Revolut ranks new products against a decade of previous bets

Nik Storonsky says Revolut grew by improving financial products and obtaining licenses rather than initially optimizing for scale. He recalls starting with one to ten customers a day and says roughly 100,000 people now join daily.

The first seven years, he says, relied on word of mouth and public relations rather than serious growth spending. The original multi-currency card removed foreign-exchange fees; later, business accounts became another successful product.

Storonsky argues the opportunity was not unique to foreign exchange: a company could improve other poorly served banking products. He says Revolut now onboards around 1,500 business customers a day, with roughly 30% attributable to paid marketing.

For a new product, Revolut typically separates a small team of around ten product, design and engineering people. Anyone in the company can suggest a “new bet,” which goes through economic validation and review by people who have built previous products.

The validation examines unit economics and whether the idea is useful enough for customers to pay. If approved, the company assembles a team; Storonsky says building and launching can take roughly nine to eighteen months, including internal governance.

Products that grow quickly receive more people; weak ones may be retired, while moderate performers can be maintained. He gives Revolut Lite, a money-transfer product launched in more than 100 countries, as an exception where paid acquisition cost made the economics unattractive.

Storonsky says past mistakes made him more willing to try plausible ideas and judge execution. He underestimated the scale of Ultra, a premium account costing about €50 a month, which he says became a much larger business than expected.

Collison distinguishes word of mouth from viral acquisition built into payments. Storonsky says peer-to-peer payments helped in markets such as Ireland and Romania, but identity checks and country-specific licensing create friction that limits exponential growth.

He recalls that earlier rules allowed some small transactions without full identity checks, but says the change in requirements made viral loops harder. Collison suggests that money waiting to be claimed can still motivate onboarding; Storonsky agrees it helps but says the observed referral multiplier remained below one.

Revolut’s early UK and European environment had low interchange fees and strict credit rules, Storonsky says, forcing it to make transactional services profitable. He says the company is developing lending but wants a relatively small portion of loans on its own balance sheet and would securitize the rest.

He prefers subscription revenue to lending partly because loan losses depend on economic conditions and provisions can prove insufficient in a downturn. He compares lending to collecting option premiums for years and then potentially losing those gains in a bad year.

Collison challenges an argument based only on valuation multiples, saying a company should assess the quality of the profit itself. Storonsky responds that lending is inherently more cyclical and uncertain, whereas a subscription book can shrink without the same loan-loss exposure.

After a product launches in a test country, Revolut expands it gradually and adds it to an automated cross-sell system. Storonsky says the company compares each product’s monthly gross-profit trajectory with approximately 100 previous bets at the same age.

Products in the top quarter receive additional teams; the bottom quarter generally loses further investment, and the middle is judged case by case. Collison asks about harm from shutting a product down: Storonsky says a product can be maintained without further investment if existing customers would otherwise suffer.

The cross-sell system uses customer events and profile information to select relevant products. One example is suggesting an eSIM to a customer arriving abroad. Storonsky says the model considers conversion, product gross profit and the negative effect of sending too many messages.

Asked whether slow-starting products can become major successes, Storonsky says he rarely sees that pattern and generally trusts the early trend. He attributes Revolut’s strict comparisons to having launched enough bets to build a feedback loop.

Collison asks how to judge a product designed for engagement rather than direct revenue. Storonsky says teams estimate both direct and indirect gross profit, including effects on churn and spending. He cites the RevPoints rewards system as having small direct but high estimated indirect profit.

Moving occasional users to a primary salary account matters because it increases use of other products. Storonsky estimates that, depending on the country, 10–20% of monthly active users receive salary into their accounts.

He says younger customers are easier to acquire and can move from children’s products to adult accounts, while older people holding substantial liquid assets are harder to move from established banks. His estimate that more than half of under-35s have some Revolut account applies to many European markets, not necessarily primary accounts.

The company simplified its retention analysis from a complicated state machine to country-based cohorts and the share active after 12, 18 and 24 months. Storonsky says some cohorts show a “smile”: activity rises again as former users return, which he attributes partly to word of mouth.

Revolut first applied AI to support and translation, then expanded to engineering, marketing and experiments with an end-to-end AI employee. Storonsky thinks one person may eventually cover more product roles, but says existing small-team structures have not yet changed.

He prefers to keep proven processes while testing alternatives. Existing systems and access are organized by specialization, so moving from separate iOS, Android, backend and data roles to fewer generalists requires organizational changes, not only better tools.

For expansion, Storonsky weighs the quality of incumbent banks, customer economics and licensing difficulty. He expects direct competition with established digital banks to be harder and says the company prioritizes markets with weaker incumbents and favorable economics.

On the US, he argues that difficult licensing and partnerships separating technology from bank compliance have held back digital banks. He says Revolut has conditional approval from the OCC and still needs two further approvals; he hopes for a larger launch around the following year.

He expects the US proposition to lead with a credit card and benefits comparable to Amex Platinum and Chase Sapphire, alongside wealth and crypto services. He also says the company wants a euro stablecoin; these are plans and ambitions described in the conversation.

Storonsky schedules weekdays in 15- or 30-minute blocks for reviews and one-on-ones, reserving weekends for freer thinking. He says he spends most of his time internally on products and business processes rather than conferences.

He says meeting a KPI can reflect luck or external factors, so he looks at product quality and development speed as well. He recalls a wealth product growing 60–80% year on year that he considered capable of more; after a team change, he says growth reached 300%.

Collison asks about Storonsky’s AI investment firm. Storonsky describes a machine-learning model used to assess startups and says the firm offers founders operational playbooks, his advice and introductions to Revolut or its customers. He says the firm’s returns have been good.

Asked about an IPO, Storonsky hopes for 2028, subject to market conditions, while saying international expansion remains work in progress. He does not say that a US launch is a strict prerequisite.

He favors quarterly performance reviews over annual ones, with assessment of skills, deliverables and culture by managers and a review process. Collison challenges whether jobs such as legal work can be quantified; Storonsky points to team-level measures such as contract turnaround and deficiencies, while acknowledging that not everything can be reduced to a KPI.