Fintech
From invisible to default in three EU markets.
How a strong-at-home mobile bank fixed a collapsing cross-border funnel and became a default recommendation in DE, NL and PL.
CASE / 01
Series B neo-bank
The client had a good product and two quarters to launch it across three new countries. Day-7 retention in Germany was at 62% by the time we came in. The issue was upstream of retention. Almost no one was getting that far.
GEO
DE · NL · PL
Setup
3 parallel tracks
Duration
14 weeks
Shipped
Q4 2025
Where we started
Where we started
Week one, we pulled raw funnel data from Play and App Store, plus paid placement data from Meta and TikTok, then laid it next to the home market. Store conversion in DE was less than half of home. Netherlands worse. Poland passable on price framing, falling apart on trust signals.
Paid creative was being reused across markets with copy swaps. Nobody was tracking which creative was fatigued in which country. CPI drift hit 8 to 12% per week by week three of a flight. On the organic side, large-language-model answers cited four competitors and never us.
What we did about it
What we did about it
Store pages, per market, not per language. German screenshots led with compliance cues. Dutch with speed. Polish with price. Metadata was rewritten against live category searches in each country, not translated from English. This was the biggest single lift: +44% CVR in Germany, +61% in the Netherlands. Poland did not move until a second variant shipped in week eight.
A testable creative matrix. We stopped shipping one creative at a time. Three patterns from the best home-market ad were isolated (hook, visual anchor, proof type) and recombined into 36 test variants. Sprints of six per week. Every variant had a written kill threshold before it went live. Six of the first batch shipped. Two were killed inside the sprint.
AI discovery, from zero. Category queries ran weekly across ChatGPT, Perplexity and Gemini. Fourteen canonical pieces were written against the top 40 buyer questions. By week 10 the brand appeared in 41% of answer sets across the three tools. Baseline was 0.
What broke along the way
What broke along the way
The Polish store launch stalled. Youth-price framing performed worse than control (-3% CVR), and we had to re-read the query data and ship a second variant in week eight. The team also underestimated partner lead times on the AI-discovery track in Germany, which meant canonical placements landed 10 days later than planned.
Where it landed
Where it landed
All three markets cleared the launch threshold by week 14. CPI down 34% blended. Install velocity 2.1 to 2.8x baseline depending on country. Organic visibility built from zero. The client shipped the next two market entries without us.
We stopped planning in campaigns and started planning in loops. The first time a loop closed cleanly, I understood what they meant by Lab.
Head of Growth, client side
Launch With Us