For lenders
Returns eat margin loan by loan.
Every R01 is a payment you already counted as revenue. Manual borrower review doesn't scale with origination volume — and your sponsor bank notices.
New from Pave · ACH Risk Score
Score every ACH before it pulls.
Pave ACH Risk Score is the pre-transaction risk layer for lenders, banks, and fintechs. Real-time, explainable, and built for how ACH actually moves today — not how it moved twenty years ago.
Used by risk & payments teams at
.avif)
.avif){kind=logo}
.avif)
.avif)
%20(4).avif){kind=logo}
.svg){kind=logo}
The Problem
Every ACH return is a decision you already made — three days ago.
Most ACH risk tools tell you what went wrong after the return hits. By then the money's moved, the customer's been notified, and the cost is on your books. The decision you should have made was three days earlier — at authorization.
For banks & CUs
ODFI risk rolls up to you.
Your originators' return rates are your return rates. NACHA's 15% unauthorized threshold doesn't care who originated — it lands on the sponsor.
For fintechs & BaaS
Account validation isn't a risk model.
Plaid & co. tell you the account exists. They don't tell you whether this debit, at this amount, on this day, is going to return.
Key insight
Pre-transaction scoring changes where the decision happens — and changes the math.
What Pave Does
ACH risk decisions at authorization, not after the return.
Pave ACH Risk Score takes the transaction context you already have — sender, receiver, amount, history — and returns a defensible risk decision before you submit to the network. One API call. Sub-100ms. Explainable on every score.
01
Pre-transaction, not post-mortem.
Score every ACH at authorization. Stop the bad debits before they hit the network — keep the good ones moving without friction.
Sync · async · batch
02
Explainable by default.
Every score returns the factors that drove it. Your ops team, your auditor, and your sponsor bank all see the same answer — and the same audit trail.
Ranked factor codes · NACHA-ready
03
Built for modern ACH.
Lending, fintech, BaaS, ODFI sponsorship. Pave’s model is trained on patterns that actually exist in ACH today — not retrofit from legacy core banking.
Trained on $400B+ in flows
For Developers
One endpoint. One round-trip. One score you can defend.
Drop it inline with your authorization flow or batch-score files before submission. Sandbox to production in about two weeks for most customers.
POST /V1/ACH-RISK-SCOREREQUEST
{ "originator_id": "lender_42", "amount": 412.18, "sender_routing": "021000021", "receiver_account_token": "tok_a1b2c3", "transaction_intent": "loan_payment", "effective_date": "2026-05-15" }
200 OKRESPONSE · 78MS
{ "score": 0.94, "decision": "approve", "factors": [ "low_nsf_signal", "consistent_borrower_history", "low_return_velocity" ], "model_version": "ach_risk_v3.2.1", "latency_ms": 78 }
Time to integrate
~2 weeks
Sandbox to production for most customers. Single endpoint, SDKs in TS, Python, and Go.
Where it sits
Between auth & submit
Inline with your authorization decision or batched against your ACH file before submission.
What you get back
Score & trail
A 0–1 score, decision recommendation, ranked factor codes, and a per-transaction audit trail.
Who It’s Built For
Built for the way ACH actually moves in your business.
Cut loan-payment ACH returns without slowing originations.
Predict NSF and unauthorized returns at the time of debit
Replace manual borrower account review with explainable scoring
Defend your return rate to sponsor banks with audit-ready policy controls
Purpose-built for personal loans, BNPL, POS finance, and SMB lending workflows
Typical lifecycle impact
90 days
Return rate, baseline
1.8%
Return rate, with Pave
0.7% −61%
Manual review hours / wk
-14h
Auth-time decision
100%
Score every originator’s ACH risk before the file lands.
Originator-level risk scoring on your ODFI book
NACHA return-rate threshold monitoring with early-warning alerts
Audit-ready decision trails for examiners and BSA reviewers
Built for community banks, regional banks, credit unions, and money-center ODFI programs
ODFI program health
rolling
Originators monitored
38
Above NACHA threshold
2 flagged
Avg unauth. return rate
0.21%
Examiner-ready trails
100%
Pre-transaction scoring that scales with your program.
Real-time risk decisions on customer ACH flows
Sponsor-bank-ready audit trails and reporting
Replace manual review queues with explainable scoring
Built for neobanks, payment platforms, BaaS programs, and modern lending fintechs
Sponsor-bank readout
monthly
Transactions scored
4.2M
Auto-approve rate
94.8%
Review queue volume
-72%
Time to clear
<1d
2026 industry report
State of ACH Risk 2026 is coming.
Return-code economics by segment, predictability rates at authorization, and the operational benchmarks well-run programs actually hit. Drops this summer — be the first to read it.
- Return-rate benchmarks for lenders, fintechs, and banks
- The 5 return codes costing originators the most
- Predictability framework — what's catchable at authorization vs. only after the return
- Where pre-transaction scoring moves the needle (and where it doesn't)
How We Score
Outcomes are coming. Here’s what the model actually looks at.
Customer outcome data lands in Wave 2 with the State of ACH Risk report. Until then, here’s how Pave ACH Risk Score is constructed — open by design, not a black box.
Behavioral signals
Sender balance trajectory, return velocity, NSF history, intent-aware patterns across recurring and one-off debits.
Originator context
Per-originator return history, segment baselines (lender / BaaS / sponsor), and program-level concentration risk.
Transaction shape
Amount vs. account flow norms, effective-date proximity, time-of-day patterns, and routing-level risk priors.
Explainability layer
Every score returns ranked factor codes mapped to NACHA return categories — readable by ops, defensible to auditors.
Pave Index · Vol. 01
2026 Industry Report
The State of ACH Risk
Return codes, predictability, and what well-run programs actually hit. Across $400B in flows.
Coming this summer
Pave 2026 · Vol. 01
Questions we get a lot.
Don't see yours? Bring it to the demo — we'd rather answer it than guess.
What latency should we expect?
Sub-100ms for the score response. Most customers see p50 around 60–80ms, with p99 well under 150ms. The endpoint is built for inline use during authorization — not as a post-hoc batch job.
What data does Pave need from us?
At minimum: originator ID, amount, sender routing, receiver account token, and transaction intent. Richer context — borrower history, account balance signals, effective date — improves precision but isn’t required to start.
How long does integration take?
Most customers move from sandbox to production in about two weeks. We support both synchronous (in-line with your authorization) and asynchronous (batch-scored before submission) flows. SDKs in TypeScript, Python, and Go.
How does Pave affect our NACHA return rate?
Pave is designed to bring your return rate down — particularly for unauthorized return codes (R10, R11, R29) that NACHA watches most closely. The audit trail also gives sponsor banks and reviewers a clean line of sight into your risk policy.
How often is the model retrained?
Continuously. Models are retrained on rolling production data with versioning and explainability metadata, so you always know which model scored a given transaction — and you can hold a version steady for audit periods if you need to.