The DEHY Credit Health Model

01Introduction

DEHY structures the public record around an issuer — insider transactions, institutional ownership, activist positions, material events, and a decade of financial statements. The Credit Health Model extends that 360° view to the right-hand side of the balance sheet: a single, interpretable read on whether a company can comfortably carry and refinance its obligations.

The design goal is transparency. The score is computed from disclosed financials and filings with published thresholds; there is no opaque model and no machine-learning black box. An analyst can reconstruct any score by hand from the same inputs we show alongside it.

02Philosophy

• Triage, not rating — The score ranks where to look first and flags deterioration. It is not a substitute for a credit rating or a covenant-level credit analysis.
• Absolute, not relative — Thresholds are anchored to S&P/Moody’s cash-flow-to-debt and leverage norms, so a score of 80 means the same thing in any sector and any year — unlike a purely cross-sectional rank that drifts with the universe.
• Graceful degradation — Each component is computed only where its inputs exist; the headline is the weighted average over whatever is available, with thin-data cases flagged.
• Cash flow first — The largest weight sits on cash-flow-to-debt — the closest public-data analog to the funds-from-operations-to-debt ratio that anchors agency methodology.

03Model Architecture

The headline is a weighted blend of five sub-scores, each mapped from a financial ratio to a 0–100 value through published piecewise bands, plus an event overlay that caps or penalizes the score for adverse filing events.

The event overlay reads recent filings: an 8-K reporting bankruptcy (Item 1.03) caps the score near the floor; a debt-acceleration or triggering event (Item 2.04) or a listing-deficiency notice (Item 3.01) applies a penalty and a cap; negative shareholders’ equity caps the score; and recent insider purchases add a small solvency-confidence nudge.

04Data and Analytics

• Financials — As-reported annual statements from SEC XBRL company facts, point-in-time, with a reporting-availability lag so no figure is used before it was public.
• Events — Structured 8-K items (1.03 / 2.04 / 3.01) and Form 4 insider purchases, matched to the issuer.
• Computation — Scores are derived at read time, never stored, so they always reflect the latest as-reported figures and the most recent filings.

05Validation

Without any tuning to fit, the model orders a cross-section of well-known issuers the way a credit desk would — cash-rich technology names at the top, levered media and autos in the middle, and negative-equity or cyclically-distressed names at the bottom.

06Strengths and Limitations

• Transparent — Every score is reproducible by hand from the disclosed inputs shown beside it.
• Absolute and comparable — Rating-grounded bands make scores comparable across sectors and over time.
• Limitation — coverage — Some issuers (notably banks and auto-finance arms) report debt under XBRL tags that thin out the cash-flow-to-debt component; those cases fall back to balance-sheet leverage and are flagged.
• Limitation — annual cadence — The fundamental base updates on the 10-K/10-Q cycle. Intra-period credit deterioration is captured by the event overlay, not the statements.
• Not a rating — The model has no access to covenants, indentures, ratings, or market-implied default risk. It is a screening read, not a credit opinion.