Heya. I love cooking up product solutions:takingill-definedproblems andmaking themsolvable.
Hi, I’m Sébastien, Tiao if you follow Duckrew. I spent three years at Flitter, an insurance broker: came in as a designer, left as a product manager, and spent most of it finding out where the real problem was hiding. Then a year building Duckrew on my own: the product, the audience, and yes, the code. Below, four problems I’m happy to talk you through.
Insurance & fintech · Regulated products · User research · Design systems · Design + code
Work
Flitter → Duckrew · 8 projects
- Screens to come
Flitter · Wallet
- Screens to come
Flitter · Onboarding
- Screens to come
Flitter · Claims × Norauto
- Screens to come
Flitter · Design system
Duckrew
Duckrew Live
- Screens to come
Cayenne
Duckrew · Identity
What I’m actually for
- 01
Find the real problem
The brief is rarely the problem. Under a simple request there’s usually a symptom, and the work starts where you dig.
- 02
Hunt the answers where they live
Support knows why people call. Engineers know what the system can do. Legal knows what you can promise. Users contradict you. None of it is in Figma.
- 03
Decide, ship, measure
Own what you ruled out. Prototype in code when it’s faster. Then read the numbers, especially the ones that say you were wrong.
The problems
01 · A constrained problem
Flitter · Wallet
- The real problem
- The State killed the green card. Flitter was losing the only physical touchpoint with its customers.
- What I did
- Put the insurance certificate in Apple & Google Wallet, with the whole contract on the back and a QR code partner garages can scan. One object, three problems solved.
Relayed by 3 trade titles the same week
Read the case · 5 min →
02 · A trade-off
Flitter · Onboarding
- The real problem
- Every document you ask for before payment loses buyers. Every one you ask for after risks the 30-day legal cutoff.
- What I did
- Pushed friction as late as possible, then made it simple enough to finish: step-by-step, live document checks, reminders that show only what’s missing.
+20% contract completion
Read the case · 4 min →
03 · A problem across two companies
Flitter · Claims × Norauto
- The real problem
- A claim handled out of network costs more and generates more calls. Sending people to a partner garage meant two companies, two systems, one driver in the middle.
- What I did
- Learned the partner’s booking interface and ours, then built one flow: the appointment is taken from the claim, in the customer area, with the data both sides need.
2 companies, 1 flow
Read the case · 3 min →
At Flitter, the constraint was given to me. I wanted to see what I’d do with none at all.
04 · A problem with no known answer
Duckrew
- The real problem
- How do you become France’s most-followed surf media in a year, alone, with no budget?
- What I did
- Treated editorial like a product: hypotheses, tests, retention curves. Turned a 49.6% skip-rate failure into two writing rules, validated on later scripts.
13 → 45,700 Instagram followers in 12 months
Read the case · 6 min →
Where the answers were
Five projects, six places to look. A dot where I went; hover it for what that source taught me.
Wallet
- SupportSupport knew the first question the reform would trigger: “so how do I prove I’m insured now?”
- EngineeringWhat a wallet pass can carry, and which of our fields could feed it. The first pass file, I wrote myself.
- LegalOn an insurance product, what you display is a commitment. Legal set what the card could say.
- PartnersRepairers feared the reform would restrict free choice. We showed up with more data than the paper certificate ever gave them.
- UsersCustomer feedback after launch reshaped what the back of the card shows. The second version came from them, not from me.
- DataPublic reviews started citing the wallet card unprompted. Three trade titles covered it the same week.
Onboarding
- SupportSupport knew why files stalled: minor mismatches between documents, almost never fraud or neglect.
- EngineeringEngineers defined what could actually be read from a document, and what each field cost.
- LegalThirty days after purchase, an incomplete file closes the contract. Legal, not commercial.
- UsersSMS reads as an emergency. I’d planned it as a reminder channel; interviews killed that.
- DataThe whole flow lifted completion by 20%. I can’t isolate the one-document reminder from the rest, and I won’t claim to.
Claims
- SupportSupport had the numbers: a claim handled out of network meant more calls, longer handling, and a worse outcome for the driver.
- EngineeringTwo companies, two APIs, one appointment. Engineers on both sides defined what could be shared, and when.
- PartnersThe partner’s booking tool had its own logic and its own fields. The flow had to fit theirs, not only ours.
Duckrew
- PartnersBrands came once the audience existed, not before. Partnerships that didn’t fit were declined.
- UsersThe audience isn’t one block: 74% under 34 on TikTok, 35 to 54 on Facebook. Three platforms, three treatments.
- Data49.6% skip rate on a hook I thought was good. The body held from 3s to 55s. The problem was two seconds long.
Cayenne
- EngineeringIf a document contradicts the code, the code wins. The rule I set so that a system built by AI agents couldn’t tell me what I wanted to hear.
- Users4,575 lines of public reviews, none of them my users yet. Frustration number one isn’t the swell, it’s the score: “how can 1 ft be good?”
- DataThe verification stage compared the engine to a copy of itself: it said yes whatever the engine did. A +15.8% bias, invisible.
And then there’s beauty. A complex product no one wants to look at is a product no one understands. Form doesn’t come last to decorate: it’s what makes complexity usable.
All work · 8
- Screens to come
01
Flitter · Wallet
The insurance certificate in Apple & Google Wallet: one object, three problems solved.
Relayed by 3 trade titles the same week
- Screens to come
02
Flitter · Onboarding
Finishing a contract after payment: one step at a time, documents checked on upload.
+20% contract completion
- Screens to come
03
Flitter · Claims × Norauto
One appointment flow across two companies’ interfaces, from the claim itself.
2 companies, 1 flow
- Screens to come
04
Flitter · Design system
Tokens, components and a shared language, so a product team could scale without losing consistency.
1 language, design to code
Short page in progress
05
Duckrew
A surf media built alone, from 13 followers, by treating editorial like a product.
13 → 45,700 Instagram followers in 12 months
06
Duckrew Live
Watching a surf contest together, in French: official stream, live chat, a voice channel anyone can join.
1,226 unique viewers · 175 h watched
Short page in progress
- Screens to come
07
Cayenne
A surf forecast that says when it doesn’t know: three words, never a score. In beta, no users yet.
Beta · 121 spots · 0 users yet
08
Duckrew · Identity
Brand, mascot, visual system and its variations.
1 duck, many hats
Short page in progress