A digital bank survives when the money one customer generates over their lifetime exceeds what it cost to win and keep that customer. It keeps refinancing when the reverse is true at scale. That is the whole test, and most challenger-bank failures trace back to losing it quietly for years.
The discipline behind that test is unit economics: the profit or loss on a single account, treated as the basic building block of the business. The word matters here. Cambridge Dictionary's business usage defines a unit as a single item counted on its own, and that is exactly the framing — one customer, one account, one month — before any talk of scale. A bank with a million customers and a bad per-customer model is a million small losses wearing a growth story.
This piece walks through the three cost lines that decide the outcome, then assembles them into a simplified model any reader can apply to a challenger bank's public claims. No spreadsheet required. Just a habit of asking what one customer is worth, and what one customer costs.
What exactly is unit economics in banking?
Unit economics is the accounting of one customer as if they were the whole company. Revenue per account minus the direct cost of serving that account, over the time the account stays open. If the number is positive and stable, growth compounds value. If it is negative, growth compounds the deficit, and the gap gets filled with new funding rounds rather than operating profit.
The mechanism sounds obvious, yet challenger banks have repeatedly grown first and priced later. Free accounts, cash bonuses and slick apps win sign-ups fast. The question that survives a Tuesday afternoon is duller: after the app is built and the customer is onboarded, does that account pay for itself this year, and how many years until it has repaid its own acquisition cost?
Traditional banks answer with the deposit spread. As our explainer on how banks price deposits and turn the spread into profit covers, a bank pays depositors one rate, lends or invests the funds at another, and keeps the difference. A digital bank plays the same game, usually paying more on savings because it has no branch network to fund — a trade-off covered in why online banks pay more on savings than branch networks. Higher deposit costs mean the spread is thinner, which makes every other cost line matter more.
Which cost lines decide digital bank profitability?
Three lines dominate, and they behave very differently as a bank grows.
Customer acquisition. Marketing, referral bonuses, paid partnerships and onboarding costs land before the customer produces anything. This is the most visible line and the easiest to inflate. A bank can cut acquisition spend and watch growth stall, or keep spending and watch losses widen. The honest question is not whether the spend buys customers — it usually does — but whether the customers it buys stay and generate enough revenue to repay it.
Compliance and risk. Know-your-customer checks, fraud monitoring, sanctions screening, suspicious-activity reporting and the staff to run them. These costs are largely fixed per institution but scale in steps as volume grows, and they are non-negotiable. A bank that underinvests here does not save money; it defers the bill to a consent order, a fine or a forced exit from a market. Compliance is also where unchartered neobanks carry extra structure: as our primer on how neobanks without charters move and insure customer money explains, running deposits through a partner bank adds a sponsor-fee layer that a chartered bank does not pay.
Per-account servicing. Core-system licences, payment processing, customer support, card issuing, disputes and chargebacks. Some of this falls with scale, some does not. Support tickets per customer are a quiet tell: a bank whose customers file disputes and call support constantly is paying a servicing tax that a cleaner product avoids. Payment rail choice feeds this line directly, since each rail carries its own cost profile — the breakdown in wire vs ACH vs RTP and what each rail costs banks shows how routing decisions show up in unit costs.
How do you build a simple per-customer model?
Strip it to four inputs. You can estimate them from any challenger bank's published figures, or test the ones a company volunteers in its own announcements — which are the attributed source of those claims, not independent verification.
- Acquisition cost per funded account. Total marketing and onboarding spend divided by accounts actually funded. Not sign-ups. Funded accounts.
- Annual revenue per account. For a deposit-led model, mostly net interest margin on average balances, plus card interchange and any subscription fees. Thin average balances make this line small no matter how many accounts exist.
- Annual servicing cost per account. Support, processing, compliance share and infrastructure. Compliance share is the line most casual models omit, and it is often the one that flips the sign.
- Retention. How long the average funded account stays open. A bonus-chasing customer who leaves after the promo period can have a negative lifetime value even with decent monthly revenue.
The model then reads simply: annual revenue per account minus annual servicing cost gives the yearly contribution. Divide acquisition cost by that contribution to get payback time. If payback runs longer than typical customer tenure, the model loses money on average, and scale makes it worse, not better. If contribution is negative from month one, every new customer deepens the hole.
Two honest caveats. First, these inputs are estimates, and companies disclose them unevenly; a model built on a marketing deck's own figures inherits the deck's optimism. Second, the mix matters: a bank whose customers carry high average balances and use lending products can run a very different model from a free-debit-app with the same brand recognition.
What this means for judging a challenger bank
Our analysis: the single most useful habit is to convert every headline metric back to a per-customer figure. "Millions of accounts" is a scale claim. "Revenue per funded account above servicing plus amortised acquisition cost" is a viability claim. Only the second one tells you whether the next funding round is growth capital or a lifeline.
Watch for the patterns that precede trouble. Acquisition spend rising faster than revenue per account. A free-everything product with no visible second revenue line. Compliance costs described as an afterthought in investor letters. A deposit base concentrated in promotional rates that will reprice. None of these is a verdict on its own, but together they describe a business buying time rather than building margin.
The reverse pattern is just as legible. A bank that discloses its per-account costs, shows a payback period shorter than its customer tenure, and funds compliance like a core function is running a real business — whatever its app looks like. The unit economics, not the interface, decide which kind of company a digital bank is. That is the durable answer to whether it survives: survive on per-customer contribution first, and let scale amplify whatever sign that number already carries.
