Knowledge Transfer
Everything you need to understand, question, and lead Hobbeet - from vision to code to launch.
How to use this page
One elevator ride to explain Hobbeet
Hobbeet connects people through shared hobbies. Want to play tennis Saturday? We find you the right partner - matched by skill, personality, schedule, and values. After meetup, you build trust and a personal hobby journal. Best connections happen through shared experiences.
Not a dating app
Match for activities, not romance. Think 'hobby LinkedIn'.
Not Strava / AllTrails
We don't track runs. We find the right running partner.
Trust is earned
No streaks, no FOMO. Return because you want to.
Everything serves one of these purposes
Matching - Find the right people
8-component algorithm, 0-100 scoring. Our competitive moat.
Engagement - Seamless experience
Create -> join -> message -> coordinate -> meet. Every step effortless.
Retention - Reasons to return
Sanctuary (journal), Bloom (trust), Connections (partner network), Feedback (learning loop).
Three ways users interact - each ends with a different retention hook
A) Create & Host
You have a plan - find the right people for it.
Create Hobbeet
1 minTap 'Create' -> pick hobby -> set title, date/time, neighborhood, skill preference -> 1:1 or group.
Publish & Wait
PassiveYour hobbeet enters the discovery feed, visible to compatible users ranked by match score. You get notified as join requests arrive.
Review & Approve
Per requestFor each request: see profile, match score breakdown, trust tier, strengths & considerations. Approve or decline. Approval -> chat channel auto-opens.
Coordinate & Meet
The momentChat to finalize details (exact location, gear). Then meet up - the magic moment everything is built for.
Complete -> Bloom
30 sec BloomGive your partner an Uplift (trust boost) or Downlift. Their Bloom tier grows from community endorsements. Your tier reflects how others rate you.
B) Get Matched
Someone's activity fits you - the algorithm found it.
Match Notification
Push notifAlgorithm found a high-score hobbeet for you. Notification: "Perfect tennis match nearby - 87% compatible!"
View Compatibility
Quick scanFull breakdown: shared hobbies, personality alignment, schedule overlap, strengths & considerations.
Send Join Request
One tapTap 'Join' -> creator sees your profile, match score, trust tier. They approve or decline.
Chat & Meet Up
The momentApproved -> chat channel opens. Coordinate details, then meet in person.
Complete -> Connections
Automatic ConnectionsYour connection with this partner is recorded per hobby. "You and Alex: 3 tennis sessions." Only accumulates, never decays.
C) Discover & Join
Browse what's happening - great for group hobbeets.
Browse Discovery Feed
BrowseScrollable feed of hobbeets nearby, ranked by your match score. Filter by hobby, date, group size. Cards show hobby, time, location, creator's Bloom tier, compatibility %.
Explore Details
ExploreTap a card -> full view: description, context (intensity, goals), member list, your match breakdown, open spots.
Join Group Hobbeet
One tapSend join request. For groups, multiple people join. Creator approves each based on score and group fit.
Participate
The momentJoin the group - book club discussion, group hike, board game night. Meet new people through shared experience.
Complete -> Sanctuary
2 min SanctuaryCapture the memory: write a moment (reflection + learnings), save the place, update an intention. Your personal hobby journal grows.
All journeys converge here
Bloom
Every hobbeet feeds trust
Connections
Every partner tracked per hobby
Sanctuary
Every experience -> a memory
Loading...
8 components, 100 points - click any to deep-dive
Weight Distribution
Smart Weight Adaptation
New users with little data get engagement/behavioral weights auto-reduced and redistributed to core signals. Weights adapt per platform phase: Early (<100), Growing (100-10K), Mature (10K+).
Bloom, Sanctuary, Connections
Bloom - Trust Reputation
Earned via peer feedback. Users see tier, never exact points. Randomized to prevent gaming.
Uplift
+15 to 25 pts
Downlift
-5 to 15 pts
Sanctuary - Hobby Journal
Each hobby gets its own Sanctuary. Three components:
Moments
Journal entries: reflections, learnings, photos.
Places
Discovered locations: courts, trails, cafes.
Intentions
Personal goals. Track active vs achieved.
Connections - Partner Network
Every completed hobbeet creates/strengthens a per-hobby connection. Tracks: who, how many shared moments, first/last activity. Visible graph: "you and Sarah: 12 tennis sessions." No decay - only accumulates.
How it's built - no code, just concepts
Tech Stack
Backend
Python FastAPI
Database
PostgreSQL (Supabase)
Auth
Supabase JWT
Hosting
Railway
Chat
Stream Chat
Mobile
SwiftUI + Kotlin
Analytics
PostHog
ML
HuggingFace
Data Flow - How a Match Happens
User opens feed -> API call with auth token
Backend fetches published hobbeets near user's city
Matching algorithm runs per hobbeet: creator's persona + user's persona
8 components scored, weights applied, total 0-100
Sorted by score, blocked users removed, paginated
Mobile renders cards with match % and summary
Tap -> full breakdown + strengths/considerations
Database - 20 Tables
Core (5)
users, hobbies, user_hobbies, hobbeets, hobbeet_members
Onboarding (4)
user_personality, user_lifestyle, user_values, user_availability
Sanctuary (4)
hobby_sanctuaries, moments, places, intentions
Trust (4)
user_bloom, hob_sharing, connections, feedback
Safety (3)
reports, user_blocks, notifications
What we measure and what good looks like
Why These Metrics?
These KPIs validate our core hypothesis: quality matching leads to real-world meetups, which drive retention. Each metric maps to a stage in the user journey and helps us identify where the funnel breaks.
Acquisition
Onboarding completion, profile quality
Activation
Match accept rate, first hobbeet
Retention
D7/D30, repeat hobbeets, connections
North Star Metric
Completed Hobbeets per Week
Captures entire funnel: create -> match -> coordinate -> meet up. Single metric that indicates product-market fit.
Beta Hypotheses to Validate
Match score predicts satisfaction
Users with 80+ scores report "would repeat" at 2x rate of <60 scores
Personality matching matters more than skill
Personality component correlation to "would repeat" > skill level correlation
Bloom tier improves trust signals
Higher Bloom tier users get 30%+ higher accept rates on join requests
Connections drive repeat behavior
Users with 3+ connections have 2x D30 retention vs users with 0-1
Cohort Segmentation
Key segments we'll analyze during beta:
By Onboarding Completeness
Full profile vs partial vs minimal
By User Role
Creators vs joiners vs both
By Hobby Category
Sports vs creative vs social
By Match Score Band
Excellent vs good vs possible
Early Warning Signals
Accept rate <20% -> Algorithm needs tuning
Completion <30% -> Flaky users or bad matches
D7 ret <15% -> Not delivering value
Avg msgs <5 -> Chat UX or coordination friction
Uplift ratio <70% -> Negative experience trend
Report rate >5% -> Safety/trust issues
Data Collection Strategy
Implicit Signals
Profile views, time on match card, join request timing, message patterns, session duration, feature usage frequency
Explicit Feedback
Post-hobbeet ratings, "would repeat" binary, positive/negative tags, Bloom uplift/downlift choices, NPS surveys
Statistical Note
With 20-30 beta users and ~50 hobbeets, we'll have directional signals but not statistical significance. Focus on qualitative patterns and user interviews to complement quantitative data.
The Funnel
Track conversion rates at each step. Biggest drop-off = highest leverage improvement.
Loading...
Mobile Screens - Backend Ready?
Native iOS (SwiftUI) + Android (Kotlin). No React Native.
Auth
Done
Onboarding
Done
Feed
Done
Hobbeet Detail
Done
Create
Done
Chat
Pending
Profile
Done
Sanctuary
Done
Notifications
Done
Settings
Done
Phased approach to launch
Open questions needing your input
You're caught up, Komal
This is a living document. Question the weights, the tiers, the priorities, the launch plan. That's exactly what this is for.