How a Top SMB Lender Improved Risk Model Accuracy by 25%

The Opportunity: Growth Without Added Risk

A top 10 small business lender wanted to grow revenue without increasing loss ratios. In lending, this is the fundamental tension: every dollar of additional credit extended creates both revenue potential and default risk. The challenge is separating the two.

Safely increasing credit lines was a core part of their growth strategy. The logic was straightforward: existing borrowers who are thriving represent the lowest-risk, highest-return opportunity for additional lending. They already have a relationship, a payment history, and a demonstrated need for credit.

But without clear visibility into their portfolio’s health, the lender was missing opportunities and taking on unnecessary risk simultaneously. Some borrowers who deserved higher credit lines weren’t getting them. Others who were deteriorating were maintaining limits they shouldn’t have.

The Data Gap: What Traditional Signals Miss

Traditional credit risk models rely on bureau data, payment history, and financial statements. These signals are valuable but backward-looking and often delayed. A business that started declining three months ago might not show distress in bureau data for another six months.

The lender needed forward-looking signals that could distinguish between businesses on different trajectories:

• Is this business still actively transacting? Transaction presence confirms ongoing operations.
• Are transactions stable or declining? Stability patterns predict future repayment capacity.
• Is the business growing? Growth signals indicate safe credit expansion opportunities.

These three questions — presence, stability, and growth — required transaction-level data that traditional credit inputs couldn’t provide.

The Solution: Transaction Signals in Risk Models

The lender integrated Enigma’s transaction data into their risk models, focusing on three key attributes:

Presence of Transactions

The most basic signal is the most powerful for loss prevention: is this business still processing card transactions? A business that has stopped transacting is a fundamentally different risk profile than one with active revenue, regardless of what bureau data shows.

Transaction Stability

Beyond presence, stability patterns matter. A business with consistent monthly transaction volumes presents predictable repayment capacity. One with erratic or declining volumes presents increasing risk. These patterns emerge months before traditional credit distress signals.

Growth Rates

For credit line increases specifically, growth rates are the key signal. A business growing at 15% year-over-year can safely absorb a higher credit line. That same credit line increase for a flat or declining business represents disproportionate risk.

The Results: 25% Better Accuracy, $30M in Revenue

The results validated the thesis: transaction data provides signals that traditional credit inputs miss.

25% Model Accuracy Improvement

By incorporating transaction presence, stability, and growth signals, the lender’s risk models became 25% more accurate for the target population. Better accuracy means fewer false positives (denying good borrowers) and fewer false negatives (extending credit to deteriorating ones).

70,000 New High-Value Businesses

The improved models identified 70,000 businesses across the portfolio that qualified for credit line increases — opportunities that were previously invisible because traditional credit signals didn’t capture their growth trajectory.

$30M in Incremental Revenue

These safer credit line increases generated $30M in incremental revenue. The key word is “incremental” — this revenue was additive to existing lending activity, generated by acting on opportunities the previous models couldn’t identify.

Loss Ratios Maintained

The $30M wasn’t purchased with higher risk. Loss ratios held steady because the credit line increases were targeted at businesses with verifiable growth patterns, not applied broadly across the portfolio.

Key Takeaways

1. Transaction data is forward-looking. Bureau data tells you what happened. Transaction data tells you what’s happening now and what’s likely to happen next. For credit decisions, the present tense matters most.

2. Growth signals unlock safe credit expansion. Identifying businesses with rising transaction volumes creates a natural alignment between credit supply and business demand.

3. Portfolio optimization beats new acquisition. The 70,000 high-value businesses were already customers. The lender didn’t need to find them — they needed the data to see them clearly.

Technical Specifications