Licensing models for digital banks are converging on the same questions: capital, liquidity, operational resilience and the quality of the deposit franchise.

Licence design is a trade-off

Regulators want competition, innovation and better customer outcomes. They also want a failure regime that does not require a public bailout. Digital-bank licences attempt to balance those goals through staged authorisation, activity restrictions and higher capital or liquidity buffers during an early period.

The design choices matter. A restricted licence can lower the entry cost but limit the products a bank can offer. A full licence opens the deposit franchise but demands a credible path to profitability and a tested recovery plan. Neither model removes the core supervisory question: can this institution survive a stress event without disrupting customers?

Illustrative schematic accompanying the analysis
Original schematic prepared for this paper. Illustrative only; not a live data feed or a map of any specific market.

The deposit franchise is the test

App-based deposits can arrive quickly and leave quickly. Supervisors look for concentration, rate sensitivity, the share of balances that behave like operating accounts, and the institution's ability to fund itself when marketing spend stops.

This is where the quality of the customer relationship becomes a prudential issue. A bank whose deposits are held for a promotional rate is different from a bank that holds a customer's salary account, direct debits and savings buffer. The licence application may look similar; the risk profile is not.

Resilience is not a slide in the board pack

Operational resilience has become the central supervisory theme for digital banks. The relevant questions are concrete: what happens when the cloud region fails, when a card processor is unavailable, when a mobile network is down, or when a third-party fraud model produces a spike in false declines?

Impact tolerances should be set for the services customers rely on, not for the systems the bank finds convenient to map. Testing should include severe but plausible scenarios, and the results should change investment priorities.

Conduct risk moves faster online

Digital distribution compresses the distance between a product decision and a customer outcome. Disclosure that is technically present but buried, default settings that encourage overdraft use, or algorithms that segment vulnerable customers can produce harm at scale.

Supervisors increasingly expect firms to test customer journeys, not just documents. That means looking at how a real person experiences a fee, a decline, a collections message or a closure request.

A more mature conversation

The debate is no longer whether digital banks should exist. It is what kind of bank a digital licence should produce: a narrow utility with strong safeguards, a full-service competitor, or a platform that distributes products manufactured elsewhere.

Each model can be supervised credibly. What cannot be supervised credibly is ambiguity about which model is actually being run.

About this paper This is an explanatory working paper, not investment, legal or regulatory advice. It contains no invented statistics or quotations. Sources and scope are set out in the methodology. Corrections: editorial@bankingwebinar.com.