Google Analytics

UA tracked sessions and pageviews as core events. GA4 tracks all interactions as events (pageview, click, scroll, video_play, form_submit). UA is session-centric; GA4 is event-centric. UA sunset Jan 2024 forced migration. GA4 model is more granular but requires custom event setup. Machine learning in GA4 (predictive analytics, anomaly detection) is native; UA didn't have it. Migration is the biggest challenge practitioners face; many lost historical data because event mapping wasn't planned.

Yes, if you do serious analysis. GA4 UI limits: 100k rows max, limited cohort overlap, no true free-form SQL. BigQuery export (available in GA4 360 or as standard export) unlocks unlimited data access. Cost: approximately $7 per GB scanned (cheap at scale). You can join GA4 events with CRM data, attribution models, product catalogs. Query pricing is fixed; use SELECT sparingly. Most mid-market companies export to BigQuery plus Looker Studio for dashboards.

EU/UK regulation (GDPR plus UK-ICO guidance 2024) requires explicit consent BEFORE collecting analytics cookies. Consent Mode v2 (launched 2024) lets you fire tags conditionally: if user denies analytics consent, GA4 still runs in 'consent denied' mode (behavioral modeling, no PII). Google reimburses lost data via machine learning. Setup: declare consent state (analytics_storage equals 'denied' or 'granted'), fire tags conditionally, let GA4 infer conversions from modeled data. Non-compliance equals Google blocking tags plus GDPR fines.

GA4 default is last-click attribution (final touchpoint gets credit). Problem: ignores early awareness channels (YouTube, display ads). GA4 advanced attribution models: time-decay (more weight to recent), linear (equal credit), first-click, position-based (40% first, 20% mid, 40% last). Enterprise: use data-driven attribution (ML balances channels based on historical conversion patterns). Setup: Admin, Attribution Settings. For serious multi-channel work, export to BigQuery plus use propensity modeling or incrementality tests to isolate true channel impact.

Traditional GTM runs client-side (JavaScript in browser). Ad blockers plus iOS privacy features block it. Server-side GTM: you host a server (Google Cloud Run or similar), browser sends events to YOUR domain (example: events.yourdomain.com), server forwards to GA4/Segment/other tools. Benefits: ad-blocker resistant, more reliable, better latency control, can enrich data on server before sending downstream. Downside: you manage server cost plus infrastructure. Stripe, Shopify, content sites use this. Setup: Firestore backend plus GTM server-side container in Google Cloud. Learning curve is weeks.

GA4 is free, ubiquitous, best for web traffic plus ecommerce plus SEO ROI. Mixpanel/Amplitude equals behavioral analytics, user funnels, cohort analysis, retention tracking with better UX. PostHog equals product analytics plus feature flags plus session replay, engineer-friendly, open-source option. Mixpanel: $800+/mo, strong mobile. Amplitude: $1500+/mo, predictive analytics. PostHog: $500+/mo or self-hosted free. Choose GA4 if: you need cheap, you care about SEO/traffic. Choose Mixpanel/Amplitude if: product retention/activation is your KPI. Choose PostHog if: you want feature flags plus analytics in one tool.

Custom dimensions equals custom properties you send to GA4. Example: user_tier (free/pro/enterprise), subscription_duration, content_topic, ab_test_variant. Setup: declare dimension in Admin, Custom Definitions, then fire events with event parameters or user properties. Limits: GA4 Standard equals 25 custom dimensions, GA4 360 equals unlimited. Use for segmentation: 'show me conversion rate by user_tier' or 'bounce rate by device_brand'. Don't over-define; every dimension you add equals more cardinality in reporting, slower queries. Plan dimensions before launch.

Tools: (1) Google Analytics Debugger browser extension shows events in real-time. (2) GA4 Real-time report (Admin, Real-time) shows events 1-3s after firing. (3) Google Tag Assistant (Chrome extension) checks tag firing. (4) Check browser Network tab for collect requests. Common issues: event name typo (events are case-sensitive), parameter name wrong, consent blocking tag, ad blocker killing requests. Debug mode in GTM equals tag-level logging. Pro tip: create a test view (filter by test user ID) to avoid polluting production data while testing.