Anatomy of Collapse: Failed Companies Against Their Survivors

OVERVIEW

The revision in brief

Coverage separated the failures years early; revenue did not.

What changed in this revision

The August version described thirty companies that all failed. This version pairs 27 of them with an industry peer that survived, pulls both sides from the same SEC filings, and asks which signals separated the two groups and how early.

The main finding

Interest coverage did. Failed firms could not cover interest from earnings three years before filing (median -0.8x against 4.0x for survivors), and coverage below 1x flagged 83% of them at that point. Revenue decline, the most quoted warning sign, flagged survivors just as often.

Timing

Coverage led by a median of 6 quarters. Operating losses followed. The current ratio gave almost no warning until the final quarter, when debt reclassification took the failed median from 1.01x to 0.33x.

False alarms

Every single rule flags a quarter to a third of healthy peers. Requiring three of four signals a year out cut false alarms to 8% while still catching 39% of failures.

01 | THE COHORT

Thirty large failures, twenty-seven with a survivor

Same industry, same quarters: the survivor is the control.

What the first version could not say

The 25 August version of this page tracked eight ratios across thirty companies that all failed. It could describe how they deteriorated; it could not say whether a healthy company would have looked different. That needs a comparison group, which this revision adds.

How the survivors were chosen

Each of the 27 non-financial failures is paired with one U.S.-listed company in the same industry that did not file for bankruptcy through September 2026: PG&E with Edison International, Chesapeake with Southwestern, J. C. Penney with Kohl's, Sears with Macy's, Frontier with Lumen, and so on. Pairs are matched on industry, not on size.

Same clock for both

Every survivor is measured over the same twelve calendar quarters as its failed twin, ending the quarter before the filing. Both sides come from the same SEC XBRL filings, pulled with the same code.

02 | LIQUIDITY

The current ratio breaks late, then completely

Failed firms fell to 0.33x in the last quarter; survivors held 1.55x.

Three years out, close together

Twelve quarters before filing the failed group's median current ratio was 1.45x against 1.63x for survivors. The gap widened slowly: 1.12x against 1.41x one year out.

Then the cliff

In the final reported quarter the failed median dropped to 0.33x while survivors stayed at 1.55x. The mechanism is accounting, not operations: once a covenant is breached or a maturity is inside twelve months, long-term debt is reclassified as current and the ratio collapses in one filing.

Why the first version showed 0.90x

The August page reported a T-1 median of 0.90x from a dataset with many final-quarter blanks. The new pull fills those in; most of the added values are the reclassification quarter.

03 | REVENUE

Falling sales did not separate failures from survivors

Revenue is the weakest of the four signals, and the noisiest.

Both groups shrank

In the last quarter before filing, failed firms' revenue was down 16% year over year at the median. Survivors were down 8%. Twelve quarters out the two groups were within four points of each other.

The same signal fires for the healthy

A 10% revenue decline flagged 38% of eventual failures three years out, and exactly the same share, 38%, of survivors. One year out the split was 44% against 23%.

Why

Most of these failures came in industry downturns: coal in 2015-16, oil and offshore drilling in 2016 and 2020, department stores through the late 2010s. The survivor sold less too. What differed was how much debt sat underneath the sales.

04 | MARGIN

Operating losses arrived early and stayed

Survivors kept a positive margin until the pandemic quarter.

A persistent gap

The failed group's median operating margin was negative in eight of the twelve quarters and never above 3.2%; survivors' median stayed between 3% and 10% until the final quarter. One year out the gap was -1% against 4%.

The final quarter

At T-1 the failed median fell to -40%, driven by impairments booked as the filing approached. Survivors' median was -8%, almost entirely the 2020 pairs.

As a flag

A negative operating margin caught 56% of failures a year out and 28% of survivors: useful, but one in four false alarms.

05 | COVERAGE

Interest coverage was the earliest and cleanest signal

Failed firms could not cover interest three years before filing.

Below one for the whole window

The failed group's median coverage was -0.8x twelve quarters before filing and never rose above 0.3x. Survivors' median was 4.0x at the start and 1.6x a year out.

Hit rate and false alarms

Coverage below 1x flagged 83% of eventual failures three years out and 32% of survivors. In the final quarter it flagged every failure and 60% of survivors, which is the pandemic again.

How early

For 22 of the 23 failed firms with data, coverage went below 1x and stayed there; the median lead was 6 quarters, and 16 firms were flagged a year or more ahead.

06 | SCORECARD

Stacking the signals cuts the false alarms

Three signals a year out: 39% of failures, 8% of survivors.

One year out

Coverage alone: 81% hit rate, 35% false alarms. Requiring three of the four signals: 39% hit rate, 8% false alarms. The combination trades some sensitivity for a much cleaner list.

The final quarter

By T-1 70% of failures showed three or more signals, but so did 28% of survivors. Late signals are loud and unspecific; the value is in the year before.

What this is not

Twenty-seven pairs is a case set, not a model. The rates above describe these companies; they are not probabilities for the next one.

07 | FOUR WAYS TO FAIL

Same signals, different mechanisms

The ratios describe the failure; the cases explain it.

Lehman Brothers: leverage that looked improving

Net leverage peaked near 16x in early 2008 and fell to about 10x by the second quarter through asset sales and capital raises. The reported ratio improved for two straight quarters; the firm failed in September when short-term funding and counterparty confidence were withdrawn.

Silicon Valley Bank: growth outrunning duration risk

Total assets more than tripled from under $60 billion at end-2019 to $209 billion at end-2022, funded largely by uninsured deposits. Rising rates produced unrealized securities losses that became a solvency question once depositors left.

WorldCom: the metric itself was the fiction

On 25 June 2002 WorldCom disclosed $3.8 billion of improperly capitalized line costs; by August the confirmed total exceeded $11 billion. The reported numbers were not off by a margin of error; they were the wrong numbers on purpose.

PG&E: a liability larger than the balance sheet

PG&E filed in January 2019 citing up to $30 billion of wildfire liability against $71.4 billion of reported assets. In our panel its interest coverage turned negative and its current ratio fell to 0.22 in the final quarter as debt was reclassified as current.

Case figures as verified for the 25 Aug 2026 version; PG&E panel values from this revision.

08 | WHAT THIS SUPPORTS

What the matched test changes

Coverage first, margin second, revenue last.

For a lender or credit analyst

Watch coverage first and treat a reading below 1x that persists for two quarters as the signal, whatever revenue is doing. A revenue decline in a sector downturn tells you about the sector.

For an equity investor

The final-quarter numbers are not the warning; they are the confirmation. The information was in the interest line one to three years earlier.

For a board

Debt reclassified as current is the last step, not the first. By the time the current ratio collapses, the refinancing window has usually closed.

What would change the picture

A larger cohort, matching on size as well as industry, and adding non-bankrupt distressed exchanges as a third group. The code and pairs are published with this article so the test can be extended.

Limits

12 of 27 pairs end in the first half of 2020, so the final-quarter survivor figures carry the pandemic. Three financial issuers are excluded. The SEC panel starts in 2009, so the three oldest cases have partial histories. Nothing is estimated; blanks stay blank.

Revisions to the earlier page

Current ratio at T-1: 0.33x (was 0.90x, from a sparser final quarter). Revenue at T-1: -16% (was -20%). Operating margin at T-1: -40% (was -7%). The differences come from more complete final-quarter data and a non-financial cohort, not from a change in method.

Conditional implications, not forecasts.