Accepting pilot institutions — Q2 2026

The Return
Was Predictable.
Most Are.

ArielIQ scores every ACH transaction before submission — six dimensions of ML-powered risk intelligence, NACHA-grounded decisions, and actual delivery estimates. Built for the institutions that can't afford a miss.

Request Free Pilot See How It Works
0.9096
ROC-AUC on out-of-time validation
44.8%
Mean return rate reduction
10x
ROI — guaranteed floor
arieliq/v1/score 41ms
Transaction
TX-20260408-044
Decision
Warning
SEC / Amount
WEB  ·  $12,500
Action
24-48hr hold

Six-Dimensional Risk
Composite Score
24.1%
ML Ensemble
26.3%
Unauthorized (R07/R10)
31.0%
Administrative
9.2%
Funding / NSF
26.0%
Aggregate Velocity
0.0%

Delivery
Same-Day ACH · 5:00 PM ET
Total Exposure
$683.77
Confidence Interval
12.7% — 36.1%
NACHA Status
Review Required

The Problem

Community banks are absorbing a
problem that's already solved — elsewhere.

Large banks spent $10M+ building internal ML systems that score ACH risk in real time. Community institutions are left with static rules written a decade ago — and the NACHA penalties are the same regardless of size.

$2.7B

Annual ACH return costs, U.S. banking system

Per-item costs range from $25 to $75 in direct processing expense — before investigation time, customer friction, or the legal exposure on unauthorized returns begins.

0.50%

NACHA unauthorized return threshold

Exceed it and your institution faces enforcement action. NACHA fines reach $500,000 per violation — and most institutions don't know their real-time unauthorized return rate until it's too late.

15%

NACHA overall return rate threshold

The published threshold every ODFI must stay below. The industry average is 8–12%. Without pre-submission scoring, you are managing this number reactively — with returns you could have blocked.


How It Works

Pre-submission intelligence.
One API call.

Submit a transaction payload before it enters the ACH network. Get back everything your operations team needs to make a confident, defensible decision.

01   Submit

Transaction payload

SEC code, amount, timing, account history, originator relationship, balance signal, velocity. Standard fields your core system already has.

02   Score

Six dimensions analyzed

ML ensemble model plus five rule-based dimensions — each calibrated to published NACHA return data. Response in under 50ms.

03   Act

Specific recommendations

Flag tier with recommended action, NACHA rule citation, delivery estimate with an actual date and time, cost estimate, and escalation path.

Six Risk Dimensions
Overall Return Risk
ML ensemble — Random Forest + Gradient Boosting + Logistic Regression, out-of-time validated
Unauthorized Risk (R05 / R07 / R10)
Highest legal exposure — counts toward 0.5% NACHA threshold. $1,800–$3,200 legal exposure per incident
Administrative Risk (R02 / R03 / R04)
Account closed, unable to locate, invalid routing — driven by account age and originator relationship tenure
Funding Risk (R01 / R09)
Most common return category — 40–50% of all ACH returns. Driven by balance signal and return clustering history
Aggregate & Velocity Risk
Transaction frequency signals, portfolio concentration, unusual originator patterns — portfolio-level exposure
Monte Carlo Uncertainty Bounds
95% confidence intervals on every score — worst-case estimate, signal strength, and model confidence reported in every response
Five-Tier Flag System

Every score maps to an actionable tier. Every tier cites a specific NACHA rule. Your team gets a defensible recommendation — not a number to interpret.

Informational
0–10% · Log and monitor — process normally
Advisory
10–20% · Verify authorization documentation
Warning
20–35% · 24–48hr hold — verify account status
Alert
35–55% · Escalate to senior officer
Critical
55–100% · Block — compliance officer required

Validated Performance

30 independent Monte Carlo simulations.
Law of Large Numbers verified.

Results reported as mean performance with worst-case floor — the number we guarantee. Not cherry-picked. Every metric is reproducible on your data during the pilot.

ROC-AUC
0.0000
Out-of-time validation
Mean Catch Rate
0%
60.6% worst-case floor
Return Rate Reduction
0%
42.4% worst-case floor
Net Annual Savings
$0
$105,582 guaranteed floor
Model Confidence

Scored against held-out out-of-time transaction data not seen during training. Performance is stable across all 30 simulation runs — not a single favorable draw.

ROC-AUC range across 30 simulations0.9096 mean
0.000.501.00
Mean catch rate 64.0%
Worst-case catch rate 60.6%
Validation methodology Out-of-time holdout
Simulations run 30

What You Get

Everything your operations team needs.
Nothing they don't.

Pre-Submission Transaction Scoring
Score every ACH before it touches the network. Six dimensions of risk in a single API call — response under 50ms.
NACHA-Cited Recommendations
Every recommended action cites the specific NACHA rule section. Auditable, defensible, examiner-ready documentation.
Precise Delivery Estimates
Not "1–3 business days." An actual date and time — Same-Day ACH window eligibility, Federal Reserve holiday calendar, Reg CC hold probability.
Customer Email Templates
Auto-generated customer communications by flag tier — from informational notices to critical hold notifications. Compliance-reviewed language.
Human Escalation Case Files
Structured escalation packages for your risk officers — every flag, every score, every NACHA citation. Decision-ready in seconds.
Risk Officer Dashboard
Portfolio-level visibility. Return rate tracking against NACHA thresholds. 30-day trend analysis. Everything your compliance team needs in one view.
REST API + Full Documentation
Clean JSON API with Swagger docs, batch scoring (up to 1,000 transactions), portfolio analysis endpoint, and webhook support.
Cost & Legal Exposure Estimates
Expected return cost by R-code probability, unauthorized legal exposure per incident, and total portfolio exposure — in every API response.
Pilot Backtest Report
Full retrospective analysis on your historical data — what ArielIQ would have caught, what it would have saved, at what confidence level.

Who It's For

Community banks and credit unions
$500M — $5B in assets.

Institutions large enough to face real NACHA threshold exposure. Small enough that a $10M ML build isn't on the roadmap. That gap is exactly where ArielIQ operates.

VP of Payments

Managing NACHA threshold pressure

Your return rates are visible to regulators. Pre-submission scoring gives you control over the number — not just visibility into it after the fact.

Chief Risk Officer

Unauthorized return exposure

One unauthorized return cluster can push you past the 0.5% threshold. ArielIQ flags the exposure before submission — with the specific NACHA rule citation your examiners will ask for.

CFO

Justify every line item

ArielIQ is priced at 10% of demonstrated gross savings. If we don't show positive net savings on your actual data during the pilot — you never pay. The ROI is the contract.


Free Research Pilot

90 days. Your data.
Zero commitment.

We backtest ArielIQ against your anonymized historical ACH transactions. You see the exact returns we would have caught, the savings, and the ROI — before you spend a dollar.

Full API access — sandbox and production
Custom model calibration to your portfolio
Net savings report at 30, 60, and 90 days
Direct access to the founding team
No long-term contract required
Performance guarantee — no savings, no fee
Apply for Pilot Access

5 pilot slots remaining  ·  Responses within 48 hours


Pricing

10% of what we save you.
Aligned incentives.

No flat SaaS fee. No seats. No enterprise negotiation. ArielIQ costs 10% of gross savings demonstrated on your data — with a floor and ceiling. If you don't save, you don't pay.

Institution Size ArielIQ / mo Net Savings / mo ROI
$300M assets $1,500 $12,500 9.3x
$500M assets $2,600 $23,400 10x
$1B assets $5,600 $50,400 10x
$2B assets $11,200 $100,800 10x
Floor: $1,500 / month   ·   Ceiling: $20,000 / month   ·   Savings calculations use conservative 42.4% worst-case return rate reduction

About

Built by a machine learning
researcher. For your examiners.

KP

Kathryn Perry, PhD Candidate

Kathryn Perry

PhD Candidate in Machine Learning · University of Texas at San Antonio

Research focus: ML applications in financial risk systems. ArielIQ is grounded in the same methods used in peer-reviewed financial ML research — not vendor marketing claims.
Ensemble architecture: Random Forest + Gradient Boosting + Logistic Regression, calibrated to NACHA-published return rate data across all eight SEC codes.
Validated across 30 independent Monte Carlo simulations using Law of Large Numbers methodology — every reported metric has a worst-case floor, not just a mean.
Out-of-time validation: model trained on historical data, tested on future data it has never seen. ROC-AUC of 0.9096 — well above the 0.80 threshold considered clinically meaningful in predictive modeling.
Every NACHA threshold, rule citation, and compliance flag in ArielIQ is sourced from official NACHA Operating Rules and Federal Reserve ACH data — not approximations.

Request Your Free Pilot

Let's look at
your data.

Tell us about your institution. We'll set up a 30-minute call to scope the pilot and confirm fit — no sales process, no deck. Just your data and ours.

Response Time

Within one business day

Location

San Antonio, TX  ·  Remote pilot delivery nationwide

Message received.

We'll follow up within one business day to schedule a scoping call. Check your inbox — it will come from kathrynperry@arieliq.com.