Product leader · AI transformation
20+ years building digital products for millions of people across four European markets.
I like the work that starts after the launch.
Results
Four results, with the context behind them.30% → 70%
Conversion funnel lifted over four focused iterations
Facile.it — Insurance
Continuous discovery (Teresa Torres); a pet line validated from zero
+150%
Cross-sell & upsell across the full shopping journey
Lastminute.com — 4 EU markets
Weekly experimentation + a single shopper view (CDP)
3mo → 1mo
Product cycle compressed, Figma phase eliminated
Loquis — AI-first
7-day AI prototyping loops
27M
Monthly active users, NPS +10 points
Vodafone — first EU telco money-back guarantee
“Soddisfatti o Rimborsati”
Led product teams of 28, 40 and 27 people at Lastminute, Zalando and Facile.it
02 · Method
How I work
I start from the business problem, then work out what I don’t know yet.
01
Problem
Frame the business problem, not the feature request.
02
Hypothesis
Name the real bottleneck and the assumption under it.
03
Bet
Place one focused bet. Iteration over big redesign.
04
Result
Measure it in money, retention, or time.
03 · Selected work
Selected work
30% → 70%
Facile.it
Insurance · Continuous discovery
ProblemConversion stuck at 30% with heavy drop-off.
HypothesisThe bottleneck was need discovery, not price.
BetFour targeted iterations on an Opportunity Solution Tree — not a big redesign.
Result70% conversion, +7% traffic in the downstream funnel, a pet-insurance line validated from zero.
3mo → 1mo
Loquis
AI-first product · Prototyping
ProblemA 3-month cycle was incompatible with an AI-first product.
HypothesisThe Figma design phase was the waste.
BetCut it. Move to 7-day AI prototyping loops.
ResultCycle down to one month; freed capacity redirected to real-time user validation.
+150%
Lastminute.com
Shopping journey · Scale
ProblemA fragmented journey across four markets, low cross-sell.
HypothesisWeekly experimentation on the whole loop, plus a single shopper view (CDP).
BetEnd-to-end redesign instead of local optimizations.
Result+150% cross-sell, −40% post-sale contacts.
<3s
Algorand Labs
Tokenized finance · Governance
ProblemTokenized bonds needed regulator trust to exist at all.
HypothesisHere, governance co-design beats time-to-market.
BetCo-design governance with Banca d’Italia; prove settlement, then decide.
ResultBonds issued in <3s in test; scope deliberately paused rather than pushed to launch.