The risk assessment is the long bit. Picking a portfolio should be the easy bit. Right after the chat, the app even said: “Hardest part is now done. Excellent work!” Yet a large share of the people who finished the chat never confirmed a strategy.

Three look-alike cards in a carousel, each description cut off mid-sentence. No clear sign of which one you'd picked. And right above the button: “Your choices can affect the expected returns of your portfolio.”

Call notes, customer feedback and a look at other investing apps told the same story. We expected people to compare performance and allocation. They were comparing flavours: global, sustainable or income. And the ones who continued weren't deciding what's best. They were checking that it made sense, then moving on.

“Can I choose the investments myself?”
“How does Selma put my investments together?”

— new users, from onboarding calls and feedback

So the question changed. Not “which portfolio should we recommend?”, but: how might we help people recognise the right theme for them, without recommending one?

• We can't recommend — we don't know enough about someone's preferences, like sustainability.

• Two weeks, existing flow — improve what's there, no new screens.

• Every reassurance has to be true — if we say you can change it later, you really can.

The carousel became three tabs: one theme at a time, always clear which one you're on, and comparing is a single tap. The page opens with “There's no wrong choice — you can change it anytime later.” The button says exactly what happens: “Select Classic”.

• Lead with the difference — descriptions went from around 40 words to under 20, opening with what sets each theme apart.

• “Most popular”, not “Recommended for you” — a fact about what others choose, not advice.

• Nothing is chosen for you — Classic shows first, but the choice only happens when you press the button.

• Reassure, don't warn — the returns warning and a Preferences disclaimer are gone.

In the weeks after the release, the share of new users who went from the risk assessment to a confirmed strategy nearly doubled. The project also added new building blocks to our native design system: a tab row, a new attribute row, and a tag on the portfolio card. To be fair: other onboarding improvements followed soon after, so not all of the lift is ours.

What I'd take with me: when you can't recommend, help people recognise themselves. Next time I'd ship behind an A/B test, so the number is fully ours. Still on the list: letting people ask Selma's AI to explain the differences between themes.

Exact figures are confidential. Amounts in screenshots are examples. Reach out for the full story.