Digital Banking Lowers the Branch Cost and Raises the Run Speed.

The digital bank has made a genuinely useful discovery.

A branch is expensive.

It wants rent, people, cash, queues, signs, a parking arrangement, and eventually somebody who knows why the printer is unhappy.

An app wants none of these at the point of sale. It can offer an account in a three-minute journey, bind a device, accept disclosures, and create an empty account before the customer has finished the bus ride home.

That is a real saving. It is not a new law of funding.

The branch never made deposits stable by possessing carpet. The app does not make them unstable by possessing a thumbprint button.

What changes is the geometry of the relationship.

Acquisition becomes cheaper to scale.

Payments and balances become more observable.

Rate comparison becomes trivial.

And withdrawal becomes a social event, at exactly the speed at which a customer can authenticate, choose a recipient and press confirm.

So the retrospective question is not whether digital banking has defeated branch banking. It has defeated a very great deal of branch cost.

The useful question is whether that saving in physical distribution has been carried correctly through the acquisition cohort, the customer’s funding behaviour, the bank’s cost of deposits, and the liquidity arrangements expected to look reassuring once the group chat develops a view.

The Cheap Account Has Not Yet Deposited Anything

Customer-acquisition cost has acquired the status of a flattering headshot.

A campaign costing $600,000 that opens 20,000 accounts has a cost of $30 per opened account. If only 8,000 of those receive a qualifying payroll deposit, the same campaign cost $75 per funded account, before any later fraud, incentive or servicing expense.

Both calculations are correct.

Neither of them is your customer-acquisition cost.

The denominator has simply stopped being polite.

The numerator needs the same discipline.

Media, creative production, referral rewards, affiliate commission, app-store work, sales salaries, identity checks, fraud screening, promotional yield and onboarding support all belong somewhere in the fully loaded figure.

A delightful cost per download tells management roughly what a delightful cost per brochure once told the branch network.

It can be useful.

It is not the thing being funded.

The expensive test starts after approval.

A salary credit turns an empty account into a customer deposit liability, and exposes a payer, an amount and a cadence. That may be the beginning of a primary relationship.

It is not proof of one.

The institution still has to learn whether the customer pays bills, uses the card, keeps a balance, receives the next salary, and stays once the welcome offer has stopped being a welcome offer.

Which is why Chime’s disclosed 2024 sales-and-marketing expense of $519.8 million, and its 1.4 million increase in active members to 8.0 million, are instructive without yielding a true acquisition cost.

Marketing also serves existing members.

Channels convert differently.

An active member is not an opened account.

A firm can have an efficient funnel and a rather expensive relationship. The two propositions are entirely capable of sharing a quarterly report.

The Deposit Is Not an App Metric

The first sensible distinction is between an account relationship and a funding relationship.

A customer who opened an account to collect a referral reward has given the firm a record.

A customer who routes salary, keeps card credentials at merchants, pays recurring billers, uses savings goals, receives alerts, trusts support and understands deposit insurance has given it operating connections.

Those connections raise the cost of leaving.

They do not repeal the right to leave.

Measure your evidence by cohort.

Take the customers receiving a first salary payment in January. Then observe how many are still receiving salary in April, the median balance retained after ninety days, and the outflow after an advertised-rate reset.

A large promotional inflow that departs after sixty days is funding, certainly. It has a different funding value from a modest monthly household float that stays after a rate cut.

One rented a rate. The other may have adopted a routine.

The rate is not a side issue.

Deposit beta records how much a bank raises what it pays as market rates rise.

A customer on a promotional yield is economically attractive only while the spread survives both that paid rate and the expected withdrawal date.

The account remains active throughout, in the same way a departing guest remains technically at dinner until the taxi door closes.

Monzo’s 2025 annual report supplies a useful picture of the ambition: £16.6 billion of customer deposits, up 48%, and more than a million customers paying for a subscription.

Those figures do not prove permanence.

They show a bank trying to put repeated utility and paid service around a balance, rather than competing only for a savings-rate glance.

A better design.

Still a design requiring observation.

The Six Tests Before Calling It Funding

Deposit stability is not a personality assessment performed on your aggregate balance.

It is six separate tests, and the list is complete: customer concentration, rate sensitivity, uninsured share, channel concentration, withdrawal speed, contingent liquidity.

Customer concentration asks whether a few customers can move the funding position before the management pack has acquired its title page. A total can conceal a handful of large business accounts or a single source cohort.

A bank funded four-fifths by deposits and one funded four-fifths by short-term wholesale liabilities have matched only at the coarsest and least useful level.

Rate sensitivity asks which customers rented the rate and which embedded the account. Compare balances after rate changes, advertised-rate resets and competitive offers.

Transaction accounts are cheap and operationally sticky, and redeemable on demand. Time deposits have stated maturities and commonly reprice at a visible market rate.

That difference is not semantic.

It is the future cost of keeping the dollar.

Uninsured share asks which balances behave differently once confidence weakens. It belongs beside estimated insured and uninsured deposits, not inside a generic retail label.

A balance is not made retail by having been acquired through a phone. The product, the customer, the amount, the insurance treatment and the observed behaviour all get a vote.

Tiresome, and better than being surprised later.

Channel concentration asks whether one acquisition route carries too much of the base.

One payroll platform.

One price-comparison site.

One employer.

One device population.

One social network.

Employer payroll channels and member referrals may improve qualification and lower media spend.

Price-comparison traffic delivers rate-sensitive savers.

Channel is both a distribution cost and a prospective behaviour field, and it should be kept as both.

Withdrawal speed asks what the customer and the payment system can do in the time available.

An instant transfer runs through device authentication, balance and limit checks, a reservation, rail submission, interbank settlement and customer-ledger postings. FedNow is designed for processing around the clock, every day of the year.

The app’s sent label may mean accepted, submitted or received, and none of those states should be casually promoted into the others. Even so, the sequence tells a treasury team something important.

The customer can start it at any hour.

Contingent liquidity asks what else can claim cash while deposits are leaving. Commitments, letters of credit, derivatives, guarantees, securitisation exposures and unused credit lines are not funded loans, and they can all become funding demands.

Deposit stress without those claims is not stress.

It is a careful study of one door.

The Lower Branch Cost Reappears as a Funding Cost

The digital model can reduce distribution expense and still produce an expensive liability.

The relevant funding record separates noninterest-bearing from interest-bearing deposits; transaction accounts from savings and time deposits; brokered and reciprocal deposits from the rest; estimated insured from uninsured balances; domestic from foreign; and all of it from federal funds purchased, repurchase agreements, Home Loan Bank advances and other borrowed money.

Total deposits is a fine headline and a rather poor answer.

Add maturity or repricing buckets, paid rate, concentration and source cohort. Then calculate the cost of deposits against the same reporting perimeter as net interest income and average earning assets.

A net interest margin can fall while loan yields rise, if deposit and borrowing costs rise faster.

Digital distribution does not alter that arithmetic.

It merely lets the repricing conversation arrive in the same pocket as the weather.

The commercial bargain shows up in the rest of the economics. Chime’s disclosed purchase volume and average revenue per active member of $245 sit alongside a cost of revenue and a transaction-and-risk loss line, with interchange supplying 76% of revenue.

The economical branchless customer still generates card, processor, bank-partner, cloud, fraud and dispute costs. The app is a distribution channel, not a monastery.

Chime also disclosed a healthy transaction margin alongside an adjusted EBITDA close to zero. Direct revenue less direct transaction cost is not profit after acquisition, technology, operations, risk and corporate expense.

Nor is it an assurance that a promoted deposit is cheap funding. A firm can improve service costs, acquire deposits efficiently, and still discover that retention economics have invited the finance function to dinner.

The Safety Net Has a Legal Perimeter

Digital banking is unusually good at making the brand appear to be the bank.

Sometimes it is.

Sometimes a partner bank holds the actual account while the digital brand supplies the interface and the relationship.

In the second case the partner’s customer-information file needs everything required to hold that account.

A failed asynchronous hand-off can leave the app announcing open while the routing number cannot receive funds, which is a compact demonstration of the difference between a screen and an institution.

The same discipline applies to the funding analysis.

A bank call report is a bank balance sheet, income statement and supporting schedules. A holding-company report consolidates the parent organisation, and can include a bank, a broker-dealer, an insurer and an asset manager.

That is appropriate for group complexity.

It is not appropriate for being appended to a bank-charter deposit ratio after lunch.

The FDIC’s Summary of Deposits measures branch-office deposits once a year.

It remains genuinely useful for physical market share.

It is not a census of online customers, and the absence of a branch is not the absence of a funding relationship.

Digital deposit gathering gives a bank national reach with few branches. It also gives it a national population whose response to a rate, an outage or a rumour is far less constrained by geography than the old branch map was designed to suggest.

So the safety net is not the app’s calming colour palette. It is the deposit-taking entity’s records, liquidity, capital, funding mix, payment access, partner arrangements, and ability to turn available resources into settlement money when customers act at the same time.

The interface can tell a customer a payment is complete. It cannot make you liquid by force of typography.

Who Got the Convenience

Customers did, through rapid account opening, notifications, payments, and the ability to send a rent contribution without arranging a branch visit.

That is not trivial.

A service that keeps accurate records, makes balances legible and processes payments at any hour makes ordinary financial life considerably less ceremonial.

Efficient entrants did, by exchanging branch density for software distribution. So did incumbents, wherever digital servicing cut the cost of routine activity.

A larger organisation spreads technology and control costs over more customers, and can support specialist product teams.

The saving is genuine.

So is the new obligation to govern device binding, identity evidence, velocity checks, review queues, payment reservations, trace identifiers and replicated balance data. The queue has left the lobby and joined the architecture.

Some incumbents also acquired a curious comparison advantage.

Branch deposits, local deposit share, customer location and national digital reach are four separate measures.

A bank whose old branch map looks modest may now hold a very large acquisition surface.

Charming for distribution.

Complicated for supervision.

Who Paid for the Convenience

Firms paid, through the retention economics.

Promotional yield.

Identity and fraud controls.

Customer support.

Payment and processor charges.

Data storage.

Partner-bank fees.

Dispute reimbursement.

Risk losses.

And the patient work of keeping a primary relationship primary.

A dormant account and a weekly bill-paying account should not be averaged into one reassuring engagement measure, however attractive the dashboard palette.

Supervisors paid, by having to watch funding that can reprice, compare and move at a pace no branch traffic ever revealed.

The supervisory task is not to decide whether a digital bank is modern.

It is to insist that observed data stay observed: source, reporting date, unit, currency and entity perimeter beside every material field; raw dollars beside ratios; customer and channel concentration beside the aggregate; contingent claims beside funded loans.

Weaker deposit franchises paid because their customers can leave before breakfast.

Which is not an indictment of convenience.

It is a statement about on-demand liabilities in a system where the cost of discovery and action has collapsed.

The digital bank lowers the branch cost and raises the run speed.

It can still produce durable, useful funding.

It simply has to earn it, price it, segment it, and hold liquidity against the morning the customer decides the app has become an exit.

Treat a digital deposit as funding only after you have measured how quickly the customer can decide it is not.