Onboarding Funnel
PRODUCTION SUITE · MIXPANEL JQL (ID: 4015502), APPSFLYER UNIFIED LTV & POSTGRES DB
All Communities
to
Onboarding Starts
--
Mixpanel: onboarding_started
Account Signups
--
Mixpanel: signup_completed
Onboarding Completions
--
Mixpanel: onboarding_completed
Total Conversion Rate
--
Signed up vs Completed onboarding
Onboarding V5 Conversion Funnel
Sequential per-user deduplicated funnel starting at Onboarding Started. Includes explicit skips ⚙️.
Social Auth Conversion & Drop Off
Visualizing user actions on the Social Setup screen: Google vs Apple button taps, returning user logins, and setup screen drop-off.
Setup Screen Visitors: --
1. Clicked Sign Up (Google/Apple)
--
Google + Apple button taps
2. Existing Users Logged In
--
Mixpanel: Logged In
3. Did Nothing & Dropped Off
--
Left Setup screen without auth
Production Postgres DB vs Mixpanel Audit
Comparative audit showing stages tracked in both Postgres Production DB and Mixpanel.
| Metric / Stage | Mixpanel Unique Users | Postgres Production DB Count | Status / Variance |
|---|
▶
AppsFlyer Unified LTV In-App Events (Live API)
LIVE APPSFLYER LTV
Real-time in-app event counts and unique user metrics from AppsFlyer Unified LTV Dashboard (iOS & Android). Click header to expand.
AppsFlyer Clicks
--
OneLink & Campaign Clicks
AppsFlyer Installs
--
Attributed Installs
AppsFlyer Loyal Users
--
High Engagement Users
Click → Install Rate
--
AppsFlyer Conversion Rate
AppsFlyer In-App Event LTV Breakdown
| AppsFlyer In-App Event | Unique Users (LTV) | Total Event Occurrences | Avg Events / User |
|---|
AppsFlyer Campaign & Media Source Attributions
| Media Source | Campaign Name | App Platform | Clicks | Installs | Loyal Users |
|---|
DASHBOARD FILTER · INVITE_TYPE
Vouch converts at every stage and propagates deepest — but volume is low, so K sits just under 1. This is the loop worth growing.
HEADLINE K-FACTOR
0.92
below 1 — the deep loop
K = i × c
i = 1.8 invites / activated user
c = 0.51 invite→activated
Viral cycle time
2.1 days
install → first invite sent
K TREND · WEEKLY
dashed line = K threshold 1.0
INVITES SENT (SHARES)
--
DB user_safety_networks (sent)
LINK CLICKS (CTR)
--
CTR --
CLICK→INSTALL
--
install→activate --
CONVERSION DECOMPOSITION · C = CLICK × INSTALL × SOCIAL AUTH TAP × ACTIVATE
where the invite chain leaks, by stage (100% Postgres DB Grounded)
ANALYSIS GUARDRAILS & QUALITY GATES
D1 RETENTION · INVITED VS ORGANIC
44%
vs 52%
Invited traffic starts weaker on day 1.
D7 RETENTION · INVITED VS ORGANIC
19%
vs 31%
Gap widens by day 7 — quality problem.
REINVITE RATE · GEN 1 (QUALITY GATE)
29%
The K multiplier only holds if this holds.
INVITE → CHURN WITHIN 7 DAYS
61%
of invited
Most invited users churn in a week.
CHART
Metabase dashboard: a big-number scorecard (K), a line "trend" question, scorecards, and a bar question for the decomposition. All wired to one dashboard filter.
SOURCE
SQL over warehouse. K = count(activated users attributed to cohort) ÷ cohort size. 100% grounded in Postgres
referral_invite_events (created, clicked, signup_button_clicked, activated).SPLIT BY
Dashboard filter →
WHERE invite_type = {{invite_type}}. Never ship a single blended K — always expose vouch vs public_share.
DASHBOARD FILTER · EDGE TYPE (HOW THEY WERE INVITED)
Did the invited also invite?
Headline tile
Reinvite rate by generation
users in gen who sent ≥1 invite ÷ users in gen — the anti-growth-theater metric
seed
Gen 0
n = 500
34%
Gen 1
n = 2,300
9%
Gen 2
n = 720
2%
Gen 3
n = 90
Blended, reinvite rate collapses from 34% → 9% → 2% after Gen 1. Split by edge type to see it is the public-share path dragging it down.
GENERATION WATERFALL
activated users per hop · Gen 0 = seeds
CHAIN DEPTH DISTRIBUTION
how many hops the longest chains reach
Amplification / seed
7.2
avg downstream per Gen-0
Median / seed
2
most seeds barely amplify
METRICS PER GENERATION
| GENERATION | USERS ACTIVATED | REINVITE RATE | INVITES / USER | CONVERSION OF THEIR INVITES | EFFECTIVE K (GEN N→N+1) |
|---|
REFERRAL LINK & USER ATTRIBUTION EXPLORER
click any invite code to view its 7-stage chronological story timeline
| INVITE CODE | TYPE | INVITER | INVITEE | GEN | INVITEE STAGE | DOWNSTREAM INVITES | ATTRIBUTED AT | ACTION |
|---|
ANALYSIS GUARDRAILS & QUALITY GATES
🔀 Step 13 Intent Fork Architecture: My City vs. Traveller
Branching Flow
Territory flow splits post-interests: users choose whether to join their local city gate or explore global travel communities.
1. Core Shared Funnel Progression (Pre-Auth → Intent Fork)
13-Step shared foundation: Welcome → Anything (Location) → Live Now → Social Auth → Account Created → Phone → Verification → Name → Username → Birthday → Photo → Interests → Intent Fork
2. Branching Flow Comparison: "My City" (Gate) vs. "Traveller" (Community)
Side-by-side sequential drop-off for the two post-intent paths
📍 Branch A: "My City" (Invite Gate)
0 users
✈️ Branch B: "Traveller" (City Community)
0 users
3. Comprehensive Button Tap Analytics (42 Tracked Controls)
Audited interaction rates for every primary CTA, choice card, permission prompt, and back navigation in onboarding_territory
4. Traveller Community Onboarding Deep-Dive
Step-by-step engagement across destination join, dates selection, greetings sent, and intro publishing
5. Territory Social Auth Conversion & Drop Off
Breakdown of auth button taps (Apple / Google), account creation, returning logins & drop-offs on Step 4
6. Location Permission Prompt Response
Breakdown of location permission grants vs. denials on Step 2 (Anything screen)
7. Invite Gate Performance & Territory Virality (K-Factor)
Viral expansion metrics for the 5-friends invite gate (Branch A)
GATE PROGRESS MILESTONE DISTRIBUTION
8. Territory Resolution Signal Quality
Breakdown of client signals used to assign territory flow (Storefront vs SIM vs Locale vs Fallback)
North star · growth
+0.07
0.99
Effective K
Read from Invite Link Clicked forward. Raw K 1.34 inflates on broadcast shares.
target 1.00
North star · company
+11%
1,284
Velio installs / week
Cross-app attributed through AppsFlyer OneLink and reconciled in warehouse.
target 2,000
Tier 1 · engine
+2.4pp
38%
Weekly-loop completion
Voting Completed through Map Opened, in one window.
target 50%
Tier 2 · thesis
−3pp
61%
Check-in conversion
Check In Completed of Attendance Intent = going. The honest denominator.
target 70%
Master weekly loop
Mixpanel funnel · 7-day window
Open → vote → unlock → verdict → map → check in. The last two steps are the thesis.
K-chain — five stages
Warehouse · joined on invite_code
Effective K is read from Invite Link Clicked forward. Raw K 1.34 — the gap is broadcast share inflation.
Invites / user
—
Viral cycle time
—
Projected next cycle
—
Unlock leak
invite : pay
Share of paid skips vs invited unlocks over time
Pay-to-skip share has climbed +14pp in six weeks. Every paid unlock is an invite the chain never sent — the leak reads directly against K.
Invite gate
Mixpanel funnel · to unlock
Gate → send → wait → join → unlock. The wait is the riskiest moment in the loop.
Cycle retention
Warehouse · own cycle_start_date
Voted in window 1 → returned and voted in window N. Never blended.
Invited retention falls off a cliff after window 1 — borrowed motivation, not a habit. The fix is the invited first-window experience, not more top-of-funnel.
Notification lift
Holdout · 5% per push_type
Return-and-vote within 24h, sent vs. held out. Lift is the only honest column.
pop_daily_reminder posts negative lift on 52,700 sends — it is training people to mute the app. Kill it and reallocate to squad-progress, the only push earning above +19pp.
Velio conversion by source screen
Click → install, of active users seeing the CTA.
Where Popped converts 5.2× better than the leaderboard. The CTA belongs on the reveal, not in navigation.
Instrumentation health
Blocks the numbers above
Every metric on this page depends on these. A red row means the number it feeds is wrong, not missing.
Live Mixpanel & Postgres Engine Status
● CONNECTED TO POPTONIGHT (MIXPANEL ID: 4053033)
Real-time event queries from PopTonight Mixpanel Project (ID: 4053033) and Postgres pop_* cluster tables
Cross-App Retention & Churn · Pop Tonight × Velio Social
Live ecosystem tracking across 10,845 total users. Measures exact last-login and voting/posting recency to segment dual-app members into 4 standardized retention quadrants with a 30-day lifecycle threshold and 14-day intervention window.
Live Sync Active
4-Quadrant Ecosystem Retention Matrix
30-Day Activity Threshold
Distribution of all 22 dual-app users across cross-app engagement combinations.
Dual-App User Recency & Engagement Directory
All 22 members with records in both Pop Tonight and Velio Social.
| User / ID ↕ | Origin Flow ↕ | Pop Tonight Recency ↕ | Pop Tonight Engagement | Velio Social Recency ↕ | Velio Engagement | Churn Quadrant & Risk ↕ |
|---|
Segmented Branch Funnels & Dropoff Flows
Zero-Blended Virality Invariant
Direct comparative dropoff flows isolating Branch A (Gate / "My City" Cohort) from Branch B (Traveller / Destination Cohort) to pinpoint where cohort intent dies.
What people search for — preview vs. join intent
Communities searched, showing % of completed onboarding who viewed preview vs % who tapped join.
Comparing preview inspection against join tap: high preview rates with low join taps identify communities where feed gating or preview friction suppresses conversion.
Where intent dies after the tap
Two failures that look like disinterest in the funnel and are not.
Communities joined outside the home country
Searched, then joined
Joins into a community whose country is not the member's own. This is the demand signal for where to seed next.
Read the last two columns together: Marrakech is joined late and rarely converts, Detty December is joined within days and converts a third of the time. Seasonality, not size, is driving the join.
The 5-Rung Maturity Ladder
Share of all in-community actions by ladder rung. Healthy communities push users downward over time.
Reach vs. Intensity Map
Horizontal reach against vertical intensity, sized by action volume, colored by rung
Action Velocity & Conversion
How effectively intent converts into formed plans and booked seats
Top-line conversion: explore → booked
—
itinerary-viewed → booking_completed across all overlap cohorts in 30-day window.
6-Step Overlap to Booking Funnel
explore_itinerary_viewed → overlap_chart_viewed → activity_detail_viewed → activity_added_to_itinerary → booking_started → booking_completed
Booking Conversion by Overlap Bucket
Confirm or Kill
If this curve is flat, the overlap chart is decoration. Bar chart, not a table, so flatness is impossible to miss.
Windows by Number of Co-Bookers
Upcoming activity windows broken down by member booker count
Cold Upcoming Windows
1 Booker · Expiring Solo
Windows at risk of expiring solo without second booker conversion
Overlap Degree Distribution
Members by count of others with overlap_pct > 0 on their current trip
Overlap Asymmetry Map
Directional overlap for top pairs: Viewer's % vs Target's % of stay
Surface Vitality & Canary Signals
Weekly resolution tracking inputs across Feed, Explore, and Itineraries
Invites sent per booker, 14 days post-booking
Co-booked trips should be more viral — you met people. If the two bars converge, the community is not the reason anyone invites.
Week-8 retention by community density at signup
Members acquired into dense communities against those acquired into sparse ones. Same acquisition channel, different destination.
K-Factor Loop Integrity (Strict Workspace Invariant)
Zero blended K-factor — uncompromised segmentation by invite_type
5-Step Post Creation Funnel
Post Composer Opened → Post Submit Tapped → Community Post Created → Post Pinned → Conversation Opened
Plan Mortality & Moderation Friction
Lifecycle distribution of formed plans and creator warning telemetry
Posts Per Creator Distribution (30d)
Creator post concentration
Plan Field Edit Rate & Time-to-Form
Extraction accuracy and p50/p90 creation velocity
Circle Strength Distribution
Inviters categorized by successful referral connection count
4-Step Viral Unlock Funnel
Invite sent → invite accepted → posting unlocked → first plan posted
Time to First Plan Posted by Acquisition Path
Percentiles (p50 / p90) and unlock rate across user cohorts
Publish Rate by Reach Bucket & Toll Economy
Potential reach configured at creation & ledger balance
Share Artifacts & Destination Funnel
Distribution across plan objects, stickers, and relay links
Headline K-Factor & 8-Week Trend
K = i × c breakdown and viral velocity trendline
Unblended Layers & Viral Conversion Funnel
Strictly unblended K-factor: Vouch vs. Public Share (AGENTS.md §3)
Generational Invite Depth (Tree Diffusion)
Distribution of signups by viral generation depth (Gen 0, 1, 2, 3+)
Viewing to Coordination Funnel
Plan viewed → chat joined → attendance confirmed (Filtered by hop distance)
Chat Engagement & Attention Telemetry
Weekly surface taps trend and messages per plan distribution
Login & Coordination-Active Retention Drop-Off Curve
Days from signup (D0 to D30) for live cities / Weeks (W0 to W8) for waitlist
Weekly Signup Cohort Retention Heatmap Matrix
Real login return rates from W0 to W8 across historical signup cohorts
Overlaid Retention Curves by Downstream Chain Depth
User retention lift segmented by the depth of their downstream invite tree
Connector Position Retention
Retention by position in other users' invite trees
Recursive SQL CTE Architecture
Query engine used to evaluate downstream tree depth
Ranked Market Density Leaderboard
Active city presences ranked by user concentration and cold-map frequency
Sub-Area / Ward Breakdown (Lagos Hub)
Hyperlocal density and cold-map rates by ward
Revealed Interest Rankings (Selected Country)
Actual plan creation and participation categories
Declared vs. Revealed Intent Gap Analysis
% declared at signup vs % of active plans created
5-Step True North Coordination Funnel
Plan viewed → Joined while forming → Still in at formed → Attendance confirmed → Plan happened
Plan Resolution & Category Attendance
Confirmed outcomes, expiry rates, and category fill rates
Time-to-First-Coordination
Days from activation to first attended plan (p50 / p90)
Attended by Hop Distance
Social distance distribution of real-world attendance