Executive Summary
A top-5 U.S. commercial bank by total assets, headquartered in Minneapolis, Minnesota, with approximately $680 billion in assets, 70,000+ employees, and operations spanning 26 states, faced mounting regulatory pressure following the Federal Reserve’s 2025 supervisory priorities, which placed explicit emphasis on Internal Liquidity Stress Testing (ILST), deposit run assumptions, and derivatives funding methodologies.
Their existing stress testing infrastructure relied on static, batch-processed quarterly models that could not ingest real-time transaction flows from FedNow, CHIPS, or Fedwire, preventing supervisors from assessing intraday liquidity exposure during fast-moving deposit events.
After deploying the IQ Platform’s Qubz analytics engine, architected by UTOFA, the bank reduced its ILST cycle from five business days to under four hours, closed two open Federal Reserve Matters Requiring Attention (MRAs), and achieved a validated deposit-assumption model accuracy within ±2.1% variance against live run-scenario benchmarks.
Reduction in Error Rates
Decrease in Processing Time
First-Year ROI
SITUATION
This Minneapolis-based bank is one of the largest depository institutions in the United States, operating a network of over 2,200 branches and processing in excess of 50 million payment transactions daily across FedNow, CHIPS, and Fedwire settlement rails. With a balance sheet of approximately $680 billion in total assets and operations spanning 26 states, the institution sits at the center of critical U.S. payment infrastructure.
The bank’s wholesale funding mix includes significant institutional and commercial deposits, a segment with historically elevated run-rate sensitivity, leaving the bank carrying a material proportion of uninsured deposits. This structural exposure made robust, real-time liquidity visibility not just a regulatory requirement, but an operational imperative.
Despite participating in FedNow since its July 2023 launch, the bank’s risk infrastructure had not been updated to ingest FedNow’s real-time settlement data for stress simulation purposes. Treasury and financial risk management teams instead operated under a quarterly ILST cadence, relying on Excel-based aggregation pipelines that pulled static snapshots from core banking systems and supplemented them with manually curated deposit segmentation tables.
TRIGGER
The March 2023 collapse of Silicon Valley Bank, which saw $42 billion in deposits flee in a single day, with another $100 billion staged to exit the following morning, proved that a concentrated, uninsured deposit base could be catastrophically drained within hours. With 93.8% of SVB’s deposits exceeding FDIC insurance limits, the run unfolded at a speed no quarterly stress model was built to detect.
The Federal Reserve’s OIG material loss review concluded that SVB “lacked several foundational liquidity risk management elements,” including granular deposit segmentation and comprehensive contingent funding plan testing. The Fed formalized those findings into its 2025 supervisory priorities, raising the evidentiary bar for ILST rigor, deposit run assumptions, and recovery plan plausibility across all large depository institutions.
This bank received two open MRAs citing regulatory compliance deficiencies in exactly the areas the OIG identified: granular intraday cash flow modeling and dynamic shock scenario generation. Quarterly ILST were no longer defensible, and the cost of inaction; restricted capital distributions, and elevated examination frequency, was now measurable in dollars.
BARRIER
The liquidity risk team identified four structural obstacles preventing compliance with the updated supervisory standard:
Siloed payment rail data:
FedNow transaction flows, CHIPS net settlement positions, and Fedwire gross settlement records were stored in three separate systems with no unified intraday view, requiring 2-3 days of manual data reconciliation before any scenario could be run.
Static deposit segmentation:
The bank’s deposit assumption models divided accounts into broad institutional tiers, producing ±18% variance when back-tested against the actual SVB-analog stress event, well outside the ±5% tolerance the Federal Reserve’s examiners expected.
Quarterly batch cadence:
Stress scenarios were executed once per quarter using overnight batch jobs, a model structurally incapable of simulating a shock event unfolding within a two-hour window.
Manual MRA remediation burden:
Fourteen full-time equivalents (FTEs) across treasury operations and risk analytics were consumed quarterly in data preparation, model reconciliation, and examiner-facing reporting packages, with no capacity to run ad-hoc scenarios between cycles.
Implementation hurdle during discovery:
The team discovered that Fedwire gross settlement records from the bank’s legacy mainframe were transmitted in a non-standard fixed-width ASCII format that heavily burdened IT support and was not covered by the IQ Platform’s standard connector library.
Solution
The IQ Platform’s Qubz analytics engine was deployed on-premise within the bank’s existing VPC, unifying real-time transaction data across FedNow, CHIPS, and Fedwire into a single multidimensional index refreshed every five minutes, eliminating the manual 2–3 day reconciliation lag entirely. Its Intelligent ETL layer re-segmented 4.2 million deposit accounts across 23 behavioral dimensions, enabling on-demand shock scenarios, including a 20% uninsured deposit withdrawal within two hours, to run in real time.
Why IQ Platform?
The bank selected the IQ Platform over incumbent RPA-based vendors on three criteria: Zero-Copy real-time data integration without ETL latency, sovereign on-premise deployment with 100% data residency, and an SLM-First inference engine built for live intraday execution. Competing platforms required 6–9 months for comparable integrations and could not run live shock simulations in an examiner-facing setting, the single capability that closed both open MRAs within one examination cycle.
Technical Implementation:
Data Integration Layer: The IQ Platform’s Zero-Copy Deep Data Integration connector was deployed to unify FedNow settlement feeds, CHIPS bilateral net positions, and Fedwire gross transaction logs into a single multidimensional index updated at five-minute intervals, eliminating the 2-3 day reconciliation lag.
Granular Deposit Segmentation Engine: Qubz’s Intelligent ETL layer re-segmented the bank’s 4.2 million commercial and institutional deposit accounts across 23 behavioral dimensions, including deposit age, account type, uninsured exposure threshold, counterparty sector, and historical run correlation, replacing the bank’s four coarse tiers with a dynamically updated model capable of sub-minute recalculation.
Interactive Shock Scenario Workbench: Examiners and treasury risk officers gained access to a scenario workbench that allowed parametric simulation of named stress events, including a 20% withdrawal of uninsured deposits within two hours, correlated FedFunds rate jumps, and intraday CHIPS settlement failures, with immediate visualization of the bank’s liquidity position, survival horizon, and contingent funding gap.
The Results
$18.4M in annualized compliance cost avoidance
Combining regulatory remediation labor, avoided examination escalation costs, and projected capital distribution restrictions lifted upon MRA closure.
ILST cycle reduced from 5 business days to under 4 hours
Enabling intraday stress runs during live market events rather than post-hoc quarterly analysis.
Deposit assumption model accuracy improved from ±18% to ±2.1% variance
Validated against a live SVB-analog shock simulation run at supervisory request, meeting the Fed’s examination tolerance threshold.
210% ROI achieved within 18 months
Measured against full implementation, managed service, and infrastructure cost basis.
Both Federal Reserve MRAs Closed in a Single Examination Cycle
2 open Federal Reserve MRAs formally closed within the first examination cycle post-deployment; supervisory rating upgraded, removing dividend restriction triggers.
Full production deployment in 14 weeks
Including legacy Fedwire data normalization, against an industry average of 6-9 months for comparable core banking integrations.
"The IQ Platform gave our liquidity risk team something we genuinely did not have before: the ability to watch a deposit run develop in real time and quantify our survival horizon within minutes. The granularity of the deposit segmentation model alone would have changed our calculus materially in a March 2023-type scenario."
"When examiners asked us to demonstrate a 20% uninsured deposit withdrawal shock within a two-hour window, we ran it live in the room. That single demonstration resolved the core substance of both MRAs. No spreadsheet-driven program would have let us do that."
Summary
- Cut ILST cycle time from 5 days to 4 hours (90% reduction), enabling real-time stress testing during market events.
- Closed 2 Federal Reserve MRAs in one exam cycle, avoiding $18.4M in annualized compliance costs (84.8% savings).
- Improved deposit model accuracy from ±18% to ±2.1%, meeting Fed tolerance via granular segmentation of 4.2M accounts.
- Unified siloed payment data across FedNow, CHIPS, and Fedwire (99.9% unification), eliminating 2-3 day reconciliation.
- Achieved 210% ROI in 18 months, with 14-week deployment including legacy Fedwire normalization.
- Boosted team productivity, reducing FTEs on manual prep by 78.6% while handling more scenarios.
Key Metrics
| KPI | Before IQ Platform | After IQ Platform |
|---|---|---|
| ILST Cycle Time | 5 business days (batch) | Under 4 hours (real-time) |
| Deposit Assumption Accuracy | ±18% variance | ±2.1% variance |
| Payment Rail Data Unification | Manual; 2–3 day lag | Unified; 5-minute refresh |
| Scenario Execution Cadence | Quarterly | On-demand, intraday |
| Open Fed MRAs | 2 | 0 |
| FTEs on Manual ILST Preparation | 14 FTEs | 3 FTEs (redeployed to analytics) |
| Annualized Compliance Cost | ~$21.7M | ~$3.3M (net of platform cost) |
Client details available under NDA for qualified prospects evaluating liquidity stress testing solutions.
