How to Build a Distress Screener

Turning default-risk factors into a ranked watchlist.

A distress screener is not a default-prediction model. It is a triage tool — a systematic way to surface names that warrant closer fundamental work before a credit event forces your hand. The inputs are available from public filings, market feeds, and structured event databases like the corporate default events dataset. The challenge is combining them without introducing look-ahead bias, weighting them against each other in a principled way, and keeping the output narrow enough to be actionable. This post walks through a four-component design that mirrors the rigor applied to dilution screens and short-sale research.

Component one: solvency and leverage

The first component asks whether a company's capital structure is sustainable under realistic operating scenarios. Three inputs anchor it:

• Altman Z-score. The original 1968 formulation uses five accounting ratios — working capital, retained earnings, EBIT, book equity, and sales, each scaled by total assets or liabilities — to produce a composite score. Below 1.81 falls in the distress zone for manufacturers; the revised Z''-score applies to non-manufacturers and emerging-market issuers. Use it as a rough filter, not a hard cutoff.
• Net debt / EBITDA. The most common leverage screen in credit analysis. Sector context matters — a 4× ratio is unremarkable for a cable company and alarming for a specialty retailer. Build sector-relative percentile ranks rather than applying absolute thresholds.
• Interest coverage (EBIT / interest expense). Coverage below 1.5× means the company is generating less operating income than its interest bill consumes. Below 1.0× is a hard warning. Use trailing twelve-month figures and re-check against the most recent quarter for trend direction.

Pull these from point-in-time financial statement data — use the "as-reported" figures from each period's filing, not retroactively restated values, to avoid look-ahead contamination.

Component two: liquidity and runway

Solvency tells you whether the balance sheet is underwater; liquidity tells you how long the company can stay afloat before that question becomes urgent. Four inputs matter:

• Unrestricted cash and equivalents. The headline number from the balance sheet, minus any cash that is pledged or restricted by covenant.
• Levered free cash flow. Operating cash flow minus capex minus mandatory debt service. Negative and deteriorating FCF is more actionable than a single bad quarter.
• Near-term debt maturities. Aggregate the face value of debt maturing within twelve months. Cross-reference against available revolving credit capacity. A company with a $200 million maturity and a fully drawn revolver is in a different position than one with substantial undrawn capacity.
• Cash runway. Divide cash by the trailing-twelve-month cash burn rate. For pre-profit issuers this is the primary liquidity metric; for profitable issuers it remains a backstop measure.

Component three: disclosure flags

Accounting disclosures and filing behavior contain qualitative signals that quantitative ratios will not capture until it is too late. These flags function as binary inputs — present or absent — and any single flag should immediately move a name into active review:

• Going-concern opinion. An auditor's going-concern qualification in the 10-K or 10-Q is the single highest-signal flag. It means the auditors themselves doubt the company's ability to continue operating for twelve months.
• Covenant breach or waiver disclosure. Disclosed in an 8-K or in the notes to financial statements. A waiver is not relief — it often comes with tighter terms and shorter timelines.
• Late filing (NT 10-K, NT 10-Q). A notification of inability to file on time suggests accounting problems, auditor disagreements, or management turnover.
• Auditor change. A sudden auditor rotation, particularly to a smaller firm, warrants scrutiny. Parse the 8-K/A filings that disclose auditor disagreements specifically.

These flags are documented in structured form by SEC filing type and are straightforward to monitor systematically. The harder part is parsing the unstructured text of the filings to detect going-concern language and covenant breach disclosures at scale.

Component four: market structure

For companies with publicly traded debt or active credit default swap markets, market prices encode information not yet visible in lagging accounting data. Three inputs are relevant:

• Bond yield and CDS spread. A spread above 1,000 basis points over Treasuries conventionally defines distressed debt. Watch for spread widening in excess of 200–300 basis points over a short window — that is the market repricing default probability, not interest-rate duration.
• Equity drawdown from 52-week high. A drawdown of 60–80% is common in the run-up to a credit event. Combine with volume — a collapse on elevated volume signals forced selling rather than routine drift.
• Short interest and borrow cost. Rising borrow rates indicate that short sellers are aggressively building positions. This is a useful confirming signal; it is not a primary input because it can reflect other theses including dilution or M&A rumors.

For names without public debt, equity market structure becomes the primary market-based input. This is where integration with finding companies at risk of default and finding stocks to short sell using data becomes directly applicable.

Inputs, sources, and a sample schema

Input	Component	Primary source	Frequency
Altman Z-score	Solvency	10-K / 10-Q (as-filed)	Quarterly
Net debt / EBITDA	Solvency	10-K / 10-Q (as-filed)	Quarterly
Interest coverage	Solvency	10-K / 10-Q (as-filed)	Quarterly
Cash runway (months)	Liquidity	10-Q cash flow statement	Quarterly
Debt maturing <12 months	Liquidity	10-K notes / debt schedules	Annual + update
Going-concern opinion	Disclosure	Auditor report (10-K/10-Q)	Per filing
Covenant breach / waiver	Disclosure	8-K, credit agreement notes	Event-driven
Late filing (NT form)	Disclosure	EDGAR NT 10-K / NT 10-Q	Event-driven
CDS spread / bond yield	Market structure	TRACE, Bloomberg, Markit	Daily
Equity drawdown	Market structure	Exchange feed	Daily
Borrow rate	Market structure	Prime broker / securities lending feed	Daily

Scoring, filtering, and avoiding look-ahead

A two-stage approach keeps the output manageable. In stage one, apply hard filters to eliminate inapplicable names: drop financial institutions from Z-score ranking (the model was not designed for them), drop names with no public debt from CDS-based inputs, and drop any name with fewer than four quarters of history. What remains is a survivable universe.

In stage two, score survivors on each component. For continuous inputs — leverage ratios, coverage, drawdown — compute cross-sectional percentile ranks within sector or convert to z-scores. For binary disclosure flags, assign a fixed additive penalty (e.g., going-concern adds 20 percentile points to the raw score). Weight the four components; a reasonable starting point is 30% solvency, 30% liquidity, 25% disclosure, 15% market structure, but calibrate against labeled historical default outcomes from your corporate default events dataset. A simple logistic regression on historical defaults will tell you which inputs actually predicted events in your universe — let the data override the prior.

Look-ahead is the most common source of false validation in distress research. Use point-in-time snapshots: the 10-Q filed on date X should only inform the screener score on date X, not retroactively. Restated financials should be excluded from backtests entirely. Market data is naturally point-in-time; filing data requires discipline to handle correctly.

Output: a triage watchlist, not a trade signal

The screener's output is a ranked list — typically the top 30 to 50 names by composite score — reviewed on a defined cadence, usually weekly. Each name in the watchlist should trigger a specific next step: pull the most recent earnings call transcript for liquidity commentary, read the credit agreement for covenant headroom, or check whether the bond has traded recently and at what spread. The screener tells you where to look. It does not tell you whether to trade.

This distinction matters for calibration. A screener that generates 50 names per week, of which three default over the following twelve months, is performing well if those three appeared consistently at the top of the list in the months before the event. Precision at the top of the ranking is the metric to optimize — not overall classification accuracy across a heavily imbalanced universe where defaults are rare by design.