Attribution & UTM Pipeline

How marketing spend flows from content/ads through UTM parameters to install attribution and back into optimization loops.

Data Flow

Blog/Ad/Reddit Post
   ↓ UTM-tagged links (utm_source, utm_medium, utm_campaign, utm_content)
App Store / Play Store listing
   ↓ Install referrer captured by PostHog SDK (captureDeepLinks=true)
PostHog events table (deep_link_opened)
   ↓ Weekly HogQL queries via attribution_feedback.py
marketing/keywords/posthog_feedback.json  →  ASO keyword rotation
marketing/data/content_feedback.json      →  Content topic selection
marketing/data/attribution-report.md      →  This dashboard

UTM Tagging Convention

All outbound links in blog posts are tagged by growth_content_pipeline.py → add_utm():

Parameter	Value	Example
utm_source	Platform name	github_pages, devto, linkedin, twitter
utm_medium	Channel type	organic, paid, referral
utm_campaign	Date-stamped	daily_blog_20260220
utm_content	Keyword slug	daily_blog, reaction_training_app

PostHog Queries (run weekly)

1. UTM Attribution

SELECT properties.utm_source, properties.utm_medium,
       properties.utm_campaign, properties.utm_content,
       count() AS installs, count(DISTINCT person_id) AS unique_users
FROM events
WHERE event = 'deep_link_opened'
  AND timestamp > now() - interval 30 day
GROUP BY source, medium, campaign, content
ORDER BY installs DESC

2. Onboarding Funnel

Queries first_open, first_timer_configured, first_timer_completed events to compute:

• Open → Configured rate: % of new users who set up a timer
• Configured → Completed rate: % who run a timer to completion
• Open → Completed rate: Full funnel conversion

3. Campaign Performance with Activation

SELECT properties.utm_campaign, properties.utm_source,
       count(DISTINCT person_id) AS attributed_users,
       countIf(person_id IN (
           SELECT DISTINCT person_id FROM events
           WHERE event = 'first_timer_completed'
       )) AS activated_users
FROM events WHERE event = 'deep_link_opened' ...

Feedback Loops

→ ASO Keyword Rotation

posthog_feedback.json maps utm_content values to install counts. The weekly aso_keyword_rotation.py uses these as real performance signals instead of simulated rankings.

→ Content Pipeline

content_feedback.json ranks campaigns by activation rate. The daily growth_content_pipeline.py uses this to prioritize topics that drive actual installs.

→ Campaign Budget

attribution-report.md shows campaign ROI. Manual review informs budget reallocation between Apple Search Ads and Google UAC.

Automation Schedule

Workflow	Schedule	What it does
weekly-attribution-feedback.yml	Sunday 08:00 UTC	Runs PostHog queries, writes feedback JSON, commits to repo
weekly-aso-rotation.yml	Monday 10:00 UTC	Consumes feedback, rotates iOS keywords
daily-growth-publishing.yml	Daily 13:15 UTC	Publishes content with UTM tags

Required Secrets

Secret	Purpose
POSTHOG_PERSONAL_API_KEY	Bearer token for PostHog HogQL API
POSTHOG_PROJECT_ID	PostHog project to query

Without these secrets, the pipeline generates empty feedback files so downstream scripts don't break.

Source Files

• scripts/attribution_feedback.py — Main pipeline
• scripts/growth_content_pipeline.py → add_utm() — UTM tag builder
• .github/workflows/weekly-attribution-feedback.yml — Automation