A large outflow tells a bank what left. The more revealing question may be what happens next: whether balances return, how quickly they return, where the money comes from and whether the customer resumes normal transactional behaviour.
A bank can lose the same amount of deposits for very different reasons. A corporate customer may move cash to meet payroll, a household may pay a tax bill, an investor may shift idle balances into a money market fund, or an uninsured depositor may leave because confidence has broken. The balance-sheet effect can look identical on the day of the withdrawal. The funding implications are not.
That distinction is pushing deposit analytics beyond the familiar question of how much money left and toward a more behavioural question: what is the recovery path after the outflow? In this article, “deposit recovery pattern” is an analytical term rather than a formal regulatory metric. It describes the timing, source and durability of balances that return after an unusually large withdrawal, together with the resumption—or disappearance—of the account’s normal payment activity.
The idea matters because the most recent evidence on bank runs has made one point unusually clear: deposit stability is heterogeneous. The FDIC’s May 2026 depositor-level study of Silicon Valley Bank, Signature Bank and First Republic Bank found that the largest and least-insured customers were much more likely to run, while fully insured retail depositors generally did not and at some points generated net inflows. The same study also found that some large business customers drew down operating-related accounts, showing that an account’s transactional purpose does not make it immune from a confidence shock.
From a balance snapshot to a funding trajectory
Traditional funding dashboards are built around balances, concentration, pricing, insurance status and contractual maturity. Those measures remain indispensable. But a point-in-time balance can hide how an account behaves through a cash-flow cycle. Two customers can each keep an average of $5 million at a bank while imposing very different liquidity demands. One may regularly draw the balance down to $1 million and refill it within three days as receivables arrive. Another may keep $5 million untouched for months and then remove the entire amount when rates or confidence change.
The first account looks volatile in daily data but may be deeply embedded in the customer’s operating cycle. The second looks exceptionally stable—until it is not. A recovery lens therefore asks whether a withdrawal is followed by recurring incoming payments, salary credits, merchant receipts, settlement flows or other activity that restores the balance. It also asks whether the returning money is the same type of funding. A deposit rebuilt through ordinary transaction flows is not economically identical to one replaced through a high-rate promotion or brokered deposits.
This is consistent with the logic already embedded in liquidity regulation. Under the Basel Liquidity Coverage Ratio framework, fully insured retail deposits can receive more favourable run-off treatment when customers have an established relationship with the bank or use transactional accounts, such as accounts into which salaries are automatically paid. The framework does not prescribe a “recovery score”, but it recognises that the function and relationship behind a deposit can matter for stress behaviour.
What the 2023 runs changed
The 2023 regional-bank failures changed the scale at which deposit behaviour had to be understood. The FDIC’s 2026 reconstruction used core-system and wire-transfer data to follow depositor activity day by day. Its findings reinforced the stabilising role of insurance but also showed how quickly concentration can overwhelm conventional assumptions. Large depositors were disproportionately likely to withdraw all or nearly all their funds, and much of the money left by wire transfer in the first days of the runs.
The implication is not that every bank should expect a 2023-style run. It is that historical averages can conceal the shape of a stress event. A funding base that normally replenishes after tax dates, payroll cycles or investment transfers can behave very differently once withdrawals are driven by solvency concerns or herd behaviour. That is why the Basel Committee’s February 2026 review of non-maturity deposit stability is cautious: recent technological and competitive changes may have influenced deposit stability, but traditional factors—especially insurance coverage and perceptions of bank solvency—still explain much of depositor behaviour.
The practical lesson for asset-liability committees is that recovery behaviour should complement, not dilute, stress assumptions. If a customer has historically restored balances within a week after every large operational outflow, that pattern is useful for forecasting normal liquidity. It should not be treated as a promise that the customer will refill the account during a confidence shock.
Why the return of money can be more informative than the withdrawal
A large withdrawal is an event. Recovery is a sequence. The sequence can reveal whether the bank remains central to the customer’s financial life. If payroll deposits, receivables and card settlements continue to arrive after a large debit, the account may still be operationally important even though the closing balance temporarily falls. If those flows migrate elsewhere, the withdrawal may be the first visible sign that the relationship itself is weakening.
This distinction is especially important in corporate banking. Operating balances often move with working-capital needs, tax calendars, acquisition payments and quarter-end cash positioning. A low balance after a known payment date can be less concerning than a smaller outflow accompanied by the disappearance of incoming collections. The same logic applies in retail banking: a customer who moves surplus savings but continues to receive salary and pay bills through the account may behave differently from one who redirects salary, direct debits and card spending to another bank.
Research on deposit structure supports the broader point that stability is shaped by more than account size. A BIS working paper on geographic diversification and funding stability finds that banks with more geographically diversified deposit bases experience lower volatility in deposit growth, particularly in demand deposits. Diversification does not measure post-withdrawal recovery directly, but it shows why the source and distribution of flows matter alongside aggregate balances.
Behavioural modelling is already moving in this direction
Non-maturity deposits are contractually available on demand, yet banks routinely model them as if parts of the balance have longer behavioural lives. An ECB working paper published in November 2025 using confidential bank data found that only about 20% of non-maturity deposits in the sample were treated as having zero maturity, while roughly 10% were assigned maturities beyond seven years. Banks with more volatile, interest-rate-sensitive and digitalised deposit bases tended to assign shorter maturities.
The same research also identified a governance problem: during the recent monetary-policy tightening, banks with more sensitive non-maturity deposits did not necessarily shorten their assumed maturities or update models quickly. That finding is relevant to recovery analytics because any measure built from historical replenishment patterns can become stale when customer incentives change. A deposit that refilled predictably in a near-zero-rate world may recover differently when alternative cash products pay materially more.
A useful recovery framework therefore needs to be dynamic. Rather than assigning a permanent label such as “sticky”, banks can observe how the same cohort behaves after different types of outflow and in different rate environments. The objective is not to convert on-demand funding into a fictitious maturity. It is to improve the institution’s estimate of when normal operating liquidity is likely to return and how much confidence it should place in that estimate.
What a recovery pattern could actually measure
The most intuitive measure is time to replenishment: how long it takes an account or customer cohort to restore a defined proportion of the pre-withdrawal balance. But the number is useful only when paired with context. A bank may compare recovery within one, seven or 30 days, then separate seasonal withdrawals from rate-driven transfers, operational payments and unexplained exits. The same analysis can track whether the restored balance comes from recurring customer activity or from a deliberate pricing intervention by the bank.
A second dimension is recovery quality. If $10 million leaves and $10 million returns because the customer receives its usual receivables, the bank has learned something about the continuity of the relationship. If the replacement comes from a new brokered deposit raised at a higher rate, the balance sheet has recovered numerically but not economically. The funding cost, concentration risk and likely stress behaviour have changed.
A third dimension is recurrence. Some deposits repeatedly leave and return around payroll, tax or securities-settlement dates. Others recover once and then decline structurally. Banks can therefore look at the persistence of the recovery pattern rather than treating one rebound as proof of stability. The most useful signal is not simply “money came back”, but whether the customer’s ordinary financial activity re-established itself after the outflow.
Current data show why aggregate deposit growth is not enough
At the system level, U.S. bank deposits have recovered strongly from the post-2022 decline. The Federal Reserve’s June 2026 Supervision and Regulation Report said aggregate commercial-bank deposits had reached a historical high of about $19.5 trillion by February 2026. It also noted that uninsured deposits had risen slightly from their recent 2023–24 lows as a share of assets and deposits, while remaining below 2022 levels.
The latest Federal Reserve H.8 release available on 28 August 2026 showed seasonally adjusted deposits at all U.S. commercial banks of about $19.49 trillion for the week ending 19 August. That headline is reassuring about system-wide funding availability, but it does not tell an individual bank whether its own growth is coming from durable operating relationships, higher-priced deposits, acquisitions or rate-sensitive balances.
Europe presents a similar distinction. The EBA’s June 2026 Risk Assessment Report says deposits remain the cornerstone of EU/EEA bank funding and that banks plan roughly 4% annual growth in household and non-financial-corporate deposits over the forecast horizon. The EBA describes those client-deposit growth assumptions as relatively ambitious and expects the sector’s LCR to decline to about 155% in 2026 while remaining well above the 100% minimum. For individual institutions, the question is therefore not only whether deposits grow, but how reliably the underlying customer flows regenerate after periods of heavy use.
Company examples show why composition matters
Recent company disclosures illustrate the difference between balance recovery and behavioural interpretation. Charles Schwab reported that client transactional sweep cash ended June 2026 at $485.7 billion, up $24.2 billion from the prior quarter-end. The company attributed the increase to typical second-quarter tax seasonality, organic growth and client asset-allocation decisions. By July, its monthly activity report showed transactional sweep cash at $476.8 billion. The point is not that Schwab’s deposits provide a model for commercial banks; it is that cash can leave and return for reasons unrelated to loss of confidence, and the explanation behind the path matters.
First Citizens BancShares offers a different example. In the second quarter of 2026, deposits increased to $173.43 billion, up $2.59 billion from March, but the composition moved in different directions: Corporate deposits increased by $4.28 billion while Commercial Bank deposits fell by $1.50 billion and noninterest-bearing deposits declined by $1.13 billion. The company also noted higher marketing expense tied largely to Direct Bank deposit promotions. Aggregate growth was real, but its funding meaning cannot be understood without looking at segment, pricing and mix.
The strongest use case is forecasting, not labelling
Banks should be cautious about turning recovery behaviour into a marketing-style customer score. The more defensible use is in liquidity forecasting, deposit segmentation and model validation. Transaction-level data can help treasury teams distinguish routine cash cycles from abnormal attrition, while relationship managers can see whether operational flows are continuing even when balances temporarily fall.
This can also improve stress testing. A bank could compare its normal-time recovery curves with stressed assumptions and ask where management has become dependent on historical refill behaviour. If a liquidity plan assumes that a large share of corporate operating balances will reappear within days, the institution should be able to explain why that assumption remains credible when rates, digital transfer speed or confidence conditions change.
The benefit for fintech providers is similarly narrow but important. Cash-flow analytics, treasury data platforms and deposit-intelligence tools can help institutions identify recurring flow patterns in near real time. The value lies less in predicting a bank run than in giving treasury teams earlier evidence that an account’s normal rhythm has changed.
The counterargument: a recovery history can create false confidence
The main danger is obvious. Historical replenishment can be highly reliable until the regime changes. A customer may have refilled an account after every large withdrawal for ten years and still leave permanently after a credit event, a pricing shock or a viral confidence scare. The Basel Committee’s 2026 literature review specifically warns against assuming that recent changes can be summarised as a simple permanent increase in deposit volatility; evidence remains mixed, and traditional drivers such as insurance and solvency perceptions remain central.
Behaviour can also be influenced by information. BIS research on the aftermath of SVB found that news about the failure increased surveyed households’ inclination to withdraw, while information about FDIC insurance and Federal Reserve communication could reassure depositors. The study on panic-driven runs and public communication is a reminder that recovery is partly a confidence process, not merely a mechanical cash-flow process.
There are governance concerns too. A bank that monitors salary credits, merchant receipts and customer transfers at high granularity needs clear limits on purpose, access and retention. Liquidity analytics should not become a pretext for intrusive customer profiling. The models also need controls against circular reasoning: a bank should not call a deposit stable simply because its own model expects it to return.
Evidence versus inference
The evidence is strong that deposit behaviour varies materially by insurance status, size, concentration, relationship characteristics, transactional use, interest-rate sensitivity and digitalisation. It is also well established that banks already use behavioural assumptions to model non-maturity deposits, and that regulators distinguish more stable from less stable deposit categories for liquidity purposes.
The inference is that post-withdrawal recovery behaviour can become a useful additional signal of funding quality. There is not, as of 31 August 2026, a universal regulatory definition of a deposit-recovery metric, a standard recovery horizon, or evidence that a single replenishment ratio reliably predicts future run behaviour across banks and jurisdictions. Any institution using such analytics should therefore treat them as a complement to—rather than a substitute for—deposit insurance analysis, concentration limits, pricing sensitivity, liquidity buffers and conventional stress testing.
Implications for banks, fintechs, regulators and investors
For banks, the practical shift is from asking only how much funding is on the balance sheet to asking how the funding behaves through time. Treasury and ALM teams can link large outflows to the customer’s subsequent transaction pattern, the cost of replacing the money and the probability that the balance will leave again. That can sharpen liquidity forecasts without pretending that deposits have a contractual maturity they do not possess.
For fintech and banking-technology providers, the opportunity is in data integration and explainability. Institutions need tools that can combine core-deposit records, payment flows, pricing history and customer-segment data while allowing model owners to see why a recovery assumption changed.
For regulators, the relevant question is likely to be model governance rather than endorsement of a new ratio. Supervisors can ask whether behavioural assumptions are timely, conservative under stress and back-tested against actual outflows. For investors, deposit growth should increasingly be read alongside mix, cost, concentration and the persistence of transaction flows. A rising balance can mask weaker funding economics; a temporary decline can mask a still-healthy primary relationship.
Conclusion: deposit stability is increasingly path-dependent
The post-2023 debate has often focused on how fast deposits can leave. The next analytical step is to understand what happens after they do. A large withdrawal can be a routine cash-flow event, a search for yield, a change in primary banking relationship or the opening phase of a run. The subsequent recovery pattern helps distinguish among those possibilities.
That does not make replenishment a new definition of “sticky” funding. Insurance coverage, solvency, concentration, pricing and liquidity buffers remain more fundamental. But as banks gain better access to transaction-level data, the shape of the balance after an outflow—how quickly it rebuilds, through which channels, at what cost and with what continuing activity—can tell management more than the closing balance alone.
References
7. European Banking Authority — Risk Assessment Report, June 2026
8. Bank for International Settlements — “Bank geographic diversification and funding stability”
10. Charles Schwab — “Schwab Reports Record Quarterly Revenue and Earnings” (Q2 2026)
11. Charles Schwab — “Schwab Reports Monthly Activity Highlights” (August 2026 release)