Live
Twin Watch
Results How it works Limits Request a pilot
Why Twin Watch?

Spot the companies that look like the ones that failed.

Twin Watch compares a company's reported financial history with thousands of past filings from companies that later went bankrupt and companies that did not. It flags the ones whose history most resembles the failures.

Start a 90-day pilot See the test results
Tested on real SEC filings
65 of 87
failed companies that Twin Watch could score were flagged before they filed. About 75%, with a 95% range of 65% to 83%.
About 11%
of healthy company quarters were flagged by mistake in that test.
Why Twin Watch?
It learns from real failures
Twin Watch holds 9,798 past filings from 641 companies, including the ones that went bankrupt. It does not rely on one ratio or one rule of thumb.
It finds the twins
For each company it looks for past companies whose reported history looked the same, then asks how many of those twins later failed. More failed twins means a stronger warning.
It beats the standard score
On a 2025 to 2026 test, it caught 33 of 37 failures against 13 of 37 for the Altman Z-score at the same false-alarm rate.
It is tested honestly
Every test rule was fixed and fingerprinted before it ran, and 3,000 companies are being tracked forward to confirm the results in April 2027.
Results at a glance
33 of 37
vs 13 of 37 for the Altman Z-score at the same false-alarm rate, on a 2025 to 2026 test
77%
chance a failing company is ranked riskier than a healthy one
Fixed in advance
every test rule was recorded and fingerprinted before it was run
Tested on real SEC filings. Being confirmed forward on 3,000 companies reporting April 2027.
Reference
Twin library
9,798
past snapshots · 641 companies · model v4
Forward test
3,000
companies frozen 2026-11-16 · reports April 2027
Time now
live · Eastern
Effective fed funds rate
3.88%
Federal Reserve · 2026-11-12
10-year Treasury yield
5.10%
Federal Reserve · 2026-11-12
Bank prime loan rate
7.00%
Federal Reserve · 2026-11-09
Effective federal funds rate. The effective rate on overnight federal-funds transactions. It is a widely used benchmark for very short-term U.S. interest rates.
10-year Treasury yield. The market yield on 10-year U.S. Treasury securities. It is a common benchmark for longer-term borrowing and valuation.
Bank prime loan rate. The prime rate posted by a majority of the largest U.S. commercial banks. It is a common base rate used to price short-term business loans.
Rates are for reference only and are not inputs to Twin Watch.

Twin Watch beat the Altman Z-score

Twin Watch caught 33 of 37 failures; the Altman Z-score, set to the same false-alarm rate, caught 13. The Altman Z-score is the long-standing distress measure. On a 2025 to 2026 test, both were run on the same companies and the same quarters: 37 that failed and 62 healthy. To compare fairly, Altman was set to the same 12% false-alarm rate as Twin Watch.

Method Failures flagged Healthy companies flagged by mistake
Twin Watch 33 of 37 12%
Altman Z-score, set to the same 12% 13 of 37 12%

At the same false-alarm rate, Twin Watch flagged 21 failures that Altman missed, and Altman flagged 1 that Twin Watch missed. This is the current model, version 4.

Alert levels
WATCH
About 1 in 5 of the closest twins went on to fail.
ALERT
More than 1 in 3 of the closest twins went on to fail.
HIGH
About 6 in 10 or more of the closest twins went on to fail.
Each level also has its own word and outline weight, so it can be told apart without color. Reddish shades are used only for these three alerts.

How it works

1

Read the filings

It takes a company's last eight quarters of public SEC filings: sales and cash-flow trends, balance-sheet strength, cash cover for debt coming due, and signs of how the company is reporting.

2

Find its twins

It searches a library of about 9,800 past company snapshots for the ones that looked most alike at the same stage, and checks how those turned out within two years.

3

Raise a flag

If enough close twins went on to fail, the company is flagged as WATCH, ALERT or HIGH. A flag only appears when there are enough good matches to trust it.

What a report shows you
Twin Watch reads a company's recent SEC filings, finds its closest financial twins, and shows what happened to them. When a much larger share of the twins went on to file for bankruptcy than is normal, it raises an early warning and names the twins, so you can see exactly why. It is a warning signal, not a prediction.
HIGH: higher risk than 92% of the companies we watch
7 close historical twins found; 5 of them filed within two years. Example report, fictional company.
Closest twins
Twin A, a regional manufacturer in 2014: 81% match. Filed 6 quarters later.
Twin B, a specialty retailer in 2016: 74% match. Filed 4 quarters later.
Twin C, an equipment supplier in 2018: 69% match. Cut debt and stayed healthy.
What is driving it
Debt is high for this kind of business.
Cash and cash flow cover less than a year of debt coming due.
Operating-risk warning: ELEVATED. Revenue falling while operations lose money.
A going-concern warning appeared in a recent filing.
What would lower the risk
Twins that survived reduced debt and kept cash covering at least a year of obligations.
Every report names the real historical twins behind the score, so you can check the reasoning yourself.
Who it is for
Lenders
Spot borrowers drifting toward default while there is still time to change terms or exposure.
Supply chains
See when a key supplier looks like companies that collapsed, and line up a backup first.
Receivables
Check that large customers will still be able to pay before extending more credit.
Investors and buyers
Find out in due diligence whether a company is weaker than it looks.
Help it learn
Report outcomes
Tell us how a company you checked turned out. Each outcome becomes a new twin.
Report misses
A company failed and Twin Watch did not flag it? Those cases show exactly where it is blind.
Disagree
Think a report is wrong? Tell us why. If you are right, that reason becomes something it learns to see.
How it was tested
Every test followed the same discipline: the rules were written down and the model files fingerprinted before each test ran, each test ran once, and the first result is the official one.
Internal tests by Recombination Finance on SEC filings, not an independent validation. False alarms are the share of quarters flagged for companies that did not file within two years.
Compared with the standard Altman Z-score, this comparison was added after the original test, and matching the false-alarm rate favors Altman. The 2025 to 2026 set is mostly small, loss-making companies, where Altman is known to be less reliable, so this does not show Twin Watch beats Altman everywhere.
A record kept in the open
Past tests only go so far. A list of 3,000 companies was frozen on 2026-11-16 and scored with a digital fingerprint, so the scores cannot be changed later. As the months pass, those saved predictions are checked against which companies actually filed. The first forward scorecard is due in April 2027. Each new outcome also becomes a new twin, so the library keeps learning.
Pilot access
Twin Watch is part of the free Nexus pilot, 90 days by default. Recombination Finance sets the end date for each pilot and may change or end it at any time. Pilot companies can request reports on the companies they care about and tell us how those companies turned out. Reports are emailed, usually within a few hours and always within two business days.

What to know before you rely on it

It does not score every company

Twin Watch needs enough filing history. In testing it could score about a third of companies overall, and 58% of the failures since 2018. Companies with short histories are not scored. Public companies are scored from SEC filings; private companies can be scored from figures you provide.

Tested, and being confirmed

The rules were fixed before each test was run, on real filings. A forward test on a frozen list of 3,000 companies reports in April 2027.

Older failures are harder

For failures since 2018 it flagged 60 of 74 (81%). For earlier failures it flagged 5 of 13, a small sample.

A flag is a warning, not a verdict

Most flagged companies do not fail, and about 12% of calm quarters were flagged for companies that did not go on to file. The score ranks risk. It is not an exact probability of bankruptcy. Treat a flag as a reason to look closer. This is not investment or credit advice.

Try us out. What do you have to lose?

Run a 90-day pilot on your own list. Upload a CSV of company names and get back which ones can be scored and how each is rated.

Request a pilot
© 2026 Recombination Finance. All rights reserved. Twin Watch by Recombination Finance, Nyack, NY Contact: recombinationfinance@proton.me The method Rule audit Terms Privacy Twin Watch is developed by James R. Oursler, Recombination Finance, Nyack, NY. Risk indicators for evaluation only; not investment, credit or legal advice.