Low for long: the rate forecast hidden in bank technology budgets

The week

The Jackson Hole symposium is the annual gathering of central bankers in Wyoming hosted by the Federal Reserve Bank of Kansas City. In August 2020 it was held virtually because of the pandemic, and the Federal Reserve used it as the stage for a significant shift in how it thinks about its job.

The announcement ended a long process. Powell had said in late 2018 that it was a good time to review how the Fed formulates, conducts and communicates monetary policy. The Fed described the result as the first comprehensive and public review of its framework. Through 2019 the Board and the regional Reserve Banks held a series of “Fed Listens” events with community groups, labor organizations and small businesses. A research conference at the Federal Reserve Bank of Chicago in June 2019 brought in outside economists, and the FOMC discussed the review at a run of meetings starting that July.

The problem the review set out to address was a decade old. Through the 2010s, inflation had mostly run below the 2 percent goal the Fed had formally adopted in 2012, even when unemployment was low. Estimates of the neutral interest rate, the rate that neither stimulates nor restrains the economy, had fallen. A lower neutral rate meant the Fed would hit zero more often in downturns, leaving it less room to cut. The new framework was designed for that world.

On August 27, 2020, Federal Reserve Chair Jerome Powell announced a revised Statement on Longer-Run Goals and Monetary Policy Strategy. That document is the Fed’s own description of what it is trying to achieve and how it intends to go about it. The Federal Open Market Committee, the group within the Fed that sets interest rate policy, made two important changes.

First, it said it would seek inflation that averages 2 percent over time. After periods of inflation below 2 percent, policy would aim for inflation moderately above 2 percent for some time. It works like a driver who has been going below the speed limit and makes up a little time by going slightly over it for a while: the aim is the average, not each moment.

Second, it said decisions would be informed by shortfalls of employment from its maximum level, rather than deviations in either direction. The Fed would worry about unemployment being too high, but would no longer raise rates simply because the job market looked very strong.

The FOMC approved the revised statement unanimously. It did not set out a formula. It did not say over how many years inflation would be averaged, or how far above 2 percent it would be allowed to run. That vagueness was deliberate. It gave the Committee flexibility, and it left banks, investors and businesses to infer for themselves what the change meant for the path of rates.

The policy rate, the short-term interest rate the Fed steers and which flows through to what banks pay on deposits and charge on loans, was already near zero, where it had been since March. Markets and commentators read the change as a signal that rates would stay very low for an extended period, even if the economy recovered.

The inference did not stay vague for long. At its next meeting, on September 16, 2020, the FOMC wrote the new framework into its forward guidance. It said it expected to keep the federal funds rate at zero to 0.25 percent until two conditions were met. Labor market conditions had to reach maximum employment, and inflation had to reach 2 percent and be on track to moderately exceed it for some time. Two members dissented. The median projection published alongside that statement showed the rate staying near zero through 2023.

The banking backdrop made the signal easy to absorb. The Fed had cut its target range to zero to 0.25 percent in March 2020 and reduced reserve requirements to zero later that month. Deposits were pouring into banks and credit unions. Stimulus payments arrived, businesses drew down credit lines and parked the cash, and households spent less. The Paycheck Protection Program had stopped accepting applications on August 8, and much of that loan money was still held in borrowers’ deposit accounts. Many institutions had more cheap funding than they could lend.

For most people, this was a story about mortgages and savings accounts. For banks and credit unions, it was something more: an authoritative signal about the economic conditions they should plan around. And plans made in that environment included a great many technology investments, some of which carried that low-rate assumption inside them without ever saying so.

Monetary policy rarely appears in a technology steering committee’s papers. It should. Rates shape deposit behavior, funding costs and the value of the balance sheet, and through those they shape what a bank or credit union can afford to build and run.

The August 2020 framework change mattered because it gave institutions an authoritative reason to expect years of near-zero rates. Many technology investment cases absorbed that expectation without stating it. That is the kind of risk that does the most damage: not the one everyone argues about, but the one that sits silently inside a spreadsheet.

The low-rate assumption hid in three places: deposit and funding assumptions, the metrics used to judge digital channels, and the commercial terms of long technology contracts. To see why it mattered, start with how interest rates connect to technology spending.

How interest rates reach the technology budget

At its simplest, a bank borrows money from depositors and lends it to borrowers. The difference between what it earns on loans and investments and what it pays on deposits is called the net interest margin, or the spread. For most banks and credit unions, that spread is the largest source of income. It pays for branches, staff, compliance and technology.

When rates are near zero, deposits are cheap. A bank can pay savers almost nothing and they rarely move their money, because nobody else is paying much either. Deposits in that world feel “sticky,” a word bankers use for balances that stay put. Cheap, sticky deposits make almost any business plan look comfortable.

When rates rise, the picture changes. Savers notice that money market funds or online banks are paying more. Deposits become more expensive to keep, and some leave. The bank’s cost of funds, what it pays for the money it lends, goes up, and the spread can narrow.

Technology investments are usually justified with a business case: a document that estimates costs and benefits over several years and concludes that the investment is worth making. Every business case rests on assumptions about the future: how many customers, how much activity, what revenue. If those assumptions include a particular interest rate environment, the case carries interest rate risk, whether or not anyone labels it that way.

Banks already have a body whose job is to manage interest rate risk: the asset-liability committee, often called ALCO. It looks at how changes in rates would affect the bank’s earnings and balance sheet, usually by running several scenarios. Technology decisions, by contrast, rarely pass through the same lens.

Where the rate assumption enters the arithmetic

A rate assumption enters a technology business case through several doors. The first is the revenue line. A digital deposit platform is usually justified by the balances it will gather, and those balances are valued at the spread the bank expects to earn on them. At zero rates, that spread depends on how little the bank pays, which in 2020 was close to nothing.

The second door is the deposit beta. Treasury teams use the term for the share of a change in market rates that a bank passes on to depositors. A beta of 0.3 means that when market rates rise one percentage point, the rate paid on that deposit rises about 0.3 points. Betas are estimated from history. In 2020 the recent history was a decade of near-zero rates and one short, modest tightening cycle, so measured betas were low. A low beta makes gathered deposits look valuable and loyal for a long time.

The third door is the discount rate. Business cases compare costs now with benefits later, and later benefits are discounted back to today. Many finance teams set the hurdle rate with reference to the institution’s cost of capital, which tends to fall when rates fall. A low hurdle rate flatters projects whose payoff comes in years four to seven. When rates rose, the same project, rescored, looked worse even though nothing about the technology had changed.

Accounting adds a slower fourth channel. Much of the internal development work on a large platform is capitalized under the rules for internal-use software and then amortized over several years. A program started in 2021 can still be producing amortization expense in 2025. Decisions taken when money was cheap therefore show up in the expense line long after the conditions that justified them have gone.

Deposit and funding assumptions

Deposits flooded into banks and credit unions in 2020 and 2021 as pandemic support payments arrived and spending fell. With rates near zero, those deposits cost almost nothing and looked sticky. Business cases for digital banking platforms, treasury management products and data programs often took both conditions as given.

The assumption shaped priorities as well as revenue lines. Tools for deposit pricing, intraday liquidity monitoring and interest rate risk analytics looked like low-value investments when rates were not moving. A deposit pricing tool helps a bank decide what rate to offer on each type of account, and how quickly to respond to competitors. Intraday liquidity monitoring shows how much cash the bank has available hour by hour, rather than only at close of business. Rate risk analytics model how the balance sheet would respond to different rate paths.

Asset-liability committees could run on monthly batch reports because little changed month to month. A batch report is one produced periodically from a snapshot of data, rather than updated continuously. Institutions that deferred these capabilities later found themselves pricing deposits on spreadsheets while competitors and money market funds repriced weekly.

The hidden mechanism is that a business case measured against a flat rate path never tests the scenario in which the capability is most needed. A deposit pricing engine has little value at zero and large value at four percent. A case written in 2020 would have scored it on the first condition.

The same blind spot affected resilience spending. Liquidity stress tooling, the ability to see large depositor concentrations intraday, and the operational capacity to move collateral quickly all look like insurance in a calm market. Insurance is easy to cut from a budget that assumes calm will continue.

How the exposure differed by type of institution

The same assumption landed differently depending on the business model.

Community banks generally fund themselves mostly with core deposits from local households and businesses. Their deposits were often stickier than average because many accounts were tied to operating relationships and loans. Their exposure lay less in sudden outflows than in slow margin compression, as they matched rising rates on certificates and money market accounts. For them, the deferred capability was usually pricing: the ability to pay more to the customers likely to leave without repricing everyone.

Credit unions carry a structural feature that made the low-rate assumption riskier. Many hold large portfolios of fixed-rate mortgages and auto loans funded by share accounts and certificates. When member deposit costs rise, those fixed-rate assets do not reprice. NCUA already required federally insured credit unions with more than $50 million in assets to maintain an interest rate risk policy and program. In October 2021 its board added a Sensitivity to Market Risk component to its rating system, which became CAMELS for examinations beginning April 1, 2022. NCUA cited the growth of mortgage-related assets on credit union balance sheets as one reason.

Larger regional banks had a different mix. More of their deposits came from commercial clients and wealthier households with balances above the FDIC insurance limit. Those depositors are rate-aware, and many have treasury staff or advisers watching yields. Their technology exposure was in commercial deposit pricing, concentration monitoring and the ability to see large movements during the day.

Broker-dealers and wealth managers faced a version of the same problem through cash sweep programs. Many sweep uninvested client cash into affiliated or partner bank deposits and earn a spread on it. At zero rates clients paid little attention to sweep yields. When rates rose, clients moved cash into money market funds and Treasury bills, a shift the industry called cash sorting. Firms whose client reporting and cash tools had been built for a zero-rate world had to rebuild them under pressure.

What an examiner would ask to see

Rate risk is a supervised discipline, so the evidence trail matters. Examiners reviewing interest rate risk look closely at the assumptions behind the institution’s models, especially for non-maturity deposits such as checking and savings accounts. Those deposits have no contractual end date. A model has to assume how long they will stay and how their rates will move, and those assumptions drive much of the result.

Supervisors expect those assumptions to be supported by the institution’s own data, tested against actual behavior and challenged by someone independent of the people who set them. Banks supervised by the Federal Reserve and the OCC follow model risk management guidance issued in 2011, which reaches asset-liability models along with others. An institution whose deposit data lived in a core system that produced only monthly snapshots had little to show an examiner when deposit behavior changed quickly. The same data gap that weakened pricing weakened the supervisory file.

The metrics used to judge digital channels

Cheap funding also changed competition. Fintechs and digital-only banks could acquire customers on thin margins, supported by plentiful investor capital. Incumbents responded with digital account opening, mobile features and faster onboarding. “Onboarding” is the process of turning an applicant into an active customer: verifying identity, opening the account and getting the first money in. Success was usually reported as accounts opened, application completion rates and time to fund.

Those metrics count arrivals. They say nothing about whether balances stay or what they cost to keep. When deposit costs rose, the same digital convenience that won accounts made balances easier to move. A customer could shift funds to a higher-yielding account with a few taps and no branch visit. A channel strategy measured on onboarding speed had no early warning for that.

This is an important point for any leader reading a digital dashboard. A metric is not neutral; it shapes behavior. If teams are rewarded for opening accounts quickly, they will optimize for opening accounts quickly. Whether those accounts hold balances, at what cost, and for how long, becomes someone else’s problem.

A more durable scorecard pairs acquisition with retention and margin. Useful measures include balance retention at 90 and 365 days by acquisition channel, cost of funds by cohort, and the share of new relationships that add a second product. A “cohort” is simply a group of customers who joined at the same time or through the same route, tracked together. These measures are harder to compute because they join digital, core and finance data. The core is the bank’s central system of record for accounts and balances. That join is the work a rate-neutral business case should have funded.

A surge of accounts that did not stay

Take a regional bank that launched a slick mobile account-opening experience during the deposit surge. Its dashboard showed applications completed in minutes and a steady climb in new accounts. The board was pleased. What the dashboard did not show was that many of those new accounts were opened with modest initial deposits, held a single product, and had no direct deposit or other ties to the bank.

When rates rose, those balances were the first to leave. Because the bank had never tracked retention or cost of funds by acquisition channel, it could not tell whether the digital program had built lasting relationships or simply rented deposits for a season. A scorecard built on retention and margin would have raised that question while there was still time to act on it.

What it takes to build the cohort view

The retention scorecard sounds simple. Building it is a data integration project with clear owners and a predictable sequence.

1. Tag the origin of every account. The digital platform knows which channel, campaign and promotion brought an applicant in. That tag has to travel into the core system with the account number, or be joinable later through a stable key.
2. Capture balances over time. Most cores report current balances well. Retention needs history, which usually means a daily or monthly balance snapshot landing in a data warehouse.
3. Attach a cost. Finance has to assign each cohort its rate paid, promotional costs and acquisition spend, so that a cohort’s balance can be compared with what it cost to win and keep.
4. Define relationship depth. Direct deposit, bill pay, a loan or a second account are common markers. The definition should be agreed with the business before the first report, or every review becomes an argument about definitions.
5. Report it where funding decisions are made. The scorecard belongs in the channel steering committee and in the asset-liability committee’s deposit review.

Ownership tends to split three ways. The digital team owns origin data. The data or analytics team owns the join and the history. Finance owns the cost of funds. Nobody owns the whole by default, which is one reason it was so rarely built. A single accountable executive, often in finance, keeps it from falling between functions.

The trade-off is real. Cohort reporting costs analyst time and warehouse capacity, and its early results can embarrass a program that leadership has celebrated. For a small credit union with one analyst, a quarterly report on three or four acquisition channels may be the right scale. A large regional bank should expect monthly reporting by product and channel.

When digital deposits count as something else

Some digital growth came through partners rather than the institution’s own app: fintech programs, deposit listing services and sweep arrangements. How those balances are classified matters. Brokered deposits face restrictions at banks that are less than well capitalized, and liquidity models and supervisors treat them as less stable than core deposits. The FDIC finalized a revision of its brokered deposit rules in December 2020, changing how many fintech and sweep arrangements were assessed.

For a technology leader, the implication is practical. A partner-sourced deposit program needs data on which balances came from which arrangement, under what agreement and with what classification. Without it, treasury cannot model the balances properly, and compliance cannot answer an examiner’s question about how they were treated. That tagging is cheap to build at launch and expensive to reconstruct later.

The commercial terms of long contracts

The third hiding place was in contracts. Multi-year agreements for core processing, digital banking and cloud services were often priced on account or transaction volumes that reflected pandemic-era growth. Some carried minimum commitments sized to projected growth rather than current use.

Technology leaders in 2020 and 2021 had reason to lock in terms while budgets were relatively loose. The exposure lay in provisions that assumed the moment would last. Three deserve scrutiny in any long technology agreement:

• Volume floors and minimums, which create cost when activity falls or customers consolidate. A bank that commits to paying for a certain number of accounts pays for them whether or not they exist.
• Escalators tied to inflation indices, which looked trivial when inflation ran below two percent and became material in 2022. An escalator is a clause that raises the price each year in line with a published measure of inflation.
• Termination and transition rights, which decide whether an institution can restructure its technology estate when its economics change. These clauses set out what it costs, and how long it takes, to leave or to move to another provider.

Credit unions and community banks felt this most. Their margins depend heavily on the spread between loan yields and deposit costs, and much of their technology spend runs through a few long vendor relationships. A core or digital banking contract signed at the height of the deposit surge fixed costs for years, regardless of what happened to the spread.

How the terms work, and where buyers have room

Core processing and digital banking contracts often run for five years or longer, and renewals frequently reset the clock. Early termination typically triggers fees based on a share of the fees remaining for the rest of the term, plus deconversion charges for extracting data and moving to another provider. Those provisions are standard in the market. They are also what makes a contract signed at the top of a cycle hard to revisit at the bottom.

Volume pricing creates the opposite exposure from the one most buyers worry about. Buyers focus on what happens if volumes grow and tiers become expensive. The exposure in 2022 and 2023 ran the other way. Account counts fell as single-product digital accounts were closed or went dormant, and minimums stayed fixed. A contract with a minimum sized to the 2021 peak charges for accounts that no longer exist.

Inflation escalators deserve a closer look than they usually get. Many link annual increases to a consumer price index. Some apply the full index change with no ceiling. Others apply the greater of a fixed percentage or the index. When the index ran near 2 percent, the difference between those formulas was small. When consumer prices rose 9.1 percent in the twelve months to June 2022, according to the Bureau of Labor Statistics, it became a budget line.

The negotiation levers are familiar to experienced buyers. A cap on annual escalators, a minimum that steps down if account counts fall, and a merger clause that lets the institution consolidate contracts after an acquisition all reduce the exposure. Each costs something at the table. Vendors may price the protection into the base fee or ask for a longer term in exchange. Whether that premium is worth paying is a rate and inflation judgment as much as a procurement one.

The regulatory frame differs by charter. Banks supervised by the Federal Reserve, FDIC and OCC follow the interagency guidance on third-party relationships finalized in June 2023, which expects contract terms, including termination rights, to be part of due diligence. Credit unions answer to NCUA, which issues third-party guidance but lacks the authority bank regulators have to examine technology service providers directly. That gap puts more of the contract burden on the credit union itself.

Why the assumption stayed hidden

If the low-rate assumption mattered so much, why did so few institutions write it down? The answer has less to do with carelessness than with how organizations divide up their thinking.

Different committees, different languages. Technology investments are usually approved by a technology or operations committee that thinks in terms of features, delivery dates and cost savings. Interest rate risk is managed by treasury and the asset-liability committee, which think in terms of scenarios, durations and spreads. The two groups seldom read each other’s papers. A business case could rest on cheap deposits for five years without anyone who models deposit costs ever seeing it.

The present always looks normal. When a condition has held for a while, it stops feeling like an assumption and starts feeling like a fact. Rates had been near zero since March 2020, and the Fed had just given an authoritative reason to expect them to stay there. Writing “we assume rates stay low” in a business case would have seemed unnecessary.

Benefits are easier to count than exposures. A business case naturally focuses on what the investment will deliver: more accounts, lower processing costs, faster service. Asking how the case would look under very different economic conditions is extra work, and it can make a favored project look less certain. Few sponsors volunteer for that.

Vendors price to the moment. Technology providers proposing multi-year contracts naturally base volume tiers and pricing on current growth. If the buyer does not ask what happens when growth slows, the contract will not answer.

Budget cycles and contract cycles run on different clocks. Most institutions set technology budgets annually. A five-year contract or a three-year platform program spans several budgets, but it is approved in one. The assumptions that justified it are reviewed once, at approval, and then become history. Later budgets inherit the commitment without inheriting the reasoning.

Internal pricing reflected the moment, not the cycle. Larger banks use funds transfer pricing to charge business lines for the funding they use and credit them for the deposits they bring in. With market rates near zero, that credit was small, so the internal numbers said a deposit franchise was worth little. Investments in deposit pricing and retention looked marginal on that basis. When rates rose, the same deposits became far more valuable, and the tools to defend them were missing.

These forces are not unique to banking or to 2020. They operate whenever a long-lived commitment is made under conditions that feel stable. That is why the remedy is a habit of making assumptions visible, rather than a better forecast. Nobody needs to predict the next rate cycle. They need only to ask, before signing, what the plan takes for granted without saying so.

Costs, customers, regulators and strategy

That does not make the investments of 2020 and 2021 mistakes. Many were overdue, and the pandemic made digital service essential. Customers who could not visit branches needed to open accounts, move money and apply for loans from home. Institutions that invested served them better.

The gap was that their macroeconomic assumptions were implicit. Treasury teams stress-test the balance sheet against rate scenarios as a matter of routine. Few institutions applied that discipline to the technology portfolio, even though a case that only works at zero rates carries interest rate risk in disguise.

For a board, that has several consequences. Costs locked in under one set of conditions can become burdensome under another, just as margins tighten. Customers won on convenience can leave on convenience. Regulators expect institutions to understand and manage liquidity and rate risk across the whole business rather than in the treasury function alone. And strategy suffers when the capabilities a bank needs in a rising-rate world, such as pricing, liquidity visibility and analytics, were deferred because they looked unnecessary in a flat one.

The strongest case against this argument

A fair critic would make three points. First, almost nobody forecast what happened. The Fed’s own median projection in September 2020 showed no rate increases through 2023. Asking technology teams to plan for a path the central bank did not expect could look like asking them to outguess the market.

Second, the losses that broke banks in 2023 came mainly from the balance sheet, not from technology contracts. Unrealized losses on securities bought when yields were low, and concentrations of uninsured deposits, did the damage. A volume minimum in a digital banking contract is small next to a bond portfolio that has lost a large share of its market value.

Third, scenario analysis has costs. Every new review layer slows decisions. Institutions that held back digital investment in 2020 to run more scenarios risked losing customers who could not visit branches.

Each point has force, and none defeats the argument. The case here is about testing a plan against more than one path, which requires no forecast. Treasury teams did exactly that for the balance sheet and modeled rising-rate scenarios even when they seemed unlikely. Technology commitments are smaller than bond portfolios, but they are also less flexible. A security can be sold at a loss, while a core contract often cannot be exited at any price that makes sense. And the capabilities that were deferred, such as deposit pricing and intraday visibility, were the ones that determined how well institutions could respond when the balance-sheet problem arrived. A one-page note and a second scenario cost little and need not slow anything worth doing.

A one-page note and a broader scorecard

The remedy is modest. Attach a one-page assumptions note to every major technology investment. It should state the rate environment, deposit growth and account volumes the case relies on, and the conditions under which it must be revisited. Route the note to the asset-liability committee as well as the technology committee. That puts technology commitments in the same risk picture the board already reviews, at almost no cost.

In practice, the discipline shows up in four ways:

• Every large technology business case shows how it performs under at least two or three rate scenarios, rather than the single one prevailing when it was written.
• Digital channels are reported on balances retained and cost of funds, alongside accounts opened.
• Long contracts are reviewed by finance and treasury as well as by technology and procurement, with particular attention to minimums, escalators and exit terms.
• Capabilities whose value appears only when conditions change, such as deposit pricing and intraday liquidity, are funded as core infrastructure rather than as optional extras.

Institutions that missed the risk tended to share the opposite habits:

• Assumptions buried in appendices. The rate and volume figures exist, but nobody outside the project team reads them.
• Treating technology as a cost center only. Judging spending on cost alone, without asking how it changes the bank’s ability to respond to markets.
• Celebrating arrivals. Reporting growth in accounts without tracking what those accounts are worth over time.
• Signing for the boom. Accepting volume commitments based on the best year in a decade.
• Cutting the insurance. Removing resilience and analytics capabilities from budgets because the conditions that would make them valuable seem distant.

Running the discipline: owners, cadence and evidence

The note needs an owner on each side. The sponsoring executive writes it, because the sponsor knows what the case assumes. A treasury or finance analyst reviews it, because that person knows which assumptions are fragile. The secretary of the asset-liability committee logs it, so the committee can see the stock of technology commitments alongside its other exposures.

A threshold keeps it proportionate. Most institutions would apply it only to investments above a set multi-year value, or to any contract with minimums, escalators or a term longer than three years. A community bank might set the line low enough to catch its core and digital contracts and little else. A large bank would set it higher and apply it across many programs.

Triggers make the note useful after approval. Each note names the conditions under which the case gets a second look: a move in the policy rate beyond a stated range, deposit growth well below plan, or account counts falling under a contract minimum. When a trigger is hit, the sponsor and finance rescore the case and report the result to the committee that approved it.

The evidence trail matters to examiners and auditors. A folder of notes, rescoring memos and committee minutes shows that management understood the assumptions behind its commitments and revisited them. That record supports the institution’s interest rate risk and third-party risk programs at the same time.

The trade-off is administrative load. Sponsors will see the note as another approval hurdle. The answer is to keep it short, standardize the scenarios so nobody invents their own, and let finance supply the rate paths. Done that way, it should add hours to a major decision, not weeks.

Rate scenarios to run before signing a long contract

• For each of our largest technology investments, what interest rate, deposit growth and volume assumptions does the business case rely on, where are they written down, and has our asset-liability committee ever seen them?
• If rates rose sharply, which of our technology commitments would look most different, and which capabilities, such as fast deposit repricing or a view of cash and large depositor balances during the day, would we suddenly wish we had?
• For customers acquired through our digital channels, how much of their balance is still with us after 90 days and after a year, and what does it cost us to keep?
• Which of our technology contracts include volume minimums, inflation escalators or costly exit terms, and when do they come up for renewal?
• Does finance run our technology portfolio through the same rate scenarios we use for the balance sheet?

The rate paths need not be elaborate. Three are usually enough: the path the market currently prices, a path where rates rise faster and further, and a path where they fall. Running a technology commitment through all three shows which assumptions carry the case. If a contract or program works under only one of them, the institution should either change the terms or accept the exposure deliberately, with the board aware of it.

Answering these questions requires no forecast of where rates are going. It requires only the recognition that rates will move, and that a technology plan which works under one set of conditions should be tested against others before it is signed.

With hindsight

Inflation rose sharply in 2021 and 2022. The Federal Reserve began raising rates in March 2022 and lifted the target range above 5 percent by mid-2023. Deposit costs climbed and balances moved to higher-yielding alternatives. The low-for-long world that so many plans had assumed ended faster than most of those plans expected.

The sequence moved faster than any 2020 business case allowed for. Consumer prices rose 9.1 percent in the twelve months to June 2022, according to the Bureau of Labor Statistics. The Fed raised its target range by a quarter point in March 2022 and then by larger steps through the year. By July 2023 the range stood at 5.25 to 5.50 percent, the highest in more than two decades.

Deposits followed. The FDIC’s Quarterly Banking Profile showed industry deposits falling for several consecutive quarters from 2022, and for the first quarter of 2023 it reported the largest quarterly decline since its data began in 1984. Unrealized losses on banks’ securities, mostly bonds bought when yields were low, were about $620 billion at the end of 2022, according to remarks by the FDIC’s chairman in March 2023.

The failure of Silicon Valley Bank in March 2023 showed, in extreme form, how unhedged rate exposure and fast-moving, digitally connected depositors could combine. “Unhedged” means the bank had not protected itself against the possibility that rising rates would reduce the value of its holdings. When depositors grew worried, they could move money quickly through digital channels, and they did. The same convenience that banks had spent the low-rate years building turned out to cut both ways.

Signature Bank failed two days after Silicon Valley Bank, and First Republic was closed and sold in May 2023. The Fed created the Bank Term Funding Program in March 2023, lending against eligible securities valued at par. It stopped making new loans on March 11, 2024. In July 2023 the federal banking agencies and NCUA updated their liquidity guidance, stressing tested contingency funding plans and operational readiness to borrow from the discount window, including pre-pledging collateral. NCUA pointed credit unions to the Central Liquidity Facility as their backstop.

Supervision moved toward the capabilities this article describes. NCUA’s new sensitivity rating took effect in April 2022, just as rates began to rise. The interagency third-party guidance of June 2023 made contract terms and exit planning a firm part of due diligence. Supervisory attention through 2023 turned to uninsured deposit concentrations, collateral readiness and the speed of deposit outflows.

The 2020 framework itself did not last. After a review concluded in August 2025, the Fed dropped flexible average inflation targeting. That ended the idea of deliberately letting inflation run above 2 percent after periods below it, and it removed the emphasis on employment shortfalls. The policy that had seemed to promise years of calm was itself replaced within five years.

Rates came down again without returning to zero. The Fed began cutting in September 2024. After three quarter-point cuts between September and December 2025, the target range stood at 3.50 to 3.75 percent. Institutions that had built pricing and liquidity tools in 2023 found them just as useful on the way down, when the task became lowering deposit rates without losing balances.

The core claim, that technology plans carried an unstated rate view, has held up well. Less settled is how much of the 2023 deposit speed came from digital channels specifically. Silicon Valley Bank’s depositors were concentrated, largely uninsured and connected through the same networks, and many community banks with modern digital banking saw no run at all. Digital convenience accelerated outflows where the depositor base was already fragile. It did not create the fragility.

For bank technology leaders, the durable conclusion is that deposit analytics, liquidity monitoring and pricing tools are core capabilities in any rate environment. They allow an institution to see and respond when conditions change, which they always eventually do. An institution that has these tools can reprice deposits quickly, spot concentrations of large balances and explain its position to regulators with evidence rather than estimates.

Technology business cases deserve the same rate scenarios the balance sheet gets. For boards, that is a simple and inexpensive discipline: ask what each major investment assumes about the economy, and make sure someone who manages interest rate risk has read it.