From problem to live product.

The first big problem to solve was the funnel between Meta ads and account creation: it barely converted. I ran a survey of 300 people to understand how Germans relate to pension investing, analyzed the existing product's performance, and defined a new path.

It took 16 prototypes and 4 final versions over a month of testing before testers approved the solution. It was my first real professional test: I learned the humility of staying true to the findings even when instinct suggested a shortcut.

The idea behind it was simple but a little brutal: let people know what was coming, so they could prepare in time. After name and date of birth, the first hit was an estimate of when you'd retire. Then, instead of showing your own numbers right away, the app told the story of a character your same age: a device to teach the mechanic before making it personal. A time-travel jump to 2063 to find out that her pension would only cover 46% of her income, according to official German statistics.

That difference was called the 'pension gap', explained with a simple diagram: full income today, halved income in the future. It was designed to shock, to create a sense of urgency, and to lightly introduce the alternatives that existed. It was a long onboarding, working against every best practice about brevity and friction, but it's exactly what triggered the jump in conversion from 1% to 25-45%.

From there, the account was created, and a series of questions followed to actually populate it: marital status, identity verified by SMS, your own state pension. By the end, the app was already showing the first solutions to close the gap: from fear to the promise of an answer, in just a few steps.

With conversion solved, the next problem was retention: after onboarding, users kept coming back to a fast but inaccurate simulator. I worked on both the interface and the messaging: every setting became clear, with no need to guess which number to enter because it was suggested and explained, and I updated the chart colors while keeping the same interactivity. The result exceeded expectations, and the company gained authority in users' eyes as a subject-matter expert.

The 'book your consultation' button remained a delicate spot: the sales that followed depended a lot on the individual advisor's skill, more than on the app's promise, a sign the product alone wasn't yet closing the loop. With few real users to test on, we chose to work with an external testing panel, protecting the company's credibility in the meantime. The result was valuable: people like to investigate on their own, before trusting whoever is selling to them.

The product aimed for a precise idea: educate people enough to let them choose the investment on their own. A short questionnaire led to a comparison table that worked beautifully, a team result I'm still proud of, with filters and selectable options. From there, whoever had made a choice could proceed in three ways: get a copy by email, book a review with an advisor for confirmation, or go straight to signing up, without needing to talk to anyone.

The critical point was exactly when a user chose to discuss it with an advisor: their choice and reasoning didn't arrive with enough context, and got lost in the handover. Further upstream, the real issue was that the financial content behind the simulations wasn't developed well enough yet: results looked flat and interchangeable, as this screen shows, three providers with the same 9% return and around €920 net monthly income each. Without real differentiation to point to, the advisor's job got harder too.

The commercial pace aimed to close fast, while the product was designed to nurture the lead over time: an interesting mismatch in hindsight. The more information we gave, the more questions arose in the user's mind, and sometimes the less transparent competitor came out ahead. We discovered, a bit too late to test for it beforehand, that our more educated users would call several market players to compare discounts, and pick whoever offered the biggest one: a valuable lesson about loyalty built through transparency, more fragile than we'd hoped.

In parallel I supported marketing, mapping the entire path from social to the website to onboarding, in a saturated market where an ad would get copied by a competitor within two days. When the original onboarding became outdated, we discovered that taxes had become the most effective angle in ads: that's where a second onboarding came from, three interactive, animated screens, each explaining one step of how tax deductions affect real income.

From that point on, the app opened on taxes, not on the pension simulation. That angle, too, was soon replicated by competitors.

A part of the work I particularly care about: moving asset entry from uploading bank statements and contracts, which an advisor then had to read and interpret, to forms designed to be filled in independently, without waiting for anyone.

The hardest part was real estate, because it uses terms that aren't common knowledge outside the industry. My architecture background helped there: I already knew those terms well, and could translate them into a form that was clear even for people who'd never heard them before.

I had spotted a pattern I believed could work: turning the product into a B2B tool supporting financial advisors broadly, not just B2C. For the founders, though, it was a time to consolidate rather than take on more risk: with the German recession underway, it made sense not to expose the capital they'd built to new unknowns before investing further in a long-term product. A decision I understood, even as the pattern I'd spotted kept nagging at me. The product team was scaled back to wait out the market.

I stayed on to support the ads side, then chose to move toward a project that offered real, concrete work. One thing I still carry with me from that time: I defined how we worked with developers, understanding what they needed and building the graphic material that fit their workflow, to the point where we could ship a new feature in just two weeks.