Fika for Substack
An appreciation-based discovery flow for Substack writers and readers: send someone a kind word and get a slice of this week's cake in return.
- Role
- Product, design & engineering
- Timeline
- 6 weeks
- Stack
- React, TypeScript, Node.js, TailwindCSS, Supabase
The problem
Substack's recommendation engine surfaces newsletters algorithmically, but discovery through an algorithm feels transactional, it optimizes for engagement, not for the small human moment of "this writer's work mattered to me." There’s no permanent place where appreciation lives. No wall that says: these are the writers this community actually values, and here’s why.
UX flow
Give a note
User arrives at Fika for Substack and gives an appreciation note for the writer or reader and takes a slice from the cake.
Read and Share
Once the note has been submitted, the user is prompted with a new modal where he can see his note, preview the shareable link or download it as an image.
Share note on Substack
From the preview link of the note, the user is taken back to Substack where he can create a Substack note, copy the appreciation in it and share it with the person who the note was written for.
Browse
One shared table, one round of 12 slices. Two view modes (round cake / slice cards). Below the cake sits the Appreciation Wall where the user can see every slice ever given.
Key decisions & tradeoffs
Single-tenant before multi-tenant
Decision: Build one shared table for my own newsletter first rather than the originally-envisioned per-writer model.
Tradeoff: Faster to validate whether the appreciation concept resonates at all, but defers the actual growth mechanism (cross-publication discovery, ranking system) until a second build phase, so this version can't test network effects yet.
No auth / no identity
Decision: Recipient and sender are free-text fields with no accounts, sessions, or verification.
Tradeoff: Keeps the "one kind word, low friction" gesture instant, but breaks the core promise's other half: recipients can't know a slice was left for them (unless they are notified by the senders), attribution is unverifiable, and there's no return trigger.
No algorithmic discovery
Decision: Explicitly no leaderboards, "most appreciated" rankings, or trending lists, which is a strategic constraint and not an oversight.
Tradeoff: Preserves the "word of mouth, not popularity contest" feel that's central to the product thesis. Cost: no built-in mechanism to surface standout writers/notes beyond manual browsing which makes it harder to demonstrate value at a glance to a new visitor.
Outcome
Shipped and live. In the first 30 days: 199 visitors and 492 page views, up 184% and 115% period-over-period.
The strongest signal is where the traffic came from, not the totals. Looking at referrers, a meaningful share of visits arrived from other writers' Substack newsletters within my network, publications I subscribe to and who subscribe to mine. That's a smaller claim than reaching a totally cold audience, but it's still the loop working as intended: a note traveling through someone else's readers, not just mine. The open question for the next phase is whether it travels further — to readers with no existing connection to me at all.
The individual share links back this up. Each /share/ link is a single appreciation note, and the top ones pulled 21, 9, 7, and 6 visits on their own, solid evidence the core loop (give a note → share it → someone new discovers the writer) works end to end. Reach extended beyond Spain too: 31% of visitors were in the US, with the UK and Canada also represented, without any paid promotion.
Bounce rate sits at 51%, though. /cake is the second most-visited page after the homepage (83 visits), which suggests most first-time visitors are browsing rather than giving a note, a direct cost of the "no algorithmic discovery, no leaderboard" decision above.
Built solo end to end: product framing, UX, UI, and implementation.