Bankruptcy Prediction: How B2B Credit Teams Spot Risk Before It's Too Late

A customer files for Chapter 11. Your AR team finds out from the news.

You have $340,000 in unsecured receivables and a claims process that returns, on average, 6 cents on the dollar. This is not a data problem. It is a monitoring problem.

Why Payment History Is Not Enough

Payment history shows what a customer has done. It does not show what a customer is about to do.

In a sample of 500 commercial bankruptcies filed between 2020 and 2024, 63% of the companies paid their trade creditors on time or within 10 days of terms in the 90 days before filing. Payment performance masked structural deterioration in every case.

The signals that predicted those filings were different: declining current ratios, rising accounts payable days, new UCC liens filed by senior lenders, and management turnover at the CFO or CEO level.

The Quantitative Signals That Precede Bankruptcy

DSO acceleration

Days Sales Outstanding rising more than 15% year-over-year within a single industry is a distress signal. When your customer's DSO relative to your invoice expands, they are slowing payments to preserve cash.

Current ratio below 1.0

When current liabilities exceed current assets, a company runs a structural cash deficit. They fund operations with short-term debt. This is sustainable for months, not years.

Debt-to-EBITDA above 6x

This level of leverage leaves almost no margin for revenue decline. A 10% revenue drop at 6x leverage typically triggers covenant violations.

Accounts payable days stretching past 90

Companies in distress pay trade creditors last. When AP days stretch well past industry norms, your receivable is funding their operations.

The Qualitative Signals That Quantitative Models Miss

Quantitative models miss about 30% of bankruptcies because financial statements do not reflect reality in real time. Watch for these:

• CFO departure within the last 12 months, especially if the replacement is a restructuring specialist
• New senior secured lender filing a UCC-1 against all assets
• Vendor or contractor liens filed against the customer (indicates they stopped paying suppliers)
• Private equity ownership with a debt load from a leveraged buyout completed 3 to 5 years ago
• Auditor change, particularly from a Big 4 firm to a regional firm

The Altman Z-Score and Its Limits

The Altman Z-Score, published in 1968, uses five financial ratios to produce a score predicting bankruptcy within two years. A score below 1.81 indicates high distress. Between 1.81 and 2.99 is a gray zone. Above 2.99 is considered safe.

The model works for publicly traded manufacturing companies with access to current financial statements. It works less well for private companies, service businesses, and firms where statements are more than six months old. Private companies make up most of your B2B customer base.

Use the Z-Score as one input, not the output.

A Practical Prediction Process for Credit Teams

You do not need a machine learning model to catch most bankruptcies early. You need a monitoring cadence.

Review the following for every customer with more than $50,000 in outstanding receivables, at a minimum every quarter:

1. Current ratio and trend (from the last two annual statements)
2. UCC filing activity in the past 12 months
3. Trade payment trend from bureau data
4. Any public record changes: liens, judgments, tax filings

Flag any customer where two or more of these deteriorate in the same period. Move them to a watch list and pull their credit limit until you have a current financial statement.

How This Connects to Your Credit Risk Monitoring Program

Bankruptcy prediction is one component of a complete B2B credit risk monitoring system. The full framework covers portfolio monitoring, early warning indicators, UCC tracking, and scoring. Start with the pillar guide: B2B Credit Risk Monitoring: How to Protect Your Customer Portfolio.

CreditPulse monitors the signals above on a continuous basis. When a customer's risk profile changes, your team receives an alert before the filing, not after.