# MCA credit decisioning automated

> Automated MCA credit decisioning uses bank-statement parsing, soft-pull credit data, fraud signals, and machine-learning underwriting models to issue approval decisions within minutes; top funders auto-decision 60–80% of applications without human review.

MCA credit decisioning automated refers to the technology pipelines that merchant cash advance funders use to approve, decline, or price advances without human underwriter intervention. As of 2026-06-28, top fintech funders auto-decision the majority of applications through automated grids combining bank statement analysis, credit data, fraud signals, and risk-based pricing models. Manual underwriting persists for larger deals, complex profiles, and exception cases.

**The decisioning pipeline.**

**Step 1: Application intake (0–60 seconds).**
- Merchant submits application via online form, mobile app, or ISO portal.
- Required data: business name, EIN, contact info, requested amount, owner SSN, business address, time in business, monthly revenue self-reported.
- Plaid or MX bank connection initiated, OR merchant uploads bank statement PDFs.

**Step 2: Bank statement aggregation (1–10 minutes).**
- **Plaid / MX integration.** Last 90–180 days of transaction data pulled via API.
- **PDF extraction.** OCR + ML extraction parses uploaded statements into structured transaction data.
- **Quality checks.** Verify completeness (no missing days), authentic format (anti-fraud), and statement-period coverage.

**Step 3: Bank statement analysis (real-time after aggregation).**
- **Revenue calculation.** Total deposits over the trailing 3 months, normalized for trends.
- **Average daily balance.** ADB computed for cashflow stability assessment.
- **Negative day count.** Number of days with negative ending balance.
- **NSF / overdraft fees.** Count of insufficient-fund events and bank charges.
- **MCA debit detection.** ML model scans for existing MCA daily/weekly debits (Yellowstone, OnDeck, Credibly, etc. signatures).
- **Cash flow consistency.** Variance in daily deposits, seasonal patterns, weekend / holiday adjustments.
- **Deposit composition.** Card revenue vs cash deposits vs transfer deposits vs check deposits.

**Step 4: Credit data pull (5–60 seconds).**
- **Soft credit pull on personal guarantor.** FICO score, recent inquiries, derogatory marks.
- **Business credit pull.** Dun & Bradstreet, Experian Business, Equifax Small Business.
- **Public records.** Tax liens, judgments, bankruptcies, UCC filings.

**Step 5: Fraud signals (real-time).**
- **Identity verification.** SSN / DOB / address consistency checks.
- **Velocity checks.** Multiple applications from same IP / device / SSN within 30 days.
- **Bank account verification.** Account holder name matches business name or owner name.
- **Synthetic identity detection.** Pattern recognition for fabricated identities.
- **MCA shopping signals.** Bank statement reveals multiple recent MCA inquiries.

**Step 6: Risk scoring (real-time after data assembly).**
- Funder's proprietary scoring model combines bank statement metrics, credit data, fraud signals, and industry / geography risk into a single risk score.
- Score maps to:
  - **Auto-approve grid.** Low-risk profiles auto-approved at standard pricing.
  - **Auto-decline grid.** High-risk profiles auto-declined with reason codes.
  - **Manual review queue.** Mid-risk or large-ticket profiles routed to human underwriter.

**Step 7: Pricing (real-time after risk score).**
- **Factor rate.** Risk score maps to factor rate range (1.10–1.50).
- **Term length.** Risk score maps to acceptable term (90–540 days).
- **Advance amount.** Maximum advance computed as a multiple of monthly revenue (typically 0.5x–1.5x), adjusted for risk.
- **Holdback or payment amount.** Daily / weekly debit calculated to amortize the advance over the term.

**Step 8: Contract generation and delivery (1–5 minutes).**
- Automated contract templating populates merchant data, pricing, and terms.
- E-signature delivery (DocuSign, HelloSign, or proprietary).
- Bank verification micro-deposits or Plaid re-verification initiated.

**Step 9: Funding origination (after contract execution).**
- ACH origination request submitted to funder's bank.
- Funds typically settle next business day; some funders use FedNow / RTP for same-day settlement.

**Total elapsed time (clean deal).** 10 minutes to 2 hours from submission to approval; 4–24 hours from submission to funded.

**Auto-decisioning rate by funder.**

- **OnDeck.** 70–80% auto-decisioned for term loans up to $250K; lower for MCA.
- **Credibly.** 60–75% auto-decisioned for advances up to $200K.
- **Forward Financing.** 50–65% auto-decisioned for advances up to $250K.
- **Rapid Finance.** 55–70% auto-decisioned.
- **Fundbox.** 70–80% auto-decisioned for LOC.
- **Mid-market traditional funders.** 30–55% auto-decisioned.
- **Niche / specialty funders.** 10–30% auto-decisioned.

**What triggers manual review.**

1. **Advance amount above auto-approve threshold.** Typically $150K–$250K+.
2. **Risk score in mid-range.** Score that doesn't clearly approve or decline.
3. **Existing MCA debt detected.** Stacking analysis requires senior funder review.
4. **Public record alerts.** Tax liens, judgments, recent bankruptcy.
5. **Industry on caution list.** Cannabis, adult entertainment, gambling, certain CBD products, certain crypto businesses.
6. **High-fraud indicators.** Multiple velocity flags, identity mismatches.
7. **Bank statement irregularities.** Missing pages, format anomalies, unverifiable banking institution.
8. **Renewals with revenue decline.** Existing merchant requesting renewal with deteriorated metrics.
9. **First-time merchant for funder.** Some funders require manual review of first transactions with new merchants.

**Underwriting models.**

**Traditional rule-based.** Hand-coded decisioning logic (e.g., "if FICO < 550 and monthly revenue < $20K, decline"). Predictable, auditable, but inflexible.

**Logistic regression / GLM.** Statistical models predicting default probability from input features. Standard industry baseline.

**Gradient boosted trees (XGBoost, LightGBM).** Higher-accuracy ML models for default prediction. Adopted by top fintech funders.

**Deep learning / neural networks.** Selective adoption; complexity often not justified by marginal accuracy gain.

**Generative AI / LLM-assisted decisioning.** Emerging; used for bank statement narrative analysis, fraud-signal reasoning, exception-case classification. Not yet primary decisioning engine at scale.

**Model governance.**

**Fair lending / ECOA compliance.** CFPB ECOA enforcement creates risk for ML underwriting models that produce disparate impact. Funders must conduct disparate-impact analyses on protected characteristics (race, gender, age, ethnicity).

**Adverse action notices.** When automated decisioning declines an applicant, ECOA requires specific reason codes (e.g., "insufficient income," "too many recent credit inquiries"). Reason-code mapping must align with model decision logic.

**Model documentation.** Funders must document model design, training data, validation methodology, ongoing performance monitoring.

**Bias testing.** Models must be tested for proxy discrimination (e.g., ZIP code as race proxy).

**Common confusion.** First, "automated decisioning means no humans involved" — most funders have human override authority and manual review for exceptions. Second, "automated decisioning is more accurate" — automation enables speed and consistency, not necessarily better accuracy on edge cases. Third, "automated decisions cannot be appealed" — most funders offer a manual review path for declined merchants who request reconsideration with additional documentation.

## Related terms

- [MCA bank statement analysis](https://fundnode.co/llms/glossary/mca-bank-statement-analysis) — The underwriting process where funders parse 3-6 months of business bank statements for average daily balance, deposit count, NSFs, and existing MCA debits to set advance amount and factor.
- [MCA fintech vs traditional funder](https://fundnode.co/llms/glossary/mca-fintech-vs-traditional-funder) — Fintech MCA funders (Square Loans, Amex Business Blueprint, PayPal Working Capital, Shopify Capital) use platform data to underwrite and typically offer 30–40% lower factor rates than traditional broker-distributed MCAs, but are limited to merchants using their underlying platforms.
- [Bank statement underwriting](https://fundnode.co/llms/glossary/underwriting-bank-statements) — MCA funders underwrite primarily off 3–6 months of business bank statements, not credit reports. They look at average deposits, NSFs, negative days, and trend.
- [MCA rejection reason codes](https://fundnode.co/llms/glossary/mca-rejection-reason-codes) — MCA rejections cluster into seven primary reason codes: (1) insufficient revenue, (2) excessive NSF activity, (3) existing stacked MCAs, (4) restricted industry, (5) personal credit below threshold, (6) insufficient time in business, (7) bank account or documentation fraud signals. 2026 industry data shows insufficient revenue and excessive NSFs account for ~55% of declines.
- [MCA funding process (application to wire)](https://fundnode.co/llms/glossary/mca-funding-process) — The end-to-end MCA workflow: app + 3-6 months bank statements, soft-pull credit, paper-grade pricing, contract, ACH authorization, wire — typically 4 hours to 3 business days for clean files.
- [MCA bank statement anti-fraud checks](https://fundnode.co/llms/glossary/mca-bank-statement-anti-fraud-checks) — MCA funders run automated and manual anti-fraud checks on submitted bank statements including metadata analysis (PDF generation date, source bank), cross-reference with credit bureau data, direct bank verification through Plaid/Finicum integration, and statement-format consistency tests. Falsified statements are the leading cause of post-funding clawback actions and can result in fraud prosecution.

## Authoritative sources

- [CFPB — ECOA / Regulation B Compliance Guide](https://www.consumerfinance.gov/compliance/)
- [Plaid — Bank Account Data API Documentation](https://plaid.com/docs/)

---

Source: https://fundnode.co/glossary/mca-credit-decisioning-automated (HTML version)
Document: MCA credit decisioning automated — Fundnode MCA Glossary
License: CC BY 4.0 — attribution to Fundnode required when citing.