Back to notes

October 10, 2026 · 5 min read

Stop Being Nostalgic for Universal Analytics, It Was Worse

Every GA4 complaint I hear ends with 'UA never did this to me.' UA did plenty of worse things, people just forgot because they stopped looking closely.

GA4Hot Take

I hear the same line in almost every GA4 complaint session: 'this never happened in Universal Analytics.' It's said with total confidence, and it's usually wrong. UA had a session model that silently fragmented on campaign parameter changes mid-session, bounce rate that misled an entire industry about engagement for a decade, and a sampling problem on any report of real volume that was, if anything, worse than GA4's thresholding.

The actual difference isn't that UA was more reliable. It's that people had years to build mental models around UA's specific quirks, learn to route around them, and stop questioning numbers that had always looked a certain way. GA4 is new enough that every quirk still feels like a betrayal instead of a known cost of doing business with the tool.

The UA problems people conveniently forgot

  • Session-scoped campaign attribution that reset mid-session on a parameter change, overcounting sessions in ways nobody ever fully reconciled.
  • Sampling kicked in on any report over 500k sessions at the sampled property level, with no equivalent to BigQuery raw export for most accounts.
  • Bounce rate as a metric actively encouraged bad UX decisions — single-page experiences got penalized for being good at their job.
  • Multi-channel funnel attribution models were a black box nobody could audit, arguably worse than GA4's data-driven model documentation today.
Nostalgia for Universal Analytics isn't nostalgia for a better tool. It's nostalgia for not having to think about the tool anymore.

What's actually fair to criticize about GA4

GA4 deserves real criticism — the interface still buries key reports behind too many clicks, the default event-based model is genuinely harder to reason about for people used to pageview-centric thinking, and thresholding's lack of a visible flag is a real design failure. Those are worth fixing. 'It's not UA' is not a real critique, it's a preference for familiarity dressed up as a technical complaint, and I stop taking it seriously the moment someone can't name a specific UA behavior they'd actually want back.