What it takes to build a profitable digital bank

Digital banks can be launched quickly, but most remain unprofitable. Global unique neobank users rose from about 146 million in 2021 to a projected 350 million by 2026, a roughly 35% compound annual growth rate. An industry analysis found fewer than 5% of challengers were breaking even in its sample. Average revenue per user at neobanks is about $45, compared with roughly $350 at traditional retail banks.

The primary commercial issue is unit economics. Many customers use a challenger as a secondary account for everyday spending while keeping savings, mortgages and investment products with incumbent banks. Those high-margin products account for most traditional bank income; challengers that lack them rely on card fees and subscriptions.

Licence choice shapes what a digital bank can sell. A full banking licence permits deposits, loans and mortgages. An Electronic Money Institution licence allows payments and issuance of electronic money but generally excludes traditional lending. Some founders launch using Banking-as-a-Service arrangements or acquire a licence-holding company to speed market entry. Jurisdictions with streamlined rules and testing environments, such as Lithuania, are frequently used early in rollouts.

Lending is the most common route to larger margins. Banks that reached profitability at scale built credit books that include consumer loans, buy-now-pay-later and mortgages. Revolut serves more than 65 million customers and had a reported valuation of $75 billion after a late‑2025 share sale; Nubank is profitable in parts of Latin America. Both expanded into credit and diversified revenue beyond interchange.

Technology must support those product ambitions. Successful digital banks use layered architectures: responsive mobile and web interfaces, client-side security, API gateways, microservices for modular operations, a central core banking system and data management and analytics layers. A security and compliance layer covers the whole stack.

Transaction data quality affects many downstream functions. Raw transaction feeds often contain inconsistent merchant descriptions, missing categories and gateway names in place of retailer names. Whether a bank standardises and enriches transaction data early determines how well personalisation, dispute handling and analytics work later.

Two recent launches illustrate the approach. Partners Banka, a Czech challenger, launched without legacy systems and used Czech BankID to let customers open full mobile accounts in under four minutes. The bank integrated national clearing so some payments settle in about ten seconds, issued single‑use virtual cards and implemented transaction enrichment at launch so merchant names, logos, locations and categories appear from first login. SwissBorg, a crypto app with more than 900,000 verified users and over $1 billion in client assets, activated transaction enrichment alongside its card product so card transactions display clean merchant information from the outset.

Customer acquisition cost is a key financial input. Referral and social channels can lower acquisition costs to roughly €26–€70, while paid channels commonly range from about €131 to €350. Some challengers use low-cost referral strategies early and begin monetising through premium tiers, transaction fees, affiliate products and interest income before scaling paid acquisition.

Founders make many of the important choices before the first customer arrives. Early decisions include licence type, channel mix, product roadmap and whether to build transaction‑feed enrichment into the core stack. These choices determine which products a bank can offer, how much it can spend to acquire customers and how well the backend supports product and support functions.