Analytics changed permanently
Universal Analytics is gone. Third-party cookies are functionally gone in major browsers. iOS App Tracking Transparency is settled.GA4: the new default, with caveats
- Event-based model
- Cross-platform integrated
- BigQuery export available, even on the free tier
- Reports require more configuration than UA
- Data retention shorter than UA by default
- Sampling on large reports
Privacy-respecting alternatives
- Plausible — open-source, cookieless, EU-hosted, simple
- Fathom — similar pitch, hosted
- Matomo — open-source, more featureful, self-hostable
- Server-side analytics — log analytics from server, no client tracking
Server-side tracking — the durable answer
- Client sends events to your server
- Server enriches with first-party data, sends to analytics platforms
- Bypasses ad-blocker effects
- Reduces dependency on the browser sending tracking pixels
- Engineering investment but resilient
Tools: Google's server-side GTM, Stape, Mixpanel CDP, Snowplow, custom.
First-party data is the foundation
- Customer data in your CRM / data warehouse — owned, durable
- Connect identity across web / app / email / support
- Use in audience building, lookalikes, attribution, lifecycle marketing
- Not affected by browser changes
Identity and stitching
- Authenticated user — server-side ID; gold standard
- Cookie ID — works for the visit; degrades over weeks/months
- Device fingerprint — privacy-controversial, partial reliability
- Email / phone match — when available, strongest cross-channel
The cookie banner
- GDPR / KVKK require informed consent for non-essential tracking
- Consent banners impact conversion measurably
- "Reject all" must be as prominent as "Accept all"
- Consent-gated tracking — events only fire after consent
Common analytics mistakes in 2026
- Treating GA4 reports as ground truth without acknowledging gaps
- Comparing 2026 numbers against UA-era 2022 numbers
- Last-click attribution as the only attribution
- Not setting up enhanced ecommerce events properly
- Tracking everything, analysing nothing
- No data retention policy
The metrics that still matter
- Conversion rate by channel
- Customer acquisition cost (CAC)
- Customer lifetime value (LTV)
- Repeat purchase rate
- Cohort retention curves
- NPS or equivalent
Data warehouse as the analytics layer
- BigQuery, Snowflake, Redshift, ClickHouse, Postgres
- Modern data stack: ELT over ETL
- dbt for transformations; Looker / Metabase / Superset for visualisation
- Customer data platform (CDP) — Segment, RudderStack
One pattern we'd warn about
Treating analytics as a pixel that someone installed once. Analytics is a maintained system.One pattern that always pays off
Quarterly analytics review — what events are firing, what are broken, what data is and isn't reaching the warehouse.What's your stack?
