18% lift in approval rate for credit-invisible applicants
A regional CDFI used AgentHive to expand credit access to consumers with fewer than 3 trade-lines — without relaxing risk thresholds.
Context
A mission-driven CDFI serving 400,000+ members lacked the data depth to make confident decisions on credit-invisible applicants. Traditional bureau scores returned either no-hit or shallow files for nearly 30% of their pipeline.
What we built
AgentHive's underwriting module was configured with an alternate-data feature set: cash-flow volatility from bank-feed data, rent-payment history (via Plaid), and utility trade-lines. The SHAP explainability layer surfaced which signals drove each decision — satisfying FCRA adverse-action disclosure requirements automatically.
Results
| Metric | Before | After |
|---|---|---|
| Approval rate (thin-file) | 41% | 59% |
| Average time-to-decision | 3.1 days | 12 seconds |
| Model-governance findings | — | 0 |
| Default rate (6-month) | 4.2% | 4.0% |
"For the first time, we could approve a qualified applicant without a bureau score and hand them a printed adverse-action notice instantly — all compliant, all explainable." — Chief Credit Officer
See it live in the Interactive Lab
Run the same decision scenarios yourself — no setup required.
Open the Lab