Fundnode · Learn

Glossary · MCA funder decisioning engine (typical)

MCA funder decisioning engine (typical)

Typical MCA funder decisioning engine in 2026 is a rules-plus-ML pipeline: hard knockouts (credit, deposit minimums, industry exclusions), then risk-pricing model, then human underwriter review for edge cases — producing decisions in 5 minutes to 4 hours.

By Keerthana Keti5 min read

Decisioning engines are the operational nervous system of MCA funders — the orchestrated pipeline that takes a submission and produces an approve/decline/refer-up decision with a factor rate attached.

The standard decisioning architecture (2026).

A typical funder runs a 5-stage pipeline:

  1. Intake validation. Application completeness check (required fields, document uploads, signatures).
  2. Hard knockouts. Rules-based exclusions that auto-decline (sub-500 credit, sub-$10K deposits, sanctioned industries, bankrupt status).
  3. Bank-statement analysis. Automated extraction of underwriting signals (deposits, NSFs, stacking, cash flow).
  4. Risk-pricing model. ML-driven default probability + recommended factor rate.
  5. Underwriter review. Human review for exceptions, edge cases, large advances, or unusual industries.

Decisioning timeline.

  • Sub-1-minute decisions: Platform-native (Toast Capital, Square Capital) for pre-qualified merchants with API data flow.
  • 5–30 minute decisions: Direct-digital funders with Plaid bank-data access and standard merchant profile.
  • 1–4 hour decisions: ISO-channel funders with PDF statement upload, automated pipeline, no underwriter touch.
  • 4–24 hour decisions: Deals requiring underwriter review (exceptions, large advances, complex ownership).
  • 24–72 hour decisions: Deals requiring manual document review, tax return analysis, or referral up to senior underwriter.

Hard knockout rules (typical 2026).

  • Personal credit score <500.
  • Trailing 12-month deposits <$120K ($10K/month).
  • Time in business <6 months (some funders 12 months).
  • Open bankruptcy.
  • Sanctioned industries (cannabis non-licensed states, firearms, adult, gambling).
  • Active stacking flag from UCC + bank-statement evidence.
  • Personal-guarantor fraud flag from credit report.
  • Out-of-funder-territory states (some funders state-restricted).

Risk-pricing model layer.

After hard knockouts, the surviving applications run through the funder's proprietary risk model (see /glossary/mca-funder-risk-pricing-model-2026). Output: predicted default probability + base factor rate + recommended advance amount.

Underwriter review triggers.

The decisioning engine flags deals for human underwriter review when:

  • Advance amount >$100K (or $250K depending on funder).
  • Industry in "review" bucket (construction, freight, restaurants with seasonal patterns).
  • Bank-statement anomalies (sudden spikes, irregular patterns, missing months).
  • Stacking-suspicion flags without conclusive evidence.
  • First-time submission from a new ISO.
  • Personal guarantor with complex financial profile (multiple businesses).
  • Repeat decline candidate with new bank statements.

Decisioning engine platforms.

  • In-house custom builds. Most top-50 funders maintain custom decisioning engines in Python/Java/Go.
  • Lendio, OnDeck-style platforms. Some funders license decisioning infrastructure from broker-tech platforms.
  • Heron Data + Ocrolus + custom rules engine. Common middle-tier architecture.
  • Salesforce Financial Services Cloud + custom rules. Some larger funders run on Salesforce as the case-management layer.

Decisioning output to ISO portal.

  • Decision (approve/decline/refer).
  • If approved: advance amount, factor rate, term, daily payment, ISO commission.
  • If declined: reason codes (often abbreviated — "INSUF_DEPOSITS", "NSF_COUNT", "STACKING").
  • If refer: pending-underwriter-review status.

Workflow orchestration.

Top-tier funders use workflow tools (Camunda, Temporal, Airflow) to orchestrate the decisioning pipeline. Each stage has SLA targets (e.g., bank-statement parse within 3 minutes, model decision within 60 seconds, underwriter review within 2 hours).

ISO-facing decisioning experience.

  • Real-time status updates in ISO portal (submitted → under review → approved/declined → funded).
  • Instant approval notifications via email and SMS.
  • Decline reason codes with enough specificity for ISO to coach merchant.
  • Refer-up notifications when underwriter review needed.

Decisioning engine SLAs by tier.

  • Tier 1 / Platinum ISO submissions: 24-hour decision SLA, often sub-4-hour in practice.
  • Tier 2 / Gold ISO: 48-hour SLA.
  • Tier 3 / Silver ISO: 72-hour SLA.
  • New / Bronze ISO: Best-effort, often 3–5 days.

2026 trends in decisioning engines.

  • AI-augmented underwriter review. LLM-summarized application context for underwriters reduces review time 40–60%.
  • Real-time monitoring decisioning. Engines that continuously re-decision during advance lifetime based on new bank-data signals.
  • Federated decisioning across funders. ISO-tech platforms (Onyx, Funder Intelligence, similar) routing single application to multiple funder engines simultaneously.
  • Explainability layer. Regulatory environments (CA, NY, UT, VA, GA) demanding explainable decline reasons.

Worked example: a clean B-paper application.

  • T+0:00 — ISO submits via API.
  • T+0:01 — intake validation passes.
  • T+0:02 — hard knockout pass (FICO 615, $32K deposits/month, restaurant, 24 months operating).
  • T+1:30 — Ocrolus parses 3-month statements; extracts 1 NSF, $32K avg deposits, no MCA stacking.
  • T+1:35 — risk-pricing model outputs 12% predicted default, 1.34 factor recommendation for $50K advance.
  • T+1:36 — auto-approval issued; ISO portal updated.
  • Total: under 2 minutes from submission to approval.

Common confusions.

First, "decisioning is all AI." False — most decisions still have rules-based hard knockouts.

Second, "fast decisioning = lower quality." Not necessarily — well-tuned engines are faster AND more accurate.

Third, "underwriters are obsolete." False — human review still required for exceptions, large advances, and edge cases.

Fourth, "decisioning engines are open." Almost always closed and proprietary.

Fifth, "all funders decide on the same data." False — data sources, parsing accuracy, and model design vary widely.

Related terms

  • MCA funder risk-pricing model (2026)MCA funder risk-pricing models in 2026 use 8–15 inputs (credit score, deposit volume, NSF count, time-in-business, industry, geography, stacking history, cash-flow stability) feeding a logistic-regression or gradient-boosted-tree default predictor that maps to factor rates from 1.15 to 1.50.
  • Bank statement underwritingMCA 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 funder application decision time by tier (2026)A-paper MCA decisions in 2026: 4–24 hours. B-paper: 24–48 hours. C-paper: 24–72 hours. D-paper: 48–96 hours. Funding follows decision by 4–24 hours for clean files.
  • MCA funder bank-statement analysis softwareMCA funders in 2026 use bank-statement analysis software like Ocrolus, Heron Data, Nanonets, Validis, and proprietary in-house parsers to extract deposit volumes, NSF counts, MCA debit signatures, and cash-flow patterns from PDF statements in 30–90 seconds.

Authoritative sources

AI agents: this term is available as raw markdown at /llms/glossary/mca-funder-decisioning-engine-typical.