % |
|---|---|---|
| sec bulk row | 433 | 100.0% |
| iapd json | 433 | 100.0% |
| form adv pdf served | 433 | 100.0% |
| schedule d parsed and reconciled | 431 | 99.5% |
| item11 known | 433 | 100.0% |
| item8 conflicts filed | 388 | 89.6% |
| registered snapshot history | 353 | 81.5% |
| form d name check possible | 431 | 99.5% |

Factors that can be scored from those inputs, whole primary-type universe (the independent search included):

| Factor | Firms with a score | Of | % |
|---|---|---|---|
| Regulatory and disciplinary record | 429 | 433 | 99.1% |
| Independent enforcement search | 427 | 433 | 98.6% |
| Fund safeguards (audit, custody, administration) | 431 | 433 | 99.5% |
| Conflicts and related-party disclosure | 388 | 433 | 89.6% |
| Disclosure consistency and timeliness | 431 | 433 | 99.5% |
| Filing transparency | 433 | 433 | 100.0% |
| Registration history | 385 | 433 | 88.9% |

## What did the scoring find?
433 firms (every firm of the type has a record), 424 rated and published, status counts: profile-not-rated 9, rated 424. Bulk-file facts re-derived by an independent parser for all firms: 165932 facts, 0 mismatches. Blind rendered-page reading of Schedule D radio answers, 78 firms: 1440 fields compared. Trial (20 firms) verification: Independent verification: 1317 facts checked, 5 mismatches found, 0 still open.

## What are the known limits?
- Part 2A brochure text (fee and expense disclosure) is not in rubric 1.8: the IAPD brochure endpoint returned 404 again when retried on Nov. 9, 2026 (all brochure version ids), and the adviserinfo.sec.gov firm-summary page is a JavaScript shell with no brochure link in its HTML. No workaround scraping was attempted.
- Form ADV PDFs for the very largest filers (Ares Management) are not served by IAPD ("A PDF version of the Form ADV is not available"); such firms have no Schedule D safeguards input and are profiles, not rated for every safeguards-dependent score. Where the PDF's fund list is newer than the SEC bulk count (Savvly Advisor) the reconciliation fails and the same rule applies.
- State-registered advisers (generally under $100M) are not in the SEC bulk files and are outside this universe.
- The scored Form D check is by identical normalized fund name. The data sets carry the Form D file number (021-), so the file numbers a fund lists in Schedule D Q22 are also looked up and shown (found / name differs), unscored: filings before 2019 are not in the data sets; 2026q2 had not been published (HTTP 404).
- The independent search covers SEC litigation releases, SEC administrative proceedings and CFTC press releases by name. It does not cover state securities regulators, NFA BASIC (robots-disallowed), FINRA disciplinary actions (robots-disallowed) or court dockets. Absence of a hit is "not found in those lists", not a clean bill.

## How accurate are the ratings?
Reviewers who did not build the records read the scored fund-level and firm-level inputs straight from the cached Form ADV PDFs, wrote their readings down before opening any record, and then compared.

- **First blind read** (seed 20261111, 10 firms): 932 fields, 39 wrong (4.2%). Almost all were one question the parser had not read (Schedule D Q27, the share of a fund's assets valued by a person not related to the adviser), so the valuation part of Fund safeguards was being dropped. The parser was fixed and every firm rescored (v1.7).
- **Fresh blind read after the fix** (seed 20261112, no firm from the first read, 10 firms): 639 fields, 1 wrong (0.16%) (a filing date: the cached PDF is a newer amendment than the SEC bulk file the record uses); under the 1% bar.
- **Re-implementation from this page alone:** a separate reviewer rebuilt the scoring from this text and reproduced every factor score, overall, confidence label and publication decision for every firm (Nov. 11, 2026, rerun after the and record changes; the first full run was the underlying records).

## Why do so many scores sit high?

Most private equity firms in the record have no serious regulatory, legal or financial problem on file, so published scores bunch near the top: the median is 93 of 100, 309 of the 424 scores are 90 or higher and 6 are 100. The score separates firms that have a record of enforcement, disputes or weak finances from firms that do not; it does not rank one clean-record firm above another, and a difference of a few points between two high scores is not meaningful. Where firms share a score, the lists of highest scores on the type's page put the larger firm first and say that ties are ordered by size.

## How complete are the fields on these profiles?

For each field, how many of the 433 private equity firm profiles carry a value; how many have the source read and the field blank or not reported there ("not disclosed in the source"); and how many have no such fact at all ("not collected").

**Field completeness, 433 private equity firm profiles**

| Field | Present | Not disclosed in the source | Not collected |
| --- | ---: | ---: | ---: |
| Name | 433 (100.0%) | 0 (0.0%) | 0 (0.0%) |
| City | 433 (100.0%) | 0 (0.0%) | 0 (0.0%) |
| State | 433 (100.0%) | 0 (0.0%) | 0 (0.0%) |
| CRD | 433 (100.0%) | 0 (0.0%) | 0 (0.0%) |
| Registration type | 433 (100.0%) | 0 (0.0%) | 0 (0.0%) |
| Latest Form ADV filing date | 433 (100.0%) | 0 (0.0%) | 0 (0.0%) |
| Private funds (count) | 433 (100.0%) | 0 (0.0%) | 0 (0.0%) |
| Private-fund gross assets | 433 (100.0%) | 0 (0.0%) | 0 (0.0%) |
| Fund audit answers | 431 (99.5%) | 2 (0.5%) | 0 (0.0%) |
| Auditor names | 431 (99.5%) | 0 (0.0%) | 2 (0.5%) |
| Valuation by non-related person (Q27) | 431 (99.5%) | 0 (0.0%) | 2 (0.5%) |
| Item 11 disciplinary answers | 433 (100.0%) | 0 (0.0%) | 0 (0.0%) |
| Website | 433 (100.0%) | 0 (0.0%) | 0 (0.0%) |
| Registration start date | 433 (100.0%) | 0 (0.0%) | 0 (0.0%) |

## What changed in each version?

- **Data note, Nov. 11, 2026: no rule change, rating program version 1.8 unchanged.**
- **Data note, Nov. 11, 2026: no rule change, rating program version unchanged.** Who brought each matter, as the filing shows it: a disclosure page whose details are blank because the affiliate reported the event on its own CRD/IARD record (Part I.B answered Yes) says so.
- **Version 1.8, Nov. 10, 2026: one-decimal factor scores.**
- **Version 1.7, Nov. 10, 2026: independent re-implementation and blind read.** What changed: tenure bands on whole years by anniversary; the overall is computed from unrounded factor scores; exact definitions. Exact definitions written for every input the independent build found ambiguous; the rating program's denominators were kept because each is the set of funds the test applies to.
- **Version 1.6, Nov. 10, 2026: data fixes.** What changed: who brought each matter; currency as filed; rank eligibility; Data-gap sensitivity. A private civil suit is "Private plaintiff" (never named), kind private party, with the court where the filing names one, and its display text says so (it had shown as an SEC matter on profile pages because the page fell back to the source's host). Amounts by currency across the four types: USD 148, CHF 3, EUR 1 (the non-USD amounts are an EUR fine and three CHF penalties, all filed for affiliates or under $250,000, so no score changes).
- **Version 1.5, Nov. 10, 2026: data fixes.** What changed: coverage rule; one matter per docket; concerns; names and history. No matter had been scored twice (every merged search row was already marked as reported on Form ADV), so no score changes; the audit now fails on a docket recorded twice.
- **Version 1.4, Nov. 10, 2026: research-data fixes.** What changed: private suits and no-outcome matters are not scored; Thin-record confidence cap; Within-type percentile and score distribution; published fields are plain.
- **Version 1.3, Nov. 9, 2026.** The v1.2 identity rule was too strict and hid the firm's own public record (Franklin Advisers, Ares Management).
- **Version 1.2, Nov. 9, 2026.** Sanction amounts re-extracted from the FULL, uncapped text of every confirmed SEC/CFTC document (the 20,000-character cap had cut 21 of 97).
- **Version 1.1, Nov. 9, 2026.** What changed: profile facts added to every record; parser.
- **Version 1.0, Nov. 9, 2026.** First version: seven factors, weights above, publication gate, confidence points.

This page is the written rubric the rating program follows, with file names and internal codes put in words. Individuals are never rated, pending matters and allegations are shown but never scored, and missing data is never counted as good or bad. Information, not investment advice.
