Chapter 29 — Case Study 1: The Day the Baseline Broke (Universal Analytics → GA4, 2022–2023)

A real, public event. The migration from Universal Analytics to Google Analytics 4 is a documented, industry-wide change every measurement professional lived through. The dates and mechanics below are public record; where this study reconstructs a typical reporting experience, it is labeled, and no specific statistic is invented.

Background

For roughly a decade, "web analytics" for most of the world meant Universal Analytics (UA) — Google Analytics' session-and-pageview model, the tool in which a generation of marketers built their reports, their baselines, and their year-over-year stories. Then Google replaced it. Google Analytics 4 (GA4) grew out of a 2019 "App + Web" beta and became the new default in October 2020. On March 16, 2022, Google announced that standard Universal Analytics properties would stop processing new data on July 1, 2023 (Analytics 360 properties a year later), and that UA's historical data and interface would eventually be retired. The clock was public and non-negotiable: migrate to GA4, or lose measurement.

This was not a cosmetic rebrand. GA4 is a fundamentally different measurement model. UA counted sessions, pageviews, and hits; GA4 counts events — every interaction, from a page view to a scroll to a form submit, is an event, and a "conversion" (later renamed a key event) is simply an event you flagged as important. GA4 retired familiar metrics (bounce rate gave way to engagement rate and engaged sessions), changed how sessions are defined, made data-driven attribution its default, added consent mode and privacy-driven data thresholds that can withhold some detail, and — critically for reporting — did not carry UA's historical data forward. Unless an organization had run GA4 in parallel with UA well before the cutoff, the day UA stopped processing data was the day their old baseline and their clean year-over-year comparison broke.

The reporting issue

SEO and analytics reporting rests on a foundation most reporters never think about until it moves: the measurement infrastructure underneath the numbers. The UA-to-GA4 migration yanked that foundation, and it exposed three fault lines this chapter is built around.

  • Baselines evaporated. Every report that compared "this year vs. last year" assumed a stable measurement system. Teams that had not run GA4 alongside UA for a full prior year suddenly had no comparable history — their frozen baseline (§29.1) was recorded in a system that no longer produced new data, and GA4's numbers were not directly comparable to it.
  • The numbers didn't match — and couldn't. Because GA4 counts differently from UA (event-based sessions, different session timeouts, consent-gated collection, data thresholds), the two tools reported different figures for the "same" traffic, often noticeably. Reporters who expected them to reconcile looked as if something were broken; reporters who understood the models could explain the gap — exactly the §29.3 lesson that GSC and GA4 disagree by design.
  • Measurement literacy stopped being optional. "Conversions" now meant "key events you configured"; "engagement" replaced "bounce"; attribution defaulted to data-driven. A reporter who kept using the old vocabulary on the new tool would mislabel the very KPIs (§29.1) their stakeholders depended on.
FIGURE CS29.1 — "What changed under the report"                 [reconstruction — illustrative shape]
  DIMENSION            UNIVERSAL ANALYTICS (UA)      GOOGLE ANALYTICS 4 (GA4)
  Core model           sessions · pageviews · hits   events (everything is an event)
  "Conversion"         goals                          key events (flag any event)
  Engagement metric    bounce rate                    engagement rate / engaged sessions
  Attribution default  last non-direct click          data-driven
  History on switch    stayed in UA                    NOT imported — starts fresh
  Privacy controls     limited                         consent mode + data thresholds
  Net effect on a      a stable baseline and clean     a NEW baseline; year-over-year broken
  report               year-over-year                  unless GA4 ran in parallel early

What it shows

The migration is the cleanest real-world demonstration of a claim that sounds abstract until it happens to you: your report is only as stable as the tool beneath it, and you do not own that tool. Three chapter lessons land hard here.

First, a baseline is infrastructure, not a formality (§29.1). The teams that came through the migration with their reporting intact were the ones who had set up GA4 early and let it accumulate a parallel history — they had a real starting line in the new system before the old one went dark. The teams that waited until the deadline lost their year-over-year story for a full cycle. Capturing and protecting a baseline is not box-ticking; it is the difference between a survivable platform change and a blind one.

Second, tool discrepancies are normal and must be explained, not hidden (§29.3). The professionals who kept their stakeholders' trust were the ones who could say, plainly, "GA4 and UA count differently, here's why the numbers moved, and here's which one we now report" — rather than pretending nothing changed or quietly swapping the source with no note.

Third, honest reporting includes 'the ground shifted under us' (§29.7). A responsible report during the transition said so: "Note: this month's figures come from GA4, which measures differently than last year's UA data; treat the year-over-year comparison as directional, not exact." That is not a weakness in the report. It is the caveat that keeps the report credible.

Outcome

Universal Analytics standard properties stopped processing new data on July 1, 2023, as announced, and the industry completed a forced migration to GA4. The organizations that had prepared — running GA4 in parallel, re-establishing baselines, re-defining conversions as key events, and educating stakeholders about the new model — carried their reporting through with a documented discontinuity and little lasting damage. The organizations that treated analytics as a set-and-forget utility discovered, at the worst possible moment, that their entire reporting narrative had been resting on a platform decision that was never theirs to make.

The lesson

Reporting is downstream of measurement, and measurement is infrastructure you must understand and protect. The specific tools will keep changing — GA4 will not be the last model — but the disciplines that survived this migration are the disciplines of this chapter: freeze a real baseline early and guard it; know what each metric actually counts, so you can explain why two tools disagree; and when the ground moves, report the discontinuity honestly rather than papering over it. A reporter who does those three things can survive any tool change. A reporter who trusts a single number from a platform they don't control is one product announcement away from having nothing to say.

Discussion questions

  1. A company comes to you in early 2023, three months before the UA cutoff, having never touched GA4. What is the first thing you do about their reporting baseline, and what have they already permanently lost?
  2. GA4 reports 20% fewer "users" than UA did for the same site and month. A panicked owner asks which number is "the real one." How do you answer without saying either tool is broken?
  3. GA4 renamed "conversions" to "key events" and made attribution data-driven by default. Why do these vocabulary and model changes matter for the KPIs on an SEO report, not just for the analytics team?
  4. Argue both sides: is a forced platform migration that breaks year-over-year comparisons a reason to distrust analytics, or a reason to hold measurement to higher professional standards? Where do you land, and why?
  5. Connect this case to §29.3's claim that Search Console and GA4 "will never agree." What is the general principle about multi-tool reporting that both the migration and the GSC-vs-GA4 gap illustrate?