GA4

GA4 launched in 2020 and fully replaced Universal Analytics on July 1, 2023. The headline change is the underlying data model. UA bundled behaviour into 'sessions', wrapping pageviews, events and transactions into a single container. GA4 flattens everything into 'events' — pageviews, scrolls, clicks, video plays and purchases all live in the same event structure. That redesign made GA4 the first free tool that could meaningfully unify web and app analytics in a single property.

Three things make GA4 genuinely useful for marketers. First, free BigQuery export of raw event data — that capability used to require GA360, which ran into six figures annually in USD. Second, deeper integration with Google Ads, which strengthens attribution modelling and automated bidding. Third, ML-based conversion modelling that fills in conversions cookies can no longer observe. The trade-off on the last point is that GA4 numbers will not match ad-platform numbers cleanly, and the gap is sometimes large.

Common GA4 onboarding traps: the UA 'reporting view' concept is gone, so data separation gets fiddly; custom dimensions and event parameters have hard limits and are awkward to redefine after the fact; and the UI looks nothing like UA, which forces a real learning curve on both marketers and analysts. In Korea, the most frequent early complaint was that GA4 'undercounted traffic compared to UA' — almost always traced back to confusion between UA's unique pageviews and GA4's event count.

AI search exposes a real limit in GA4. When ChatGPT, Perplexity or Claude cite your brand and a user clicks through, that visit gets bucketed as organic search or direct, and the actual originating channel — the AI answer — is invisible without custom tracking. Some teams now customise GA4 channel grouping to break out chatgpt.com, perplexity.ai and claude.ai as separate channels, but users who view an AI answer and return via direct search days later still slip through. This is a big part of why media mix modelling, incrementality testing and 'how did you first hear about us' survey questions are seeing a renaissance alongside GA4.

Two recommendations to get more out of GA4. First, turn on BigQuery export from day one — data only starts accruing once it's enabled, so flipping the switch a year later means permanently losing that first year for any historical analysis. Second, don't rely on the default GA4 dashboards. Build your reporting in Looker Studio or BigQuery plus Tableau against your specific KPIs. GA4's stock UI is built for generic analytics and rarely fits SaaS or e-commerce models cleanly without customisation.