A large share of the Google Analytics advice still circulating tells you to open Behavior, then Site Content, then All Pages, and to judge your worst pages by bounce rate. That menu path has not existed since 1 July 2023. The advice is not merely dated, it points at an interface Google switched off, which makes it a useful test: if a guide names those reports, it was written for a product you no longer have.
Key takeaways
- Standard Universal Analytics stopped processing hits on 1 July 2023, the 360 tier on 1 July 2024.
- Bounce rate in GA4 is simply the inverse of engagement rate, not the old single-page metric.
- Site speed lives in Core Web Vitals now: LCP 2.5 s, INP 200 ms, CLS 0.1 at the 75th percentile.
- Standard properties keep user-level data for 2 or 14 months, and that limit hits explorations only.
Start by checking which product your reference material describes
Universal Analytics is gone, in two stages. Standard properties stopped processing new hits on 1 July 2023, and Universal Analytics 360 properties followed on 1 July 2024, after which the old interface and API were withdrawn for everyone. Anything written before those dates that talks about goals, sessions as the primary unit, or the Behavior and Site Speed report groups is describing a discontinued product.
This matters more than a naming change because the underlying model changed. Universal Analytics organised data around sessions and pageviews. Google Analytics 4 organises everything as events, including the pageview, which means a metric that survived the migration by name has usually not survived it by definition. Reading GA4 with Universal Analytics habits is the single most common reason a team concludes that “the numbers dropped” after migrating.
Three definitions that change what your numbers mean
Before comparing anything to a benchmark or to last year, pin down these three. Each one is a Google definition, not an interpretation.
An engaged session is a session that meets any one of three conditions: it lasts longer than 10 seconds, it includes a key event, or it includes at least two page or screen views. System events such as first_visit, first_open and session_start are excluded from that calculation even if you have marked them as key events.
Engagement rate is the share of sessions that were engaged. Bounce rate is the share that were not, which makes it the exact arithmetic inverse and nothing like the Universal Analytics metric of the same name. A visitor who reads one page for eight seconds and leaves produces a 100 percent bounce; the same visitor staying eleven seconds produces zero. Any target inherited from the old metric is meaningless against the new one.
Key events are what conversions were called until Google renamed them in 2024, to separate what Analytics measures from what Google Ads optimises against. Nothing changed in how they are created or reported. What changed is that “conversion” now belongs to the advertising side of the stack, which removes a genuine source of confusion when two teams compare numbers from two products.
Google Analytics is no longer where you diagnose a slow page
The Site Speed reports went with Universal Analytics and were not rebuilt. Page performance is now assessed through Core Web Vitals, and those live in Search Console, PageSpeed Insights and the Chrome User Experience Report rather than in Analytics.
The current set has three metrics, each assessed at the 75th percentile of page loads and split between mobile and desktop. Largest Contentful Paint should occur within 2.5 seconds. Interaction to Next Paint should come in at 200 milliseconds or less. Cumulative Layout Shift should stay at 0.1 or below. Note the second one: Interaction to Next Paint replaced First Input Delay as a Core Web Vital on 12 March 2024, and it is a stricter measure, because it looks at every interaction on the page rather than only the delay before the first one is handled.
The practical division of labour is clean once you accept it. Analytics tells you which pages carry traffic and which ones people leave without engaging. Core Web Vitals tooling tells you whether the page itself is the reason. Trying to answer the second question inside Analytics is how teams spend a quarter optimising the wrong thing.
Two settings that quietly decide whether the data is usable
Both are configured once, both are easy to miss, and both are irreversible in the sense that data not collected cannot be recovered later.
Data retention. A standard property can retain user-level and event-level data for either 2 months or 14 months, with longer options up to 50 months reserved for 360 properties. The setting does not affect standard aggregated reports, so a year-on-year comparison in a canned report still works. It does affect explorations and funnel reports, which is exactly where the interesting analysis happens. Age, gender and interest data are held for 2 months regardless. Leaving a new property on the shorter option and discovering it fourteen months later is a common and entirely avoidable loss.
Consent mode. Since 6 March 2024, advertisers and publishers serving the European Economic Area have needed consent mode v2 to keep passing usable signals to Google’s advertising and measurement products, a change driven by the Digital Markets Act obligations that apply to Google as a designated gatekeeper. Without it, data on new European users stops flowing into the audience and conversion features you are presumably paying for. This is a consent-management question before it is an analytics question, and it belongs with whoever owns your cookie banner.
The pass we run on a property before trusting any report
- Confirm the property type and creation date. A property inherited from a migration behaves differently from one built natively, particularly around event naming.
- Set retention deliberately. Pick 14 months unless there is a documented privacy reason not to, and write down who made the call.
- Audit the key events list. Most properties carry at least one event marked as key that measures nothing anyone acts on, which inflates engagement and distorts every downstream comparison.
- Check consent mode status against the markets you actually serve, not the market you are based in.
- Pull Core Web Vitals separately for the ten pages carrying the most traffic, before drawing any conclusion about content quality from engagement figures.
None of that is analysis. It is the work that makes analysis worth doing, and it takes an afternoon. We would rather spend that afternoon than build a quarter of recommendations on a property where half the key events fire on page load.
https://www.youtube.com/watch?v=_rm4SNmhDXA
Questions we get after the first audit
Can I still see my old Universal Analytics data?
No. Access to the Universal Analytics interface and API ended for all users, including 360, after the July 2024 cutoff. If nobody exported the historical data before then, it is not retrievable, and any year-on-year comparison has to start from the GA4 collection date.
Why does Google Analytics disagree with my ad platform on conversions?
Because they are counting different things by design. Analytics attributes a key event to a session it observed; an ad platform attributes a conversion to an impression or click it served, within its own lookback window. Both can be internally correct and still differ by a wide margin. Reconcile the definitions before reconciling the numbers.
Is a high bounce rate in GA4 still a bad sign?
Not on its own. Since it is the inverse of engagement rate, a page that answers a question in under ten seconds and sends the visitor away satisfied will look identical to a page nobody wanted. Read it alongside key events and scroll behaviour, never alone.
Does GA4 show me which keywords brought traffic?
Not directly. Organic query data comes from Search Console, which can be linked to a GA4 property; Analytics itself sees the referral, not the query. This is a limitation of how search engines pass data, not a configuration you have missed.
One last piece of context worth holding onto. A measurement platform only sees what passes through it, and a growing share of referral and sharing activity never carries a traceable source, which is a structural gap in attribution rather than a reporting bug. Treat Analytics as the best available account of what happened on your site, not as a complete one.
Numbers clean, conversions still flat?
At that point the measurement is rarely the problem, and the sequence of steps you are asking people to take usually is.
Sources: Google, for standard Universal Analytics properties ceasing to process new hits on 1 July 2023 and Universal Analytics 360 properties on 1 July 2024, with the interface and API withdrawn thereafter. Google Analytics Help, for the definition of an engaged session (longer than 10 seconds, a key event, or at least two page or screen views), for engagement rate and bounce rate as inverse metrics, and for the exclusion of first_visit, first_open and session_start from engaged session calculations. Google Analytics Help, for data retention options of 2 or 14 months on standard properties, up to 50 months on 360 properties, the 2-month limit on age, gender and interest data, and the fact that retention affects explorations and funnel reports rather than standard aggregated reports. Google, March 2024, for the renaming of conversions to key events in Analytics. web.dev, Google Chrome team, for the Core Web Vitals thresholds of 2.5 seconds for Largest Contentful Paint, 200 milliseconds for Interaction to Next Paint and 0.1 for Cumulative Layout Shift, each assessed at the 75th percentile of page loads. Google Search Central, for Interaction to Next Paint replacing First Input Delay as a Core Web Vital on 12 March 2024. Google, for consent mode v2 becoming a requirement for European Economic Area traffic from 6 March 2024 under Digital Markets Act obligations. Product interfaces and thresholds in this area change frequently; verify against current Google documentation before acting. Updated August 2026.
