Interest Rate Risk KRIs: 8 Metrics for Banks and Credit Unions Monitoring IRRBB

TL;DR

• FDIC data shows $325.1 billion in unrealized losses across the banking system in Q1 2026—the IRR problem isn’t resolved, it’s just quieter
• Most IRRBB monitoring programs track NII sensitivity but skip the metrics that give earlier warning: EVE sensitivity, NMD assumption drift, and deposit beta variance
• The 8 KRIs covered here are what examiners want to see in your ALCO dashboard and board reports
• SVB’s failure was traceable to gaps in exactly these metrics; regulators have been raising the bar on IRRBB monitoring since March 2023

The post-SVB period gave risk managers a brief window where interest rate risk was a board-level conversation. That window is closing. Rates have stabilized, the immediate crisis has passed, and IRRBB is drifting back into the ALCO deck that nobody reads carefully.

But the numbers haven’t gone away. FDIC data for Q1 2026 shows the banking system is still carrying $325.1 billion in unrealized losses on investment securities and loans—down from the 2022 peak, but still significant. Net interest margin for community banks declined 8 basis points to 3.31% in Q1 2026, continuing a compression trend that started when deposit repricing accelerated faster than asset yields. The 54 banks on the FDIC’s problem bank list as of Q1 2026 include institutions where this compression has eaten into the capital buffer.

Interest rate risk in the banking book doesn’t announce itself. It accumulates in model assumptions, in duration mismatches, in deposit behavior that diverges from historical patterns. The point of IRRBB KRIs isn’t to predict the next crisis—it’s to give your board and ALCO the 60-to-180-day early warning that something is drifting in a direction that will matter.

Here are the 8 metrics that should be in every institution’s IRRBB monitoring framework.

What Regulators Actually Expect From IRRBB Monitoring

Before the metrics, context on what examiners are looking for. OCC examination guidance expects institutions to test a minimum of four rate shock scenarios: +100, +200, +300, and +400 basis points. The FDIC Risk Management Manual Section 7.1, updated March 31, 2026, reinforces that institutions with concentrated long-duration asset portfolios must demonstrate that their models capture non-parallel rate shifts—steepening and flattening twists—not just parallel shocks.

NCUA regulations apply comparable expectations to federally insured credit unions, with heightened scrutiny for institutions over $500 million in assets. For state-chartered banks and credit unions, the applicable state prudential regulator typically mirrors federal standards.

The Federal Reserve’s 2026 stress test found that the 23 largest U.S. bank holding companies could absorb $708 billion in hypothetical losses under the severely adverse scenario—but that stress test applies to the largest institutions. Community banks and credit unions face a different risk profile precisely because their IRR monitoring programs are less sophisticated, not more.

What examiners are specifically asking now: Can you show that your KRI thresholds are calibrated to your specific balance sheet, not copied from a generic template? Can you demonstrate that your board received timely alerts when thresholds were breached? And can you show that your model assumptions—particularly for non-maturity deposits—reflect actual behavior rather than historical averages that no longer apply?

KRI 1: NII at Risk (Earnings at Risk)

What it measures: The change in net interest income over a 12-month forward horizon under defined rate shock scenarios, expressed as a percentage of base-case NII or as a dollar amount relative to capital.

Why it matters: NII at risk is the most widely used IRRBB metric—89% of banks report using NII sensitivity as their primary IRR measure according to survey data. It’s intuitive: if rates shift, here’s what happens to your income. But it’s also lagging—it measures current period earnings exposure, not structural balance sheet risk.

Threshold guidance: A decline of more than 15-20% in NII under a +200 bps parallel shock warrants ALCO review and should appear in board reporting. A decline exceeding 25% in the same scenario should trigger a formal board discussion of hedging strategy and asset-liability positioning.

SVB lesson: SVB’s NII models were optimistic about deposit stability under rising rates. When modeled NII at risk looked manageable, the structural risk—concentrated in the EVE metric—wasn’t receiving equal attention.

KRI 2: EVE Sensitivity (Economic Value of Equity)

What it measures: The change in the present value of the institution’s net assets under rate shock scenarios, expressed as a percentage of base EVE or as a percentage of capital.

Why it matters: EVE captures the structural balance sheet mismatch—what happens to the economic value of long-duration assets and liabilities when rates shift. Unlike NII at risk, EVE sensitivity shows risk that is locked into the balance sheet today, even if it won’t show up in income statements for years.

Threshold guidance: A decline exceeding 15% of capital under a +200 bps shock is a common internal threshold. BCBS d368 (the Basel interest rate risk standard) uses a 15% of Tier 1 capital threshold as a supervisory outlier criterion—institutions exceeding this are expected to explain why the level is appropriate given their risk appetite.

What most institutions miss: Many community bank IRR models calculate EVE but don’t track it as a KRI with formal thresholds and escalation triggers. It sits in the ALCO model output without a defined governance response.

KRI 3: Net Interest Margin Trend

What it measures: Net interest income as a percentage of average earning assets, tracked quarterly against the institution’s own historical trend and peer benchmarks.

Why it matters: NIM compression is often the first visible signal of interest rate risk crystallizing. When deposit costs reprice faster than asset yields, the margin narrows. When it narrows faster than modeled, it signals that your IRR assumptions are more optimistic than your actual experience.

Threshold guidance: A NIM decline of more than 25 basis points over two consecutive quarters warrants ALCO review. A decline of more than 50 basis points over four quarters should trigger a formal model re-validation to determine whether the divergence from modeled outcomes reflects model error or balance sheet repositioning.

Peer context: The industrywide NIM decline of 8 bps in Q1 2026 provides a benchmark—if your institution is compressing faster than peer average without an explainable structural difference, that’s a KRI signal.

KRI 4: Duration Gap

What it measures: The difference between the weighted average duration of earning assets and the weighted average duration of interest-bearing liabilities, adjusted for the equity leverage ratio.

Why it matters: A positive duration gap means assets reprice more slowly than liabilities—the institution is exposed to rising rates compressing margin and EVE. A negative duration gap means the opposite: assets reprice faster, creating exposure to falling rates. Duration gap is the single most compact representation of structural IRR positioning.

Threshold guidance: A positive duration gap exceeding 2-3 years warrants monitoring; exceeding 4-5 years typically triggers examiner questions. SVB’s duration gap was extreme—56% of its securities portfolio had repricing horizons exceeding 15 years, a concentration that should have been a red-line KRI breach.

Calculation note: Duration gap calculations require assumptions about liability duration that often understate actual repricing behavior. The gap between the duration you calculate and the duration you’re actually running is itself a risk indicator—which connects to KRI 5.

KRI 5: NMD Assumption Drift

What it measures: The divergence between modeled non-maturity deposit (NMD) behavior—average life, decay rate, core vs. rate-sensitive classification—and actual observed behavior, tracked quarterly.

Why it matters: NMDs (checking, savings, money market deposits without contractual maturity) are the foundation of most community bank and credit union IRR models. Their assumed stability and duration drive EVE calculations and NII projections. When actual deposit behavior—decay rates, beta coefficients, outflow under rate stress—diverges from model assumptions, every downstream calculation is wrong.

How to measure it: Compare modeled deposit decay rates and beta assumptions against actual quarterly experience. If your model assumed 15% of rate-sensitive deposits would move when the Fed raised rates and 40% actually moved, that drift needs to be documented and fed back into model re-calibration.

Why SVB is the case study here: SVB’s deposit base was heavily concentrated in venture-backed companies with correlated withdrawal behavior—when funding conditions tightened, deposits left simultaneously. The model assumptions, calibrated on historical data from a different deposit composition, didn’t capture this concentration risk.

KRI 6: Repricing Gap by Maturity Bucket

What it measures: The net difference between rate-sensitive assets and rate-sensitive liabilities maturing or repricing within defined time buckets: 0-3 months, 3-12 months, 1-3 years, 3-5 years, and 5+ years.

Why it matters: The repricing gap quantifies the mismatch that drives NII sensitivity. A large negative gap in the 0-3 month bucket means liabilities reprice faster than assets in the short term—margin pressure in a rising rate environment. A large positive gap in the 5+ year bucket means the institution is carrying long-duration assets that will lose value if rates rise.

Threshold guidance: A cumulative repricing gap exceeding 20% of total assets in any single time bucket warrants monitoring. An institution with a large positive gap in the 5+ year bucket and a large negative gap in the 0-3 month bucket is running a structural carry trade that requires explicit board-level risk acceptance.

Practical application: Present this as a waterfall chart in ALCO reporting. The visual representation of where gaps concentrate—and whether they’re growing or shrinking—is more actionable than a single aggregate number.

KRI 7: Unrealized Loss Coverage Ratio

What it measures: The ratio of unrealized losses on available-for-sale (AFS) and held-to-maturity (HTM) securities to the institution’s Tier 1 capital, updated quarterly.

Why it matters: This is the KRI that crystallized in the SVB crisis. Unrealized losses aren’t economic losses until securities are sold—but if the institution faces a liquidity event that forces sales, those losses become real capital losses. At SVB, unrealized HTM losses exceeded Tier 1 capital before the bank run began.

Threshold guidance: An unrealized loss coverage ratio below 1.0x (losses exceed capital) is a crisis-level indicator. A ratio below 2.0x warrants enhanced monitoring and contingency liquidity planning. Any deterioration exceeding 25% quarter-over-quarter should appear in board reporting regardless of absolute level.

Data source: FDIC quarterly banking profile data allows institutions to benchmark their unrealized loss position against peers—the $325.1 billion industrywide figure for Q1 2026 provides context for where your institution stands relative to the system.

KRI 8: Deposit Beta vs. Modeled Assumption

What it measures: The actual rate paid on interest-bearing deposits divided by the change in the benchmark rate (Fed funds, SOFR), compared against the beta coefficient embedded in your IRR model.

Why it matters: Deposit beta is the most consequential assumption in most community bank IRR models. If your model assumes a 40% beta (deposits reprice at 40 cents for every dollar of rate increase) and actual beta is running at 65%, your NII projections are materially overstated and your EVE calculations are understated.

How to track it: Calculate actual beta quarterly using your ALCO data. Maintain a rolling 4-quarter average. Any quarter where actual beta exceeds modeled beta by more than 10 percentage points should trigger a model assumption review.

The current environment relevance: Deposit beta assumptions calibrated on 2010-2019 data are particularly suspect—the post-zero-interest-rate period saw compressed betas that don’t reflect how depositors behave when alternatives (money market funds, Treasury bills) are yielding 4-5%. Many institutions are discovering this divergence in their quarterly ALCO reviews right now.

Building an IRRBB KRI Dashboard

The eight metrics above work as a coherent monitoring system when you run them together. Here’s a simplified structure for presenting them:

The thresholds above are starting points. Calibrating them to your institution’s specific balance sheet composition, peer group, and risk appetite is what turns a generic framework into a governance tool that examiners will respect.

So What? What to Do With This

If your ALCO deck already tracks all eight of these with calibrated thresholds and a defined escalation path, you’re ahead of most peers. More likely, you’re tracking NII at risk and NIM trend, have EVE somewhere in the model output, and don’t have formal thresholds on the others.

The quick win: pull your last four ALCO packages and check which of these eight metrics appear in the board reporting, with what thresholds, and what happened when those thresholds were approached or breached. Gaps in that documentation are gaps that examiners will find.

The bigger lift: building NMD assumption drift and deposit beta variance into your quarterly monitoring requires some model infrastructure—you need to be capturing actual deposit behavior data in a format that lets you calculate these ratios. For many community institutions, this is the work that follows reading this article.

For institutions building or refreshing their KRI library, the KRI Library includes pre-built IRRBB KRI templates with threshold guidance and board reporting formats calibrated for community banks and credit unions—structured to meet current OCC and FDIC examination expectations without requiring a full model rebuild to implement.

Related Reading

• KRI Fundamentals: Building a Key Risk Indicator Library That Survives Examination
• Operational Risk Management Framework: What Community Banks Get Wrong

Sources:

1. FDIC Quarterly Banking Profile, Q1 2026 — Unrealized Losses and Net Interest Margin data
2. FDIC Risk Management Manual, Section 7.1: Interest Rate Risk — updated March 31, 2026
3. OCC Comptroller’s Handbook: Interest Rate Risk (current edition)
4. Federal Reserve 2026 Stress Test Results, June 2026
5. Basel Committee on Banking Supervision, “Standards — Interest Rate Risk in the Banking Book” (BCBS d368)