Assay

Methodology [top]

The grade is a rank of reported financial condition among sector peers, with one market-informed estimate of failure risk among its inputs. It is not a return forecast.

How the Rank Is Made

1. Fourteen inputs in three sleeves. Profitability: what the business earns on what it owns. Solvency: whether it can carry what it owes. Growth and financing: whether it grows without leaning on new shares or debt. The table below gives every formula.
2. Percentiles within the sector. Each input becomes a percentile among the graded companies of the company's sector. Lower is stronger everywhere on the site. A sector needs at least 100 graded companies.
3. Equal weights. A sleeve score is the mean of its inputs' percentiles, and the composite is the mean of the three sleeves.
4. Letters are fifths. A is the strongest fifth of the sector, E the weakest. A grade needs at least 7 of the 14 inputs and a score in every sleeve.
5. Boundary letters. A percentile has sampling error because the peer group is finite. When it lies within one standard error of a band edge, the page shows both letters, as in B/C.

The Fourteen Inputs

Point in Time

Only facts the SEC had received by the rescan date enter that rescan. The cutoff is the receipt date, not the period the numbers describe. A comparative figure that arrives up to 400 days after its period is accepted. Missing values stay missing; nothing is filled in. Across the current universe a filing arrives 57 days after its period end at the median and 90 days at the 90th percentile.

Failure Model and Reliability

One input uses the market: the twelve-month failure probability from Campbell, Hilscher, and Szilagyi (2008), a logistic model of accounting ratios and market equity. It was fitted on 1963-2003, and fitted pre-2009 and not presented as a current calibration. The figure is a conditional month-12 failure probability, not cumulative. Its inputs are trimmed at the pooled 5th and 95th percentiles of the run. It resolves for 486 of 3,545 eligible filers, and the distress warning marks the top tenth among them.

Reliability is measured against one outcome: 8-K Item 1.03 within 365 days, with the directional acquisition rule. The model is 11 accounting level inputs plus loss and burn thresholds, fit separately within asset quintile, L2 penalty selected per stratum by 5-fold CIK-grouped cross-validation on the training window, fitted on filings 2012-01-01 to 2019-12-31 and tested on filings 2020-01-01 to 2025-12-31. AUC is the chance that a company that went on to fail ranks above one that did not; 0.5 is a coin toss and 1 is perfect. Status: post hoc pending confirmation; pre-registered on 2026 filings once their 12-month outcome windows close; supported if the paired per-quintile gain over the pooled model has a lower bound above 0.05.

Standard universe: revenue above $1M and total assets above $10M; the screen every published distress figure was measured inside. Inside: AUC 0.889, 617 events. Outside: AUC 0.745, 74 events. Failure rates are the annualized 12-month Item 1.03 frequency, filings 2012-01-01 to 2025-12-31.

Warnings

Peer Groups

A company's sector comes from the four-digit SIC code on its SEC record. Banks, insurers, brokers and REITs, SIC 6000 to 6999, are outside the universe because their statements do not carry the inputs the grade needs.

Sources

Accounting facts from the SEC EDGAR bulk XBRL archives, refreshed nightly. Prices from Alpaca (sip feed); the last close is the iex feed. S&P 500 market value $67.1T as of Jul 31, 2026, from S&P Dow Jones Indices, used for relative size in the failure model. No market price enters the letter.

Diagnostics That Do Not Feed the Grade

The Piotroski F-score, cash runway, filing gap, late-filer status and median dollar volume appear on every company page and never enter the letter. Price measures appear as measurements, with no verdict.