How UMFR measures and what it refuses to claim
1. The number path
Every figure is read from its source — a tagged XBRL fact or the text of a credit agreement or compliance certificate — or computed deterministically from figures read there. No language model reads the documents or decides the numbers on any served path. Where a figure cannot be read or computed, the report declines rather than estimates, and names what is missing.
2. The public report
On the public report, every figure read from a filing cites its concept and period. Checks: balance-sheet identity and cross-foot; retained-earnings roll-forward (retired shares, cash repurchases and discontinued operations are printed beside the residual, never subtracted; base floor: this report's own floor of 1% of total assets); supplier-finance leverage and share of trade payables (this report's own floor of 1% of funded debt; tiers of 2% and 10%, this report's own uncalibrated tiers); mark-to-cost; unexplained debt movement; accrual versus cash; PIK share of investment income (investment companies only); restatement discontinuity corroborated by 8-K Item 4.02; forensic screens — Beneish M, Altman Z'', Piotroski F, Montier C, Sloan — with their inputs printed. Thresholds: Beneish −1.78 (Beneish 1999, p. 17); Altman Z'' 1.10 (Altman 2002, §8, p. 22); Altman 2.6, Piotroski ≤3, Montier ≥4 and Sloan >0.1 are this report's own uncalibrated flags. Banks and broker-dealers: cash-flow screens do not apply and say so. Reporting currency is detected; USD-only checks say so.
3. The private report
The agreement and its amendments are resolved to the operative covenant-EBITDA definition. Each term is classified: recomputable from reported lines, or not — with the reason, a quotation from the agreement and the page it sits on, or the reason a page cannot be given. The compliance certificate's arithmetic is recomputed from its own components; headroom is stated once, at the certificate's precision; the cap is printed as the document writes it. Where dated restatements exist, each certificate is resolved against the text that governed its period. Chain completeness is never confirmed; it is stated as not confirmed. Internal-consistency checks do not run on private packages, and the report says so. Borrowing-base and cross-document collateral reconciliation are reserved and not built.
4. The measurement
Measured on two pre-registered draws of 24 credit agreements each, drawn by a fixed rule from EDGAR 8-K Item 1.01 exhibits (filed 2024 and 2023), with no agreement in common: of the terms that send a covenant-EBITDA definition outside the agreement, the share no one outside the deal can recompute had a median of 79% (2024, 11 agreements measured) and 78% (2023, 12 agreements measured). In 21 of 23 it was half or more; two definitions had none; three agreements were set aside because this reader could not establish where one of their definition's terms ends. The draw, the rule and every per-agreement result are on the Methodology page.
| Filed year | Agreements drawn | Resolvable | Measured | Min / median / max (%) | Set aside |
|---|---|---|---|---|---|
| 2024 | 24 / 24 | 12 | 11 | 0 / 79 / 100 | 1 |
| 2023 | 24 / 24 | 14 | 12 | 0 / 78 / 100 | 2 |
Per-agreement shares (%), as the measurement prints them:
- 2024: 0, 60, 67, 76, 78, 79, 79, 84, 87, 100, 100
- 2023: 0, 50, 57, 62, 75, 77, 79, 82, 82, 84, 86, 100
Across the 24 agreements of the 2024 draw, schedules are cited 215 times; 7 are attached; 8 are withheld by the filer; 102 are absent with no marking; 98 could not be determined.
The measured draw was committed on 2026-09-21, before the public sample existed; the definition-end rule was corrected on 2026-10-10 and the whole draw re-derived by the same code.
5. The receipt
Each report carries a SHA-256 digest over the assembled report and attaches the canonical result object inside the PDF; a recipient reproduces the digest without UMFR. The digest proves that the report you hold is the one the digest was computed over; it does not by itself prove who issued it. Signed receipts and a public log of receipt digests are the next step.
6. Limits, stated
Not an audit, not an agreed-upon-procedures engagement, not an opinion, not a rating, not advice. Public-filing results are not equivalent to private-package results. Forgery by file forensics is not performed: an internally consistent forged document passes.
What runs on public data — and what needs the documents
One engine, run deterministically — each proof below is a shipped output. Read from source or computed from it. Every report and covenant card carries a receipt you can recompute in your browser to confirm it is unaltered since generation.
Tier 1 — deterministic-public · public SEC XBRL · zero compute cost · reproducible
These run on an issuer's own public SEC or ESEF filings, free. The worked cases are open on the demo.
The engine reads IFRS filers by their home ticker at the financial-statement tier — demonstrated on a real Form 20-F reporter (Ads-Tec Energy) — and is architecturally standard-agnostic. Definition-level covenant support for IFRS is a bounded extension, to be measured on a real IFRS agreement before it is claimed.
Restatement discontinuity
The Coverage Statement — live
Tier 1 runs on an issuer's own public SEC filings, deterministically, at zero compute cost — the worked cases are open on the demo. Tier 2 goes deeper, on the source agreements and schedules public data does not carry — the document path is built and running today, on the borrower's own documents. Both are deterministic in their output. Neither uses a language model to decide the numbers a reviewer sees. We are as clear about which tier a result comes from as we are about the result.
Tier 2 — document / agreement forensics
The deeper checks the engine runs on a borrower's own agreements, schedules and source documents — built and shipping, demonstrated here on real public cases. Read from source or computed from it.
What the document path does today
A lender hands over the credit agreement, its amendments, and the compliance certificates. The engine reads them in memory and returns a rendered verification document — one a credit officer can read, save, and take into an investment-committee file — in which every layer lands in one of three explicit document-path states — run (a built check consumed a supplied document and produced a result), reserved (a check we have not built yet, named as a reserve, never omitted), or absent (a built check with no document to consume, a gap the lender can close). Nothing is silently skipped, and a state is never a verdict.
Before it computes anything, the engine identifies what each document actually is by reading it — a credit agreement, a compliance certificate, a borrowing-base certificate, or something it sets aside with context; and within the credit-agreement family, a base agreement, an amendment, an amended-and-restated, a joinder, or a conformed copy, telling a guaranty apart from an amended-and-restated.
A covenant is recomputed against the operative amended text, and where the chain cannot be shown complete the engine declines to certify completeness and says so. All three public cases resolve as chain-complete, so the decline does not fire on them — the same discipline the rest of the page runs on: silent when things are clean, speaking only when they are not.
How your documents are treated
Receipt-provenance honesty.A private result is marked as the unverified-provenance class, visibly distinct from a public SEC-filed receipt — the system says, on its own output, that the source was your private document and not an independent public filing.
No persistence. The bytes live only in memory; nothing is written to disk or blob storage; logs carry counts and states only, never document contents or a filename.
Session isolation. The raw filename is never read; files are handled positionally; only the anonymised borrower label is carried.
The office-format front end — a scanned-PDF, spreadsheet, or OCR'd certificate from a private borrower, with no public XBRL to cross-check — is scoped, not built: it will be measured on a real lender certificate before it is claimed.
Definition-level covenant forensics
- Invesco — the agreement's Covenant Adjusted EBITDA runs +50.7% above plain GAAP EBITDA ($1,557.0M vs $1,033.0M); both covenants recompute as comfortably compliant — a difference of definition, not a deficiency.
- Flow — an agreement-governed add-back of 31,400 is larger than the entire 25,486 Adjusted EBITDA it sits inside. Surfaced for review, never asserted as a breach.
- Yellow — a capped add-back appears to exceed its stated cap: a cap-breach signal on a presented measure (a management deck under waiver), never an in-force or adjudicated breach.
See the live covenant card · See the Paya Holdings teardown— a finished covenant read on a named public issuer, assembled from its public SEC filings.
Hidden leverage
Document authenticity
At-scale discipline
Structural early-warning — a capacity read, not a recompute
See it run
Different issuers, different credit types, one engine. Read from source or computed from it.
Five assurance levels
An open framework UMFR publishes and uses — and one the market may adopt — but not an industry-adopted standard. Each level states what it claims, what it does not, and where liability sits. A case earns its level from its own artifacts; a level is never granted.
| Level | Claims | Does not claim | Liability |
|---|---|---|---|
| L1 Document received | The document was received in session and identified by type; its SHA-256 is recorded in the receipt. Nothing is retained after the run. | Does not claim the document is authentic or complete. | None; nothing is kept. |
| L2 Data Extracted | These values were extracted from the document text by the pipeline, each carrying a source page and quote. | Does not claim the values are correct or the document truthful. | Extraction quality only. |
| L3 Cross-Referenced | The extracted data is consistent across the documents provided and passes the internal recompute / reconciliation checks that ran. | Does not claim the underlying business reality matches the documents. | Cross-reference accuracy. |
| L4 Checks cleared | No rule-based duplicate-pledge, fabrication or arithmetic check fired on the data supplied; each check that ran is named, and each that could not run is named with its reason. | Does not claim the absence of sophisticated fabrication that the data patterns may not surface. | Methodology, not outcomes. |
| L5 UMFR Verified | The full pipeline ran and every check passed or its warnings were acknowledged; the documents meet UMFR's verification standards under the stated methodology. | Does not claim investment suitability and does not constitute advice. | See Disclaimer. |