analytics

analytics

Covers measurement, attribution, and reporting for growth systems: web analytics, conversion funnels, cohort analysis, and revenue tracking.

1) GA4 event taxonomy

// Standardised event naming: object_action
gtag('event', 'article_view',    { article_slug, word_count, cluster });
gtag('event', 'email_signup',    { source, form_id, article_slug });
gtag('event', 'purchase',        { value, currency, product_id, funnel_stage });
gtag('event', 'outbound_click',  { link_url, link_text, page_path });
gtag('event', 'scroll_depth',    { percent_scrolled: 75, article_slug });

// Conversion events (mark as Key Events in GA4)
// email_signup, purchase, trial_start, demo_booked

2) Conversion funnel tracking

# Funnel stages — track at each transition
FUNNEL = [
    ("landing",    "user hits any page"),
    ("engaged",    "scroll > 60% or time > 90s"),
    ("intent",     "visits pricing, about, or contact page"),
    ("signup",     "email captured"),
    ("activated",  "completes onboarding action"),
    ("converted",  "purchase or subscription"),
    ("retained",   "returns within 30 days"),
]

# Supabase funnel query
FUNNEL_QUERY = """
WITH events AS (
  SELECT user_id, event_name, created_at
  FROM analytics_events
  WHERE created_at >= NOW() - INTERVAL '30 days'
)
SELECT
  COUNT(DISTINCT CASE WHEN event_name = 'page_view'   THEN user_id END) AS landing,
  COUNT(DISTINCT CASE WHEN event_name = 'engaged'     THEN user_id END) AS engaged,
  COUNT(DISTINCT CASE WHEN event_name = 'email_signup' THEN user_id END) AS signup,
  COUNT(DISTINCT CASE WHEN event_name = 'purchase'    THEN user_id END) AS converted
FROM events;
"""

3) Attribution models

# First-touch: credit goes to first touchpoint
# Last-touch: credit goes to last touchpoint (GA4 default)
# Linear: equal credit to all touchpoints
# Time-decay: more credit to recent touchpoints
# Data-driven: ML-based (GA4 Pro only)

# Simple first/last-touch implementation
def attribute_conversion(touchpoints: list[dict], model: str = "linear") -> dict[str, float]:
    if not touchpoints:
        return {}
    if model == "first_touch":
        return {touchpoints[0]["source"]: 1.0}
    if model == "last_touch":
        return {touchpoints[-1]["source"]: 1.0}
    if model == "linear":
        weight = 1.0 / len(touchpoints)
        result: dict[str, float] = {}
        for t in touchpoints:
            result[t["source"]] = result.get(t["source"], 0.0) + weight
        return result
    raise ValueError(f"Unknown model: {model}")