Nik Storonsky: The most uncompromising banking CEO
- Revolut scaled by making familiar banking products materially better before spending seriously on acquisition. Storonsky says daily customer intake rose from 1–10 to about 100,000, with the first seven years “purely organic”; serious growth budgets began around 2022. Business accounts demonstrate repeatability beyond the original zero-FX-fee card: approximately 1,500 businesses join daily, with only 30% coming through paid marketing.
- Product development runs as a portfolio of bets whose early gross-profit curves determine further investment. Revolut benchmarks launches against roughly 100 historical bets, scales the top quartile and redeploys teams from the bottom quartile. “Trend is your friend”: Storonsky can barely recall a slow starter becoming a breakout, while Revolut Lite shows that even broad distribution cannot rescue acquisition economics that require uneconomic performance marketing.
- Revolut wants subscription revenue with tightly limited retained credit risk. Storonsky targets lending exposure below 10% of the balance sheet, securitizing the rest. Collison challenges his preference for technology multiples over banking multiples, prompting the substantive explanation: credit provisions depend on uncertain future repayments, and a downturn can erase years of profits. Lending is “similar to selling options”—collect premiums for 10 years, then lose everything in year 11.
- Salary deposits remain the critical conversion from occasional utility to a broader banking relationship. Salary accounts represent 10–20% of monthly active users, depending on the country, while Storonsky claims any-account penetration above 50% among under-35s in a majority of European markets served. Switching adults is “a very slow burner,” so children’s accounts provide a route into adult banking; older customers remain harder to capture despite holding most liquid assets.
- AI could make product teams much smaller, but Revolut has yet to reorganize around that possibility. Storonsky thinks a team could probably be two people, but says teams still operate within “the old framework,” with access and processes divided among specialists. He thinks a decade of hiring, KPI and performance-management systems might become redundant in 2–3 years, but wants evidence that replacement processes work before changing the model.
- The US strategy combines ownership of the regulated stack with a premium credit-card proposition. Storonsky reports conditional OCC approval, two approvals outstanding and a hoped-for launch around next year. Revolut intends to confront Amex Platinum and Chase Sapphire with “the same benefits or better,” supplemented by its wealth, crypto trading and other products.
- Storonsky hopes for a 2028 IPO, conditional on favorable markets, while prioritizing the expansion roadmap. He wants stronger evidence that Revolut’s European success translates internationally, especially into credit-led markets, although he concedes a US launch is unnecessary for going public. His framing is operational rather than valuation-led: “Let’s focus on the roadmap.”
1. Better products carried growth before marketing did
Storonsky rejects the premise that rapid scale was his original optimization target: “I was just working on the product and then solving problems end-to-end.” Daily intake progressed from 1–10 customers to about 100,000. The first seven years relied on word of mouth and PR; he recalls serious growth budgets beginning around 2022, alongside formal customer economics and referral marketing.
The initial product was a multi-currency card with zero FX fees. Collison sees FX fees as a particularly powerful acquisition wedge; Storonsky agrees tentatively, but argues the starting category was not special. In his view, almost any banking product could have been a wedge if Revolut made it much better and cheaper. Business accounts are his repeatability example: they grew from about 10 daily customers without marketing to approximately 1,500 daily sign-ups, with only 30% now coming through paid marketing. The product philosophy is renovation rather than invention: “We haven't built any product which worked, which wasn't an old product built badly by other organizations.”
Collison argues payment loops can spread within countries, like Facebook spreading campus by campus. Storonsky counters that KYC creates steep drop-off. Low-value onboarding without KYC once supported virality, but he recalls that allowance ending around 2017. P2P helps in Ireland and Romania; nevertheless, every growth loop Revolut has tried over the last 12 years has had a K-factor below 1.
2. New products must earn their place on the curve
A proposed bet first passes validation of unit economics and financials, then review by people experienced in new bets and general managers. Storonsky says many ideas make economic sense because a small team’s cost can be low relative to the potential value—tens of millions of dollars if useful to customers. Approved bets get a separate small team. From the decision to launch, the process usually takes 12–18 months, including hiring, development and internal approval.
Collison asks why virtually any 5–10-person project would not work economically with millions of customers. The counterexample is Revolut Lite: launched in 100+ countries without banking or e-money licenses as an international Venmo, it failed to grow organically fast enough. Storonsky says they expected distribution costs to be much lower; performance-marketing paybacks did not work.
Past errors changed Storonsky’s confidence in his own judgment: “I just had too many failures in the past when my judgment was wrong, so now I'm fully optimistic.” A strong team can make an average idea work with sufficient effort and time. Ultra surprised him: the subscription costs about €50 monthly, he thinks, and became very large despite his initial expectation that it “probably won't be very large.”
Launches start in a test country, expand geographically and enter automated distribution. Against roughly 100 historical bets, Revolut compares gross profit at equivalent months after launch: top-quartile products gain teams; bottom-quartile teams are redeployed; middle outcomes receive discretion. Products remain maintained if closure would harm customers. Slow-burning successes are exceptionally rare in his experience: “Usually trend is your friend.”
Machine learning selects offers using customer events and profiles; an LLM acts as an “account manager.” A UK customer landing in Ireland and spending with Airalo might receive an eSIM offer. Revolut sees transactions, not bill details. Selection weighs conversion, gross profit and a penalty for messaging customers. Direct and indirect profit both count: RevPoints has little accounting gross profit but substantial indirect gross profit.
3. Credit brings a risk profile subscriptions avoid
Britain and Europe forced Revolut to make its transactional business profitable: interchange is low and credit heavily regulated. Had it started in Brazil or Mexico, Storonsky says it would have focused on credit instead. Deposits now enable lending, but his target is retained lending exposure below 10% of the balance sheet, securitizing the rest.
Storonsky initially contrasts technology valuations of 40–50 times earnings with banks being lucky to receive 10 times. Collison’s pushback is substantive: “The business is good or it isn't.” Investors once disliked payments businesses too; a low prevailing multiple should not determine strategy. Is credit intrinsically a worse profit pool, or merely misunderstood?
Storonsky’s answer is repayment uncertainty: provisions depend on models and economic conditions, while loans may not return for three to five years. Unemployment rising from 3% to 6% in two to three years can eliminate profits and produce losses. Collison contrasts subscription cancellations, which shrink revenue without that credit loss. Storonsky agrees: lending resembles selling options—collect premiums for 10 years, lose everything in year 11.
4. The salary account remains the harder conversion
Salary deposits unlock broader product use; loading $20 for a P2P transfer is much less profitable. Salary accounts comprise 10–20% of monthly actives, depending on country. Storonsky claims over 50% of under-35s in a majority of European markets served have some Revolut account. Older customers, holding most liquid assets, are harder to move. Children’s products provide a bridge into 18+ accounts because switching adults is “a very slow burner.”
Revolut abandoned complicated state-machine funnels for country-specific cohorts tracked at 12, 18 and 24 months. Storonsky describes a retention “smile”: out of 100 customers, perhaps 60 are monthly active initially, rising to 65–70 after two years. He calls this “negative churn,” attributing returning customers to a growing share of the population using Revolut and to churned customers hearing about it from friends and family.
5. AI puts the operating playbook back under review
AI adoption began around two years ago with support and translation; it now spans engineering, marketing and experiments with an end-to-end AI employee. Storonsky says the people-based systems he spent a decade optimizing—KPIs, roadmaps, hiring and performance management—might all become redundant in 2–3 years.
Collison presses the staffing implication: why retain conventional product teams? Storonsky thinks one developer could cover frontend, backend and possibly data science; a team could “probably” be two people. Actual team sizes have not changed. Specialized systems and access remain obstacles, and he prefers continuing what works while testing replacements before reorganizing.
Storonsky spends 95% of his time internally. His contribution includes looking beyond successful KPIs, which can reflect luck or external conditions. Wealth and trading grew 60–80% annually two or three years ago; he thought it should grow 300%. Digging deeper exposed weak features and quality. Revolut changed the team; he says growth is now 300%.
Performance reviews occur quarterly because managers can be biased against exiting underperformers, and annual reviews delay action. Assessment combines skills, deliverables measured through roadmaps and KPIs, and culture, with managerial and additional review. Even lawyers can have contract-turnaround SLAs and deficiency metrics, though he is unsure about individual coverage. His limit matters: “Can you quantify everything? No.” Some work needs other leading indicators.
6. US entry depends on licenses and premium credit cards
Market selection combines weak incumbents, attractive customer economics and ease of licensing; interchange and credit-rate rules matter. Collison challenges why Europe, poor on those criteria, produced Revolut. Storonsky estimates only three or four of roughly 50 digital banks survived: difficult conditions, he argues, made Revolut fitter for expansion. Strong digital-bank competition makes a market less attractive.
Storonsky attributes US neobanking disappointments partly to separating technology companies from partner banks’ compliance organizations. His prescription is ownership through licensing and technology-driven compliance. He says the regulatory process can now take a year, versus potentially four or five previously. Revolut has conditional OCC approval, with two further approvals outstanding, and hopes to launch around next year.
The US proposition must be credit-card-led. Asked directly about Amex Platinum and Chase Sapphire, Storonsky says Revolut will offer equal or better benefits alongside wealth, crypto trading and its broader platform. His separate euro-stablecoin rationale is brief: “It's good business,” and he sees no credible euro equivalent to dollar stablecoins. The technology was built “a long time ago.”
7. Investment access requires value, while IPO timing requires markets
Collison corrects the “AI hedge fund” description: this is a venture firm. Storonsky says a feature-based machine-learning model trained on startup data has supported 40–50 deals over 2–3 years, with returns in the global top 3–4% of funds. Collison asks how identification translates into access. The pitch to founders combines Revolut operating playbooks, Storonsky’s advice and help selling into Revolut or its customers.
With 2026 as his reference point, Storonsky hopes to go public in 2028, “subject to market being good.” He wants US entry, more expansion markets and proof that European success can travel, especially into credit-led markets. Collison challenges whether the US is necessary given Revolut’s existing scale. Storonsky concedes it is optional: the objective is completing the roadmap rather than optimizing IPO valuation or readiness.
Full transcript
Do you want to go head-to-head, Amex Platinum and Chase Sapphire?
Yeah.
How will you compete?
Nik Storonsky is the co-founder and CEO of Revolut, one of the world's leading neobanks. They started with foreign exchange, but now do everything from bank accounts to business services to mobile plans. And they're expanding to the United States.
Cheers.
Cheers.
As you look back on it, how did Revolut get to scale? What worked and caused growth?
To be honest, I was never optimizing to get scale quickly. I was just working on the product and solving problems end-to-end. Initially, when I started, we had probably 1–10 customers per day. Then it went to 100 customers, then to 500, and then to 1,000. Now we have about 100,000 customers joining every day. I was never trying to optimize to get as many customers as possible. I was more focused on building the product, gaining licenses, and so on.
But you presumably thought a lot about distribution dynamics and how to get customers to refer other customers and things like that. The growth, consumer growth side of things?
Yeah, we actually started doing growth, I think, in 2022. That was the first time we started having serious budgets for growth.
But you didn't for the first 7 years of—
No, we didn't. It was purely organic. That's when we built all the models, paybacks or TVs per customer, and also built a referral marketing product.
But the first 7 years were just organic, building financial products you thought were good?
Yeah.
How did people discover them?
Word of mouth. Plus, we've obviously done PR as well.
What were the best products from an acquisition point of view?
The best product was initially the first product that we launched. It was a multi-currency card with zero FX fees.
It's interesting that both you and Wise were able to get to scale based on consumers' frustrations with FX fees. The frustration with the incumbents in that product was so high that it birthed multiple successful fintechs.
Probably. If you look at banks, the majority of products are actually very bad. It's not only multi-currency transfers, but also credit, business credit, business products—almost every single product is really bad. Acquiring used to be the bank, right? I think it wasn't about multi-currency transfers or cards. It was more that we picked a certain product and made it so much better. It could have been acquiring, which you've done. It could have been credit as well, wealth and trading. Robinhood did it.
So you think there's nothing special about starting with multi-currency cards. You think you could have, in another universe, started with checking accounts—
Yeah. I think as soon as you build a much better product than banks, which is also much cheaper, then it's pretty easy.
What was the second product that worked? When did you prove to yourself that this wasn't a fluke, and that you could repeatedly build financial products?
Business accounts. We built business accounts. The offering wasn't about multi-currency cards anymore; it was a digital bank account. Without any marketing, we first started with probably 10 business customers a day. Now we're onboarding maybe 1,500 business customers a day. Only about 30% of it is paid marketing. The rest is still organic.
One of the things that, looking at Silicon Valley's successful product companies, tells you is there are many, many paths to heaven. There are many ways to build a successful product organization. Apple famously doesn't do any focus group testing. They just build the products that they know in their heart of hearts are right, in this very functional organization. And then you have Google or Facebook developing products in totally different ways. What is the Revolut system for building products?
We just put together a small team of 10 people: smart people, smart product owners, good designers, good developers. We just build what makes sense from a first-principles point of view. It can be either replicating an existing product but building it so much better, or trying new products that we think make sense. But to be honest, everything new is forgotten old. We haven't built any product that worked that wasn't an old product built badly by other organizations. We have an internal “new bets” process where everyone in the company can raise an idea, and then it goes through validation. If the idea is validated, it goes to what we'll call a committee of people who've done it before, and then they decide whether to fund the idea or not.
What's validation?
First, I think it goes to the team that validates unit economics, financials, and so on—whether it makes sense. In the majority of cases, it makes sense because the cost of the team is pretty low compared with the value of the idea if it's useful for customers. They pay for it, and you can generate tens of millions of dollars, compared with the cost of the team, which is small. Then it goes to the people who have built new bets before, and then to general managers. They also validate it from a purely common-sense perspective.
If everything is approved, then we separate a small team. We hire additional people there. Then they build a product within 9–18 months, and then we look at the growth of the product. If there is a lot of growth, then we scale up the team. One team becomes two teams, three teams, four teams, a department, and so on. For the bets that don't grow, we sunset them. For bets that grow but not very fast, we just leave them alone.
I'm surprised that ideas would need to be validated economically, because you're distributing products across millions of customers. Won't the cost of a—can't any idea support a 5–10-person engineering team to go develop a V1 of it? It's very unlikely that a product won't work economically when you have such a large customer base and you can fix it with pricing.
An example is Revolut Lite, which we launched across 100+ countries without the banking license and without our e-money license. It was purely a money transfer business. The idea was to build Venmo internationally. It didn't really work out because we thought distribution costs would be much lower. It didn't grow organically fast enough, so we had to do performance marketing. The unit economics with the performance marketing didn't really work in terms of payback.
In the majority of cases, 90% of the time, yes, you're right. If it works out, it works out very well financially. But there are some exceptions that you still need to validate.
Brian Armstrong from Coinbase has talked about how, when he was first pitched USDC, he thought it was kind of a silly idea. He wasn't convinced. But, sure, okay, if you guys want to go work on that, have at it. Are there ideas that came to you that you were initially not fully convinced by, but that ended up being successful?
To be honest, in 99% of cases, I'm always up for new ideas and just try them—
You're just open-minded, yeah.
Yeah.
Maybe it's an idea that surprised you. You didn't expect it to work. You were open-minded, but you were pessimistic. Then it outperformed your expectations.
I just had too many failures in the past when my judgment was wrong, so now I'm fully optimistic. I think it's not about the idea, even though ideas are important. I think it's about execution, to be honest with you. You can make almost any idea work.
Can you? Even the bad ones?
If it's not that bad, then you can make it work, provided you spend enough effort and enough time on it.
And people are bringing you the plausible ideas.
If the team is very strong, then even if the idea is average, you can still make it work. There are cases when the idea is great and the business model is great, but the team can be bad. It can still fly.
What's your favorite example of the “great idea, great team” intersection?
I think our subscriptions team is very good. When they came and said, “Okay, we want to launch an additional Ultra subscription,” I thought the idea was okay. It ended up being an amazing bet because it's a very expensive subscription. Ultra is effectively the Ultra bank account. You pay about €50 per month, I think, but then it gives you a lot of additional benefits. I thought, “The idea is okay. This business probably won't be very large.” It ended up being a very, very large business.
The closest comparison is kind of like an Amex Premium card—
Amex Premium, yeah.
Yes, yes. I'm so interested to hear you say that growth dynamics were not critical in the early days of Revolut. Firstly, so many financial products spread because a payment transaction has two legs. It's just very well suited to viral product growth, whether it be Bill.com or PayPal, where you can send money to an email address even if they are not signed up for PayPal. It's a very classic acquisition dynamic. It's interesting to me that that wasn't the case for you. And also, the peer-to-peer product is now so successful.
Peer-to-peer is definitely an acquisition driver, especially in countries where they don't have a very good payment system.
Yes.
So, for example, in certain markets in Europe, we—
Ireland.
Ireland, yeah, is one of them.
Romania is another. So yes, it worked in many markets where the bank payment system is pretty old.
Okay.
But I wouldn’t say it was the primary acquisition channel.
Yeah. I would draw a distinction between word of mouth, where someone has to go to a bar and tell you, “Hey, I discovered this great new product last night,” and viral product adoption mechanics, where if I want to pay you money, you become a Revolut user, and then you can tune it. It’s like, “Have you thought of sending money?” You can promote that. Meta does a very good job at this. When they want to grow new parts of the product, they promote them in the app. Again, you have knobs and dials that you can use. The in-product growth stuff feels to me—
Yeah, of course. We tried a lot of things, P2P included. The problem with banking products is KYC onboarding, right? There’s a lot of friction. Even if you have these viral things, then—
There’s a steeper drop-off.
Yeah. There’s a lot of drop-off. It’s not really viral. Plus, you cannot really sell it worldwide, right? You always need to sell into countries where you have licenses.
Yes, but it can be viral in the same way Facebook was viral in the early days, where they would go college campus by college campus. You can’t do the cross-border transfers, but there’s so much activity within a country that you can get viral growth within a country.
Yeah, if you can onboard users in one click. Like Facebook, then yes.
Yeah, yeah. Do you think the KYC is a bigger impediment?
When we first launched, there was a lot of virality because we were doing onboarding. I think it was up to €1,000 or €100 that you could do without KYC. Back then, the regulation allowed us to do it. So there was a lot of virality, but then I think it was stopped in 2017 or something.
So the KYC changes actually made the viral growth loop a little harder.
Oh yeah, 100%.
Though you can presumably still do a lot of nice viral growth, where if there’s money waiting for you and you just need to go through KYC to claim it, then you can do one leg of the transaction and not complete it, right? And that will still be plenty enticing for a customer.
That’s true. We also have it. But again, this K-factor is never a legitimate one. You cannot really experience exponential growth. It’s always less than 1.
Could you not? There could be a K > 1 product. I’m not too pessimistic.
Everything we’ve tried so far in the last 12 years has always been less than 1.
Okay. You’re in the K-is-less-than-1 business. The other thing I find interesting about Revolut’s business is that you’re a bank whose primary revenue source is not net interest margin, which is obviously unlike other banks. Did you arrive at this accidentally? Was it part of your thesis for Revolut? You’re very different in that way, and it’s quite asymmetric versus other banks.
I think the answer is pretty simple. We started in the UK and Europe, where conditions are very tough for startups. Interchange is very low, and credit is extremely regulated. You cannot really charge customers a very high interest rate, as you can in Mexico or Brazil, for example. So you’ll have 10, 20, 30 startups starting in financial services, and the majority will die, right? We were forced to optimize our transactional business and effectively make it profitable. If we had started in Brazil or Mexico, we would obviously have focused on credit business because it’s less regulated compared to Europe. We would make a lot of money on credit business. But in Europe, it’s not possible.
But aren’t you developing a balance sheet? You have one side of it, which is the deposits. You could build the other side with loans.
Yeah, but before we got deposits, we couldn’t build a credit business. Now we’re doing credit business. Now, yes, we—
Will you build a traditional bank credit business where you have the deposit side and the loan side, or will it look atypical in some way?
Well, I think our business model, and my personal opinion, is that we want to have the least amount of risk. Even if we lend, it won’t be a large portion of the balance sheet. At the moment, the target state is less than 10%, and then the rest we would securitize.
Why?
First, the simpler the model, the better from a regulatory point of view. Plus, if we want to be a tech business, we’d rather be in the subscription business. As soon as you go into the lending business, your multiples become similar to banking multiples. If you look at tech business multiples, you can usually get 40–50 P/E. In the banking business, you’ll be lucky if you get 10 P/E.
But is that because—I'm always skeptical of multiples-based arguments because, ultimately, the business is good or it isn’t. You can educate investors if the business is in fact good. When we started Stripe, people thought payments businesses were horrendous and deserved low multiples and everything like that. Over time, they came to appreciate payments businesses more. But if we had been fixated on the multiple that people ascribed to the business, we might have been scared off this path. I think you just need to decide: “Is this a lower-quality or higher-quality profit pool?” So do you think it’s actually a lower-quality profit pool, or do investors just have a weird hang-up?
I think it’s more the business itself. If you look at the credit business in banking, the general credit business, every time you do a loan or a credit card, you need to put provisions against it. In reality, the amount of these provisions is highly dependent on the models that you use and on economic conditions. You don’t know whether, in 3, 4, or 5 years, you’ll receive the loan back, right? You model it. On average, you hope to receive the money back, but things can change, right? This uncertainty is priced in by investors, generally through lower multiples. The bank can produce a lot of money now from the credit portfolio. But look, in 2–3 years’ time, unemployment goes from 3% to 6%.
Yep.
And all the profit is gone, right? Plus, they’re losing money. So I think this strategy—
Okay. You think, fundamentally, making your money via credit is this very cyclical profit stream that can be entirely wiped out in a downturn. It’s very sensitive to you doing the math right. Whereas the nice thing about subscriptions is, ultimately, people are paying you this money on a recurring basis. In a downturn, maybe some people cancel their subscriptions, but you can’t make negative money on subscriptions. You can only shrink the book slightly, so it’s much higher quality.
Yeah, exactly. It’s similar to selling options, right? You collect the premium for 10 years, and then in year 11, you lose it all, right? That’s a problem.
Okay. I want to go back to your product development methodology, though. You’re describing the early stages: there are some top-down product ideas and some bottom-up product ideas. They have to be validated so that, if they work, they can actually be interesting economically. You have a talented product leader and a few engineers working on it. Then what?
How do you decide which products get the in-product distribution? How do they earn that? How much are you willing to give them credit for, “Oh, I think this will become a good product over time”? Product developers always tell you, “I would be doing very well if you just gave me more distribution.” The product people like to blame distribution. The distribution people like to blame the product. Maybe you can talk a little bit about how the scale-up works.
Basically, it’s pretty simple, right? We’ve got the team. They’ll work on the product for up to 18 months. Then hopefully, they’ll launch the product.
They’ll work up to 18 months before launching anything?
From the moment we decide to launch to launch is usually 12–18 months. It takes time to get the team, onboard the team, build the product, and go through the approval process on the banking side. So 18 months is probably a good number.
Because you’re a regulated bank, do you have to go to your bank regulator to get sign-off on the product?
It’s not a bank regulator, but more like the internal governance process.
Oh, I see. You’re internally—yeah.
We will launch the product, and then we release it. We first release it in a test country, for example, Ireland or the UK, or whatever. Then, if it works—if everything is fine—we release it to more countries, and more countries, and more countries. We add it to our cross-sell engine as well, which is automated. There is no distribution problem, right? It’s automatically being added to, hopefully, the whole user base if everything works fine. And because we’ve done it for a long time now, we already have historical data on the performance of about 100 bets.
Then we just look at month 1 after launch: “What is the gross profit compared to previous bets?” Month 2. Month 3. Month 4. Month 12. And then: “How does it stack up with all the other bets?” If it’s in the top 25% quantile of all the bets, we just scale it up. We hire more people. We have 2 teams, 3 teams, 4 teams, and so on. For the bottom 25%, we don’t do much. We redistribute the team to do something else.
Then in the middle, sometimes we keep the team. Sometimes we don't keep the team.
Is a new product like that actually an option, or are there costs to shutting it down? Are consumers annoyed if they were relying on a product and then it goes away? Or are there some products where, if you had a pension product, you actually just can't turn that off because it's their pension? Do you think about the cost of shutting off a product when you decide to fund it? Does that differ by product?
Yeah, of course. We look at customer detriment, right? If there are already customers using the product and there is detriment, we just keep it as it is. We stop investing in it, so we just maintain it.
Can you tell me about the automated cross-sell machine?
Automated cross-sell machine. Well, basically, we have many products on the platform. Then we effectively have a machine learning model that decides, based on certain events and the customer profile, which product is most suitable at this point in time. For example, you land in Ireland, you have a UK account with Revolut, and then we offer you an eSIM, because you're landing in Ireland and you spend money on Airalo. We see all the transactions, so we can offer you one of our products. On top of that, we also have an LLM model that effectively behaves as an account manager. Through human reasoning, it looks at all the products, all the events, and all of the suggestions generated by the machine learning model, and then decides which product to push to the customer at this particular point in time. It's all automated.
I hadn't thought that you're in a much better position to cross-sell products than anyone else because you see what people are spending money on—
Yes, as well. Yeah.
You can just... If they're overpaying for something, you kind of know.
I mean, we don't see the full details of the bill, right? We just see the transaction.
The cross-sell engine has some propensity model for the customer. You're basically trying to find the conversion-maximizing sale, like the thing they're most likely to buy next?
It's a bit more complex than that. It's not only conversion. It also takes into account GP and the product, and we also try not to spam customers, so there's a certain negative value to our message to a customer. Altogether, the model makes a decision.
It's interesting that you describe products needing to perform well on a revenue cohort curve. It's a little like a race, where your job is to stay in the pack or stay at the front of the pack, but you're judged relative to other products. Have you ever had slow-burn products that, for whatever reason, had a gradual start? Maybe the team was just learning exactly how this works, but that wasn't actually a negative signal about the product. It just took a while for them to get going, and then it really took off?
To be honest, it's almost never happened. I don't even remember an example when something was growing slowly and then started growing fast. Usually, the trend is your friend. If something is growing very quickly in the first 12 months, it will continue to grow very quickly.
A classic bigger-company behavior might be to say, “Oh, this isn't working yet, but there's a good reason for it,” and blah, blah, blah. You think those are all excuses, and the trend is your friend?
Yeah. I mean, the general statistical trend is your friend, yeah.
Why don't other companies operate in this Revolut model of developing many additional products to cross-sell to the customer? It feels applicable to many other consumer companies. Grading them against the curve of how other products have done at a similar age, I don't think most companies—even those that sell multiple products—are that strict. Instead, they make excuses for why the product isn't performing: “Oh yeah, we just need more staffing.” They're not as strict.
I think the majority of companies don't do many bets at all, right? Because we've done a lot of bets, we have a feedback loop compared to other companies. Because of that, we're very strict.
Yes.
Other companies probably do 7–8 bets. The feedback loop is delayed, plus management changes. As a result, they never end up implementing any strict process on new bets.
Are there products you've done that have not been well suited to this model? I can imagine a product that involves changing many different parts of the Revolut experience would be hard to fit into this new-bet model, because it's fundamentally much more infrastructure rewiring or something like that. Do you sometimes have to do products that don't fit into this standalone-bet model?
To be honest, you can fit any product into this model, even if it's a product that's very new to us. For example, it might not be a financial product; it can be a B2B product. We've done a bet called Revolut People. We built an HR system internally for ourselves, and we decided to make it a new bet. Then we separated it into a standalone team, and we sell it externally. It's kind of a new bet, but we cannot really cross-sell it to our retail customers.
What about products that just don't generate revenue by design? Imagine if I'm a product leader and I'm building a financial insights product. It generates a lot of engagement with the app because it sends people a summary of their transaction activity and ways to spend money. The point is not revenue. The point is app engagement. How would you grade my success?
We still have P&L attribution to it by indirect attribution of P&L, because you can also calculate it.
You have some churn model.
We'll have 2 numbers. One is direct GP, and another one is indirect GP.
So I have some synthetic profit that I can attribute to myself because I saved this many customers from churning by having them come back to the app and/or spend incrementally.
I'll give you an example. We have RevPoints. It was also a new bet.
What is RevPoints? Is that rewards?
RevPoints is more like a rewards system, yes. You get points, and then you can effectively use these points to buy additional products that we offer. Accounting GP is very small, but if you calculate indirect GP, that's very high.
That's very interesting. I feel like a lot of the most successful companies get to an understanding of their economic equation and then take that very seriously. Amazon says, “An incremental Prime subscriber is extremely valuable, so we're going to set up a movie studio to drive more Prime subscriptions.” It sounds crazy, but if you really believe the math on how much a Prime subscriber is worth, then it makes sense. Google was always very fixated on just how valuable the incremental search customer was because of the lifetime value of the advertising revenue. So they'd do all sorts of crazy things, like pay a lot of money to become the default Safari search box. But again, it was all that attributed P&L. You think the attributed P&L revenue is very important, so that you can create accountability with product teams and know if something's working.
Yeah, ultimately you need to have one number that defines whether a product is successful or not.
Yeah. I think a lot of companies actually don't do that, to a surprising extent. They don't chase through the full accountability with a synthetic revenue number.
How important is the way people think about neobanks? As I understand it, whether there's an account or “an account” is whether people deposit their paychecks. A majority of Ireland uses Revolut to some degree. They might just use it for peer-to-peer or something like that, but they have an account. Then there's using Revolut as your primary bank account, which in the banking industry people basically think of as: “Do you get your paycheck deposited there?”
How important is it that you move people from being a lightweight user to a heavyweight user, from occasionally using certain features to actually being the place where the paycheck is deposited?
I mean, it's super important, obviously, right? Because as soon as you get a salary into the account, people use a lot of other products that we have. If they're just using P2P, they load $20 and send money to someone. It's not very profitable for us. We spend a lot of effort to move people's salaries to Revolut. Overall, I mean, it depends on the country. 10–20% of our monthly active users have salaried accounts.
What moves people over?
It's a very slow burner, right? We actually spend a lot of time and effort building products for kids. As soon as we capture kids, we can easily move them to the 18+ account, because it's very hard to move adults.
Yes. It's a tale as old as time. I remember when I was in secondary school, maybe a similar experience.
The local bank came and gave us a soda and a bag of chips to sit through a presentation on how to save money. Then they gave us all a bank account. They were in the school, acquiring customers aged 13 or 14, on the theory that bank accounts are very sticky.
You see very different behaviors from people. Presumably, if you have 10–20% primary banking share among the regular population, that’s much higher among the younger population because they’re Revolut-native.
If you look at different cohorts of users, then we probably capture more than 50% of users up to 35 years old in almost every single country we’re in, in Europe. We have difficulties capturing people aged 50, 60, or 70+, who still hold the majority of liquid assets. It’s very hard for them to move from a legacy bank to a digital bank.
Sorry, what’s that statistic about 50% of under-35s?
Basically, if you take all the countries in Europe and look at cohorts of people aged 18–25, 25–30, 30–35, and then older, in the majority of countries, we’ve already captured more than 50% of people up to 35 years old.
Having any kind of Revolut account?
Yeah.
How is this product development engine different as a regulated business versus if you were just a normal tech company?
In a normal tech company, it’s much easier to do. In banking, you need to go through many approval processes. The way we’re structured—because we have many banks—the approval process works as what we call NPAP, the New Product Approval Process.
How do you minimize churn?
By trying to build the best product.
But more than that?
More than that—
Because again, you’re very quantitative, you’re very product-oriented. This is the thing that quantitative, product-oriented people obsess over.
Yeah, we used to have this framework of a state machine. You’re probably familiar with it: how many people were onboarding, how many people went through from being P2P customers to salary customers, and how many of them churned. Then it just became too complex. We ended up just having a cohort of customers who joined per country, and then looking at how many of them are still with us within 12 months, 18 months, and 24 months.
What we observe is that we have a “smile.” Out of 100 customers, maybe 60 are monthly active. But after 2 years, it will be 65 or 70 customers. People return to us. I would say we actually have negative churn.
Why do people come back? Why do people leave but then come back?
I think it’s because a larger and larger portion of the population started using Revolut. People who churned for different reasons still hear about it from friends and family, and then they come back.
I see. You have multiple word-of-mouth shots on goal. Okay, so, yeah, I was curious to ask you about AI. How AI-pilled are you in how you run the company?
I obviously started with AI about 2 years ago. The goal was just to optimize customer support and translation. Now it’s almost in every single area of the company, including engineering and marketing. We’re even trying to create an AI employee, end to end. Yes, we’re very involved.
For the last 10 years, I’ve optimized a lot of things in the company based on people, on their advantages and disadvantages. We built all the systems for KPIs, roadmaps, processes, hiring, and performance management. Now it might all be redundant in 2–3 years’ time because of AI. We’re all reinventing how the company should work.
How has it changed how you hire, either at an overall company level or at an individual team level? Should the 8-person teams now be 3-person teams? Can you have higher aspirations for what an 8-person team can accomplish?
Technically, one person can accomplish much more now. I would even say you don’t really need to have separate Android and iOS developers. One frontend developer can do things. You can even have one developer covering frontend, backend, and possibly data science as well. So, you can have one person doing all aspects of the product.
You should probably still have a designer, although it’s so much easier to design that one developer can do design as well. A team can probably be 2 people.
You work in this relatively standardized way of product bets, where you put a small team on something to initially test it. Has the size of that initial product-bet team changed? And shouldn’t it change, if not?
Not yet, to be honest with you. We’re still operating in the old framework.
Isn’t that very old-fashioned of you?
It’s pretty old-fashioned, yeah. But I just prefer to continue doing things that work, while experimenting with new things. When we have a proven track record that new things really work for us, we will start changing.
That’s interesting. You think that a 4-person team cannot bring a new product to market, even with all the new productivity enabled by AI?
No, a 4-person team definitely can. It’s just that all the processes within Revolut are still set up by specialization. You need to have an iOS developer. They have the system; they know how it operates. Then you need an Android developer, a backend developer, data science, and an Ops manager. It will be very hard to just hire one person who will get access to all the systems and then produce all the required code with AI.
I see. So partly, it’s just that you need to reorient Revolut a bit.
We need to change the process, yeah.
How does this change the shape of the Revolut organization?
Well, first of all, we’re just trying to automate a lot of things. Hopefully, when automation is done, we can start hiring fewer people. At the moment, I don’t think we’re at the stage when automation is done.
You traditionally competed with legacy banks. People might go to their bank for FX, or they might use a Revolut multi-currency card. You had a lot of advantages: you had no tech debt, no retail branch presence to afford, and many other advantages versus traditional banks.
You’re very successful. As you go to Brazil, you compete with Nubank. And as you go to Kazakhstan, I guess you compete with Kaspi? Is that their name?
Yeah, there is a Kaspi.
Kaspi, yeah.
We’re not in Kazakhstan yet.
Okay, but when you do, how is the world going to look different when you’re competing with neobanks rather than traditional incumbents?
Well, obviously it’s more difficult. Ultimately, when we go into a country, the decision is a combination of how bad the banks are, because there’s less competition, what our unit economics per customer are, and how difficult it is to get a license. Obviously, if a country has a lot of digital banking competition, it’s difficult to get a license, and the unit economics aren’t great, we will not go there. Or we’ll try to deprioritize it.
Then we try to prioritize markets where there are a lot of legacy banks, the unit economics are great, and hopefully it’s easy to get a license.
Okay, so you look for markets with sleepy incumbents.
Exactly, yeah.
That’s a good market for Revolut. What predicts the unit economics of a market, other than how rich people in that market are?
Well, it’s interchange legislation or laws, and legislation on credit rates as well. Generally, if inflation is high, then credit rates are even higher, and the margins are even higher.
Isn’t it surprising, then, that you grew up in Europe, where both interchange rates and lending rates are regulated? Wouldn’t that be a bad market according to this rubric?
Yes, a very bad market, yeah.
Does the rubric miss something?
I think it’s just because we started in Europe, where it is very hard. Out of 50 digital banks, probably 3 or 4 survived. We really survived in tough competition. But that means we’re more fit to compete in other markets.
Let’s talk about the US market. The US has historically been very resistant to neobanks. Remember Simple?
Yep.
That was what, 2012, 2013?
I think they were the earliest one, yeah. Even before Europe, yes.
Yeah, around then. Various foreign neobanks have tried to enter the US market. The US is such a quick adopter of so many technologies, and it invents so many of them. Leaving aside Revolut for a second—we’ll come back to that—why is the US so slow to get neobanks?
Well, I think historically it was a very tough regulator. The US regulator was one of the toughest. That’s number one. Number two, all these startups couldn’t really get a banking license, so they chose not to get one because it was difficult. Traditional tech companies don’t want to be regulated because it kills their business. As a result, they jumped into partnerships with legacy banks.
It doesn’t really work because you’ve got tech, and then you’ve got compliance people in banks. It’s a different company. They don’t really match together. So the model was wrong.
You think you have to be willing to own the stack end to end, including all the licenses.
Including licenses—
Which was tricky in the US because it was such a stern licensing regime.
Yeah. Plus, the culture was: “Okay, you need to have tech separate from compliance.”
You need to have tech separate—
In reality, it’s one business. You also need to automate tech, automate compliance with tech, and so on.
Yes.
Why did we not see a version of Simple that worked? Simple tried to do the tech via partners. Why didn’t someone try to start a neobank that owned the licensing end-to-end in the US?
I think it was very difficult to get a license. I know Square and Cash App have a banking license now. They’ve activated it, but it took them 4–5 years.
Yeah.
Times have changed, definitely, with the new administration.
Is the current regulatory environment better in the US?
The process is quicker, yes. Now you can get a license within a year. Before, it could take 4–5 years.
And you guys have gotten that? You will get that?
We received conditional approval. There are 2 more approvals that we need to receive.
This is from the OCC?
OCC, yes.
Okay, very good. And so, you guys will be going big in the US soon, when you get the—
Around next year, hopefully.
—the rest of the licenses. What will need to change about the product to be successful in the US?
Well, it definitely needs to be credit card-led. You’ve got a lot of benefits in the US from Amex and Chase. We’ll do the same, with a credit card-led proposition and a lot of benefits as well.
Okay, so you want to go head-to-head with Amex Platinum and Chase Sapphire?
Yeah.
How will you compete?
By offering the same benefits or better, plus our whole proposition. They’re offering just payments products. We’re offering wealth, crypto trading, and so on.
Why launch your own euro stablecoin?
It’s good business, obviously. That’s number 1. Number 2 is that there’s a USD stablecoin, but there’s no equivalent—or credible equivalent—for euros in Europe. We should do it.
Yes. You said you’ve been doing this for a very long time. It’s been multiple years?
Yeah, the tech was built a long time ago.
How do you organize your day? Are you waking up and going wherever your curiosity takes you, or is it hyper-scheduled?
Monday to Friday, it’s hyper-scheduled, in 15- or 30-minute slots. Usually it’s business reviews, product reviews, and one-on-ones. Saturdays and Sundays are more freelancing and free thinking.
Revolut’s the kind of company that has a work-from-home policy where you can work from home 2 days a week: Saturday and Sunday.
Yes, true. I try to limit our external meetings as much as possible and spend 95% of my time internally.
Do you think most CEOs are too externally facing in how they spend their time, going to conferences and all this kind of stuff?
100%. I do have an allocation for conferences, maybe 2–3 times a year, but that’s it.
Yeah. You just want to be within the building, reviewing the actual product?
Exactly. Products and business processes.
What are you actually reviewing? Are you looking at metrics dashboards with people? And what are you, Nik, adding? You could also have an automated system that shuts down any product that’s failing to meet its revenue goals.
Basically, we have all these KPIs, roadmaps, and processes targeting execution. Before, I was doing it manually. Now it’s all semi-automated, with the performance team also reviewing everything: whether the KPIs are correct and whether the roadmap is being executed for each team, department, and at the company level.
What I do is product: product direction, product features, and detecting bullshit as well. There can be a lot of cases when everything is rosy, KPIs are good, and everything is met. But things can be even better. Then I just know it.
What is the bullshit that you’re detecting if the KPIs look good? In what way are they trying to bamboozle you?
You need to dive deeper than KPIs. A KPI can be met because of external factors and because of luck. Then you dive deep and look at the product, product features, and speed of development as well.
I’ll give you an example. Our wealth and trading product grew 60–80% year-on-year 2 or 3 years ago. Then I thought it wasn’t enough. It should have been around 300%. I dived deep, and what I realized was that the quality of the product wasn’t great. The features weren’t good. Then we changed the team, and now it’s growing 300%. You need to have intuition for these things.
What’s the easiest way to annoy you in a product review?
Incompetence.
What kind of incompetence? There are many forms.
We have certain standards for quality. Everyone knows these standards. If somehow I see a product below the standard of quality, and we’ve done many products so far, then it really annoys me.
Okay, so shoddy product quality is the criteria—
Yeah, that we’re going into the same mistakes.
Yes. How’s your AI hedge fund going?
It’s actually going well. In terms of returns, it’s now in the top 3% or 4% of funds in the world. The model really works.
What is the model?
Our model is a machine learning model. It’s trained on a lot of data on startups. It’s a feature-based model. Based on our number of features, it tells you which startups to invest in.
It’s not a hedge fund, it’s a venture firm.
It’s a venture firm.
It’s an AI venture firm. Okay.
They’ve already invested in 40–50 deals in the last 2–3 years, and the returns are good.
Isn’t the classic problem with this kind of approach that, with private company investing, it’s about picking the right companies and then being able to invest in them?
Exactly.
Sometimes it’s very obviously good companies, but you just can’t invest in them.
Yeah.
How does that work here?
As you correctly noticed, the first part is to create a good model that will tell you with high precision which startups to invest in and which not to. Then, second, to enter the deal.
Our sales proposition to founders is that, first, we will help them with Revolut’s operational playbooks. We tell them exactly how Revolut hires people, how performance management works, and so on, so they don’t reinvent things from scratch. Second, I help with my personal advice. Third, we help them sell either into Revolut or to Revolut’s customers as well.
I see. So, you’re a desirable investor and—
So, an added-value investor. Exactly. You can trade the added value for that.
When’s Revolut going public?
Well, I hope in 2028. It’s now 2026, so in 2 years’ time, subject to the market being good.
What needs to be true for everything else to be ready, or is it just already ready?
It’s still a work in progress. We need to launch in the US. We need to launch in a lot of new expansion markets. We still need to prove that we can be as successful in Europe in markets outside of Europe, especially credit-led markets.
Do you actually need to launch in the US before going public? Isn’t Revolut a huge bank already, with billions of dollars in profit? You’re plenty scaled already.
No, of course. It’s not necessary, but it’s good to have.
Because investors are based in the United States? Again, this home-country bias means they’ll be more interested in Revolut if they’re familiar with the product?
I’m not really optimizing for valuation or readiness for an IPO. It’s more: “Okay, this is the roadmap that we want to complete. Let’s focus on the roadmap.”
You mentioned helping startups understand the Revolut playbooks. Where do you think the Revolut playbook is most differentiated? You talked about performance management and a few other things.
Obviously, hiring, performance management, and product building as well. I think with AI, it might be irrelevant in the next 2–3 years.
How do you do performance management differently?
I think it’s very important to do it quarterly. The majority of companies do it once a year, and I think that’s too long. The problem is a lot of managers are inherently biased against exiting underperformers.
Mm-hmm.
By having a process once a year, you’re not really helping the company. Then you’re forcing managers to make performance-based exits once a year. In reality, you need to do it as fast as possible. You can’t really do performance management once every month, because it’s too time-consuming. But you can do it once every quarter.
Do you try to make performance more quantitative than other companies? Mapping products to where they should be on the cohort curve is maybe one example of that. Do you try to do that in other places?
Yeah, we’re trying to be as quantitative as possible. Every person is assessed on their skills and deliverables. Deliverables are measured by the roadmap and KPIs, and then there’s culture. All these 3 aspects are assessed by the manager, plus the functional manager. There’s also a performance review on top of it, checking that everything is correct, and then a person is assessed.
Some domains are just fundamentally more quantifiable than others. A salesperson—you can measure how much they sold.
Whereas a lawyer, you can only make that so quantifiable, right?
Well, in our case, we have metrics for almost every lawyer, I think. Definitely for the team of lawyers. If you look at commercial lawyers, for example, they process contracts. Are they meeting their SLAs? That’s the number one metric. How many deficiencies are in the contracts? Again, you can measure it, and so on. On the team level, we definitely have KPIs. On an individual level, I’m not sure, but it’s possible to build them.
I don’t think most companies really believe you can measure lawyers on a grid of KPIs. Do you think, again, that you can be more quantitative than you think?
You can definitely be more quantitative than the majority of companies, I think. Can you quantify everything? No. You need to have some other leading indicators rather than KPIs.
Okay, we’ll leave it there. Thank you, Nik.
Thank you.