Lessons from SVB & Signature Bank: What Their Liquidity Failures Mean for Your CFP

$42 billion left Silicon Valley Bank in a single day. By the time the FDIC walked through the door on March 10, 2023, the bank had attempted to liquidate its investment portfolio and failed to raise emergency capital in the same 24-hour window. Forty-eight hours later, Signature Bank — $110 billion in assets — was closed by the New York Department of Financial Services. Two months after that, First Republic Bank followed.

Three banks. Over $500 billion in combined assets. All liquidity failures. None of them technically insolvent from credit losses when the runs began.

SVB had a contingency funding plan. It had a treasury department. It had stress testing. It had risk committees and board oversight. And none of it worked. The Federal Reserve’s April 2023 post-mortem didn’t hide from this: the SVB supervision review documented exactly how a $209 billion bank with all the right governance structures failed to manage a liquidity crisis that its own models had been predicting for months.

If your CFP was designed before March 2023 and hasn’t been stress-tested against these failure modes, there’s a real chance it would have made the same mistakes.

TL;DR

• SVB’s internal stress tests showed an $18B 30-day liquidity deficit in August 2022 — management changed model assumptions instead of fixing the balance sheet; the Federal Reserve cited this explicitly
• Three compounding CFP failures: concentration risk not modeled realistically, HTM securities miscounted as liquid, contingent funding sources (especially the discount window) not operationally tested
• The July 2023 interagency addendum (OCC Bulletin 2023-25 / FDIC FIL-39-2023) — the first CFP guidance update in 13 years — now requires operational testing of discount window access, not theoretical listing
• Community banks and credit unions face the same exam expectations as large institutions; examiners are applying post-SVB scrutiny at every size tier
• The most important lesson isn’t about SVB specifically — it’s about the governance architecture that allows a model to be gamed rather than a balance sheet to be fixed

The Speed of Failure

The SVB bank run set a record. $42 billion in a single day — roughly 25% of total deposits — is a withdrawal velocity that no pre-2023 runoff assumption table had modeled. The mechanism was new: a combination of VC group chats and social media amplified institutional concern into simultaneous action across a highly concentrated, technically sophisticated depositor base that could move money instantly.

The Federal Reserve’s post-mortem put the specific numbers on the timeline: by March 9, $100+ billion in withdrawal requests were already queued for March 10. The bank was effectively over before regulators could respond.

The concentration math was extreme: 94% of SVB’s total deposits were uninsured. That’s not just unusual — it’s the highest concentration among large US banks. Standard LCR runoff assumptions assign 40% runoff rates to operational corporate deposits and 100% to non-operational wholesale deposits. Neither table had a row for “94% uninsured deposits concentrated in a single industry, in an environment where VC partners are actively advising portfolio companies to move funds.”

Understanding the speed and mechanism is essential context for everything that follows, because the CFP failures weren’t random — they were predictable given what SVB’s actual risk profile looked like.

Five CFP Failures That Doomed SVB

1. Concentration Risk Wasn’t Modeled Realistically

The standard deposit runoff assumptions in SVB’s CFP scenarios were calibrated to diversified depositor bases, not to what SVB actually had. When 94% of your deposits are uninsured and concentrated in a single industry, your runoff scenarios need to model what happens when that industry gets stressed simultaneously — not what happens to a typical regional bank.

This is the scenario design problem. Most CFP stress scenarios use industry-standard runoff tables because they’re defensible and efficient. But defensibility and accuracy are different things. SVB’s depositor base was not average — it was structurally more concentrated and flighty than any standard runoff table assumed — and the CFP never calibrated to that reality.

The liquidity stress testing guide covers the assumption documentation requirements in detail. The critical point: your assumptions must be calibrated to your actual depositor profile, not industry defaults.

2. $91 Billion in HTM Securities Was a Liquidity Trap

SVB held $91 billion in held-to-maturity (HTM) securities. These securities had substantial unrealized losses by 2022 as interest rates rose — losses that hadn’t hit the income statement because HTM accounting defers recognition.

The practical problem: those $91 billion in assets could not be liquidated without recognizing the losses, which would have destroyed capital and triggered regulatory action. They were real assets on the balance sheet that could not actually be converted to cash under stress. A CFP that includes HTM securities in its liquidity analysis is documenting a fiction.

Post-SVB, examiners specifically check whether HTM securities are excluded from liquid asset calculations. Any CFP that counts HTM positions as available liquidity is going to generate a finding.

3. Management Changed Model Assumptions Instead of the Balance Sheet

This is the failure that the Federal Reserve documented most sharply, and the one with the broadest implications for CFP governance.

By August 2022, SVB’s own internal liquidity stress tests showed a 30-day deficit of approximately $18 billion and a 90-day deficit of approximately $23 billion. These weren’t regulatory findings — they were SVB’s own models, run by SVB’s own treasury team.

In October 2022, management changed the model assumptions, reducing the reported deficit by approximately $13 billion. The balance sheet didn’t improve. The liquidity didn’t materialize. The model changed.

The Fed had already issued Matters Requiring Immediate Attention (MRIAs) for CFP and stress testing deficiencies in 2021 — the most serious supervisory findings, requiring corrective action. The board approved the CFP framework while those MRIAs were still outstanding.

This is the governance architecture problem: when stress test results drive executive compensation, board presentations, and regulatory conversations, the incentive to manage the model rather than manage the risk becomes real. The CFP’s value depends entirely on whether the governance structure around it treats model results as information to act on — not as outputs to optimize.

4. The Discount Window Was Never Operationally Tested

SVB’s CFP listed access to the Federal Reserve discount window as a contingent funding source. It was theoretically available — SVB was an FDIC-insured institution and eligible to borrow. It had never actually borrowed. Collateral was not pre-positioned. The operational procedures for executing a discount window draw under crisis conditions had never been tested.

When the bank run hit and SVB needed to access the discount window at scale and speed, it couldn’t execute. The plumbing had never been tested.

This specific failure drove the most concrete change in the July 2023 regulatory addendum: operational testing of discount window access is now an explicit requirement. Not theoretical eligibility — operational readiness, verified through actual test draws.

5. Early Warning Indicators Didn’t Drive Action

SVB had early warning indicators in its CFP framework. By August 2022, the internal stress test results were themselves EWIs flashing significant alerts. The unrealized loss position in the HTM portfolio was an EWI. The shifting interest rate environment that was pressuring SVB’s depositor base (the startup funding market) was a market-level EWI.

None of these triggered the governance response they should have: CFP activation review, board escalation, pre-positioning of contingent funding sources. The early warning indicators guide covers the EWI governance framework in detail. The SVB lesson is that EWIs are only effective if the governance structure around them has a credible escalation path — one where triggering an EWI actually initiates a response, not a conversation about whether the EWI threshold is calibrated correctly.

Signature Bank and First Republic: The Pattern

Signature Bank closed March 12, 2023 — 48 hours after SVB — via NYDFS order. The mechanism was different: Signature had significant concentration in cryptocurrency-related deposits, and contagion from the SVB failure triggered simultaneous depositor concern. At approximately $110 billion in assets, it was the third-largest bank failure in US history at the time.

First Republic Bank failed in May 2023, approximately two months later. Together, three banks holding over $500 billion in combined assets failed in less than 90 days — all from liquidity events, not credit losses.

The pattern across all three: concentrated depositor bases, assumptions not calibrated to actual depositor behavior under stress, and contingent funding access that proved less available than the CFP assumed. Each failure reinforced the same lessons at slightly different angles.

What Regulators Changed

The regulatory response was fast. On July 28, 2023 — four months after SVB’s failure — the OCC, Federal Reserve, and FDIC jointly issued an addendum to the 2010 Interagency Policy Statement on Funding and Liquidity Risk Management. This was the first update to that guidance in 13 years.

OCC Bulletin 2023-25 and FDIC FIL-39-2023 specified three major additions:

Operational testing requirement: Institutions must operationally test their contingent funding sources — not just document that they’re theoretically available. The Federal Reserve discount window specifically: examiners now expect evidence of actual test draws, not just eligibility confirmation.

Realistic availability assessment: Contingent funding sources must be assessed for what’s actually available under stress conditions — after accounting for collateral constraints, credit facility terms, and operational execution capacity. Listing “FHLB advances: $500M available” isn’t sufficient if the FHLB line has never been drawn and collateral hasn’t been verified.

Material-change update triggers: CFPs must be updated when the institution’s risk profile changes materially — new business lines, significant changes in depositor composition, shifts in the market environment. The 2022 interest rate cycle that stressed SVB’s depositor base should have triggered a CFP review. Institutions that update their CFP only on an annual schedule regardless of intervening changes are non-compliant with the spirit of the addendum.

What Your CFP Should Do Differently

If you’re reviewing your CFP in light of the 2023 failures, five specific questions are worth working through:

1. Are your runoff assumptions calibrated to your actual depositor profile? Pull your depositor concentration data: what percentage of deposits are uninsured? What percentage are concentrated in a single industry or a small number of large depositors? If your stress scenarios use industry-standard runoff tables, test whether those tables would have predicted what actually happened at SVB given its specific composition. See what is a contingency funding plan for the baseline documentation requirements.

2. Are HTM securities excluded from your liquid asset calculations? If your CFP or liquidity stress test counts HTM securities as available liquidity, fix it now. These positions are not liquid under stress — they’re capital-destroying if sold. Your HQLA calculation should exclude them.

3. Have you operationally tested your contingent funding sources? For each listed source: when was the last time you actually drew on it? Is your collateral pre-positioned? Do your operations staff have the procedures to execute a draw under crisis conditions at 2am? The discount window specifically requires documented test draws.

4. Does triggering an EWI actually do something? Review your EWI governance: when an EWI threshold is breached, what specifically happens? Who is notified? What decision does it trigger? If the answer is “it gets discussed at the next risk committee meeting,” that’s not a credible response architecture.

5. Would your board recognize a problem? The Fed’s post-mortem cited board approval of the CFP framework while MRIAs for stress testing deficiencies were outstanding. Does your board presentation include the actual stress test numbers — deficits, survival horizons, assumption sensitivities — or does it show a summary that smooths over the underlying results? Board members can only challenge results they can see.

So What?

The 2023 bank failures aren’t an abstract case study for large-bank risk managers. They’re a documented set of CFP design and governance failures that the July 2023 regulatory addendum directly addresses — and that examiners at every institution size are now checking for.

The assumption problem is the most portable lesson: CFPs that use industry-standard assumptions without calibrating to the institution’s actual risk profile aren’t doing the job. The governance problem is equally important: stress test results that drive decisions rather than inform them create an incentive to manage the model.

The CFP regulatory requirements guide covers the full baseline of what’s required. The SVB lessons sit on top of that baseline — they’re not a replacement for it.

External sources:

• Federal Reserve: Review of the Supervision of Silicon Valley Bank (April 2023)
• OCC Bulletin 2023-25: Addendum to Interagency Policy Statement on Funding and Liquidity Risk Management
• FDIC FIL-39-2023: Addendum to Interagency Policy Statement on Funding and Liquidity Risk Management
• Federal Reserve OIG: Material Loss Review of Silicon Valley Bank (September 2023)
• FDIC: Failed Bank Information — Silicon Valley Bank