On 16 September 2026 Revolut applied for a FINMA banking licence and committed CHF 150 million to Switzerland. The licence unlocks Swiss IBANs, salary accounts and eBill. The return on that investment depends on one question: which existing customers will make Revolut their primary bank, and in what order. We built a simulation of the Swiss retail market to answer it before the licence lands.
Switzerland is a primary-bank market with unusually sticky products. Whether a customer moves their salary depends less on Revolut's offer and more on how anchored they are at their house bank. That side of the equation is invisible in Revolut's own data.
Cohort: 6,506 simulated Swiss adults from a 10,000-client, twelve-month simulation, of whom 1,693 carry a Revolut account under an adoption model calibrated to SNB survey rates. Each user's ties to their house bank are scored on products, retirement and property lock-in, wallet share, balances and age.
| Products at house bank | Acquirable | Defended | Share acquirable |
|---|---|---|---|
| 2 | 580 | 7 | 99% |
| 3 | 93 | 689 | 12% |
| 4 | 5 | 280 | 2% |
| 5 | 0 | 37 | 0% |
| 6 | 0 | 2 | 0% |
| Monthly income | Acquirable | Defended | Share |
|---|---|---|---|
| Under CHF 3,000 | 266 | 39 | 87% |
| CHF 3,000 to 5,000 | 155 | 242 | 39% |
| CHF 5,000 to 7,000 | 166 | 288 | 37% |
| CHF 7,000 to 9,000 | 52 | 210 | 20% |
| Over CHF 9,000 | 39 | 236 | 14% |
| Metric | Acquirable | Defended |
|---|---|---|
| Average age | 37.3 | 41.8 |
| Has Pillar 3a at house bank | 4% | 98% |
| Products at house bank | 2.2 | 3.4 |
| Wallet share of house bank | 62% | 64% |
| Median balance at house bank | CHF 22k | CHF 118k |
| Monthly income | CHF 4.1k | CHF 7.3k |
| Stress index (0 to 100) | 63 | 56 |
The easy salary accounts sit in the low-income, low-balance tail. The deposit and wealth opportunity sits with customers who are anchored by a Pillar 3a. Those are two different launches.
A logistic-regression classifier fitted live in your browser on the 1,693 simulated Revolut users, scored out-of-sample with five-fold cross-validation. Toggle what the model is allowed to see, move the threshold, and read the consequences on the scatterplot and the ROC curve.
Shares are of the 1,693 simulated Revolut users. Applied to 1.3 million customers, a few percentage points of difference in sequencing is tens of thousands of salary accounts.
Already run most of their spending through Revolut, hold little at the house bank and have nothing that needs paperwork to move. Ninety-six percent of this segment clears the switching threshold.
Salary account with instant IBAN switch, employer letter generated in-app, eBill migration. Price on the things they already use: FX and travel. Expect volume, not deposits.
One retirement product holds them. Wallet share and balances say they are ready to leave. The 3a is the only thing that has to travel. Small now, but this is the segment that grows the moment Revolut can hold a 3a.
Salary account plus a 3a transfer path in one flow. Until Revolut can hold the 3a, market a salary account without asking them to close anything. Second wave.
This is where the deposits and the future lending book live. A salary-account campaign will not move them. They use Revolut for travel and will keep doing so.
Win a second product first: CHF deposit with Swiss protection, or the 3a once available. Measure wallet share monthly and re-score. Primary status follows the second product, not the other way round.
One artefact worth noting: the switching-cost score also flags 135 retirees over 65 with few products, a fifth of the acquirable group. Commercially they are deposit prospects, not salary accounts. A production scoring model would carry a value dimension alongside the switching-cost dimension.
Both profiles are synthetic individuals generated from Swiss population marginals, taken from an earlier four-month pilot run. Nothing here is a real person.
The engine is a population simulator built for Swiss retail banking. It carries the house-bank side that neither Revolut nor any single bank can observe, which is exactly the side that decides primary-bank conversion.
Synthetic Swiss adults drawn from official population marginals, real address stock and merchant data. Typed financial state: income, accounts, products, balances in CHF.
Monthly event-sourced ticks: salary, rent, spending by category, product uptake, life events. This study uses the 10,000-client, twelve-month run: 8,046 individuals, 1.5 million events, 2.4 million transactions.
Revolut ownership assigned per adult from age base rates calibrated to SNB neobank survey data, with uplifts for travel, tech affinity, urban canton and low wallet share.
Each Revolut user scored on product count, mortgage and 3a lock-in, wallet share, balances, age and stress. Top 40% labelled acquirable. Gradient boosting transfers the score onto features Revolut can observe.
Designed so the first phase starts before data access is negotiated, and so every phase ends with something the Swiss leadership team can act on.
Port the switching-cost framework onto Revolut's actual Swiss customer base. Observable features only: age, tenure, top-up patterns, share of salary-like inflows, spend mix, travel share, canton. Calibrate the unobservable house-bank side with the simulation as a prior.
Runs inside Revolut's environment. No customer data leaves the perimeter.
Build a synthetic twin of the Swiss base at 1.3 million scale and run the launch scenarios that cannot be A/B tested on real customers: salary-first versus 3a-first, canton rollout order, FX and deposit pricing, the employer-letter flow.
Each scenario returns salary accounts, deposits and cannibalisation over 24 months.
Monthly re-scoring against actual conversions. Wallet-share tracking as customers move inflows. Uplift measurement per segment and per campaign so the model earns its keep on real outcomes.
Handover to Revolut's data team with documented features, code and a reproducible pipeline.
Revolut's Swiss challenge is to take primary relationships away from cantonal and regional banks. Our references are those banks and the core-banking vendor that runs them. We know what the defending side sees, measures and worries about.
Questions on the data, the segment logic or what Phase 1 would look like inside Revolut's environment: reply to the email this page arrived with, or reach Pascal directly.