Google Analytics 4 (GA4) vs. Universal Analytics: What Actually Changed?

The honest answer to the GA4 vs Universal Analytics question is that Google did not upgrade a reporting tool, it replaced the measurement model underneath it. Universal Analytics counted sessions and pageviews; Google Analytics 4 counts events, and almost every reporting quirk people complain about traces back to that single decision. Standard Universal Analytics properties stopped collecting data on July 1, 2023, and the old interface and API went dark a year later, so the comparison now matters mostly for people trying to make sense of the numbers they are left with.

If your reports feel wrong, they probably are not wrong. They are measuring something different.

The Short Version of the Difference Between GA4 and UA

Universal Analytics organised data into sessions, and every hit (pageview, event, transaction) lived inside a session. GA4 flattens that: every interaction is an event with parameters attached, and sessions are reconstructed afterwards from a session_start event. Google documents the full GA4 event structure for anyone who wants the technical version.

That change cascades into five practical areas:

• Data model: session-based hits became event-based records with up to 25 parameters each.
• Metrics: bounce rate, pageviews and goals were renamed, redefined or removed.
• Reporting: dozens of prebuilt reports were replaced by a smaller default set plus Explorations.
• Attribution: last non-direct click gave way to data-driven attribution.
• Infrastructure: free BigQuery export and consent-based modelling arrived for every property, not just 360 clients.

Sessions Out, Events In

In Universal Analytics, a session ended at midnight, after 30 minutes of inactivity, or whenever the campaign source changed. That last rule inflated session counts: a user who arrived from an email, left, and came back from a paid ad generated two sessions in one sitting.

Google Analytics 4 dropped the campaign-change reset, so session totals usually come in lower than the Universal Analytics equivalent for the same traffic, commonly by 5% to 15% depending on how much of your traffic bounces between channels. GA4 also applies a timezone-based day boundary rather than splitting a session at midnight.

The practical upside is that one user journey now reads as one journey. The downside is that any spreadsheet comparing 2022 session volume against 2026 session volume is comparing two different definitions.

Metrics That Changed Names, Meanings or Disappeared

This is where most teams lose trust in the data, usually because a familiar word now means something else:

• Bounce rate was removed at launch, then reintroduced in mid-2022 as the simple inverse of engagement rate. It no longer means “single-page session”.
• Engaged sessions are sessions lasting at least 10 seconds, or with two or more screen views, or with a key event. That 10-second threshold is adjustable between 10 and 60 seconds.
• Goals became conversions, and in 2024 conversions were renamed key events, with “conversions” reserved for the Google Ads side.
• Pageviews became views, combining web pageviews and app screen views in one metric.
• Users defaults to active users in GA4 reports, not total users, which is why the headline number often looks smaller than the old property.

Several Universal Analytics staples never made the trip at all: view-level filters as you knew them, the full set of behaviour flow reports, custom dashboards in their old form, and the 26-month-plus data retention ceiling. Standard GA4 properties cap event-level retention at 14 months, which is the single most overlooked change in the whole migration.

Reporting Got Thinner on the Surface and Deeper Underneath

Universal Analytics shipped with a long left-hand menu of prebuilt reports. GA4 ships with a short one, built around the lifecycle collection (Acquisition, Engagement, Monetisation, Retention), and pushes anything custom into the Explore section.

Explorations are genuinely more capable than the old custom reports: funnel exploration, path exploration, segment overlap and cohort analysis are all included at no cost. They also require a user who knows what dimensions and metrics are compatible, because invalid combinations return blank cells rather than an error message.

Two reporting behaviours catch people out. Data thresholding hides rows when Google Signals is on and the audience is too small to anonymise safely. Cardinality limits push high-variety dimensions such as page paths with query strings into an “(other)” bucket once a report exceeds roughly 500 rows per day in standard properties.

Attribution and Google Ads Reporting Work Differently

Universal Analytics defaulted to last non-direct click and offered first click, linear, time decay and position-based models in the Model Comparison Tool. Google retired those rule-based models in 2023, leaving data-driven attribution as the GA4 default with last click and a couple of Google-specific options alongside it.

Data-driven attribution distributes credit using your own conversion patterns, which tends to move budget credit away from brand search and toward upper-funnel touchpoints. It is a better reflection of reality for most accounts, but it does mean your paid channel will rarely report the same numbers it did under the old model.

Linking is also stricter now. If your account was migrated in a hurry, it is worth re-checking the connection, because auto-tagging and the key event import settings control what Google Ads can actually bid on. Our walkthrough on how to link Google Ads to Google Analytics 4 covers the settings that commonly get skipped, and if your spend looks busy but flat, the diagnostic steps in clicks without conversions usually point at a measurement gap rather than a bidding one.

Free BigQuery Export Is the Quiet Upgrade Nobody Talks About

Under Universal Analytics, raw hit-level data export to BigQuery was a 360 feature with a six-figure price tag attached. Every GA4 property can stream its event data to BigQuery for free, subject to a daily export limit of one million events on standard properties.

That matters more than any interface change for three reasons: it preserves event detail beyond the 14-month retention window, it sidesteps both thresholding and cardinality limits, and it lets you join analytics data to CRM or offline revenue data. Teams that track phone calls and offline conversions get the most obvious benefit, since the final sale almost never happens on the website.

Turning the export on costs nothing and takes about ten minutes. It is the first thing worth fixing in a neglected property, because retroactive data cannot be recovered once the retention window closes.

Privacy, Consent Mode and Modelled Data

Universal Analytics was built in a cookie-friendly era. GA4 was designed around the assumption that a meaningful share of visitors will never be identifiable, which is why it blends device IDs, User-ID and modelled behaviour rather than relying on client-side cookies alone.

Since March 2024, advertisers serving the European Economic Area have needed Consent Mode v2 signals for Google Ads personalisation and modelling to work. Where consent is declined, GA4 fills the gap with behavioural modelling once the property hits Google’s volume thresholds, which is why some reports carry a small modelling notice.

IP addresses are no longer stored or logged at all, so the old “anonymise IP” setting has no GA4 equivalent. Google’s Universal Analytics deprecation documentation remains the reference point for what did and did not carry over.

What to Fix in Your GA4 Property in 2026

Most properties were migrated in a rush during 2023 using the auto-created setup assistant, and that assistant made guesses. Three years of those guesses compound. A short audit usually finds the same handful of problems:

1. Retention still set to two months. Change it to 14 months in Data Settings, then confirm the BigQuery export is running.
2. Key events that are not key. Pageview-based “conversions” inflate counts and train Google Ads bidding on noise. Mark form submissions, qualified calls and purchases only.
3. Unfiltered internal and developer traffic. Define internal traffic rules and activate the data filter rather than leaving it in testing mode.
4. Missing cross-domain configuration. Booking systems and checkout subdomains split one user into two without it.
5. No annotation of the UA cutover. Anyone reading a long-range chart needs to know where the definitions changed.

Because Universal Analytics data was deleted and cannot be exported anymore, year-over-year comparisons that cross mid-2023 are gone for good. The pragmatic fix is to rebase your targets on GA4-native numbers and judge performance against the last four full quarters. If you are measuring something like email signups, define the key event properly before you build the campaign, which pairs well with the approach in our guide to lead magnets that grow an email list.

For behaviour questions that GA4 answers poorly (why a page fails, where attention dies), pair it with a session recording tool. The practical setup is covered in our piece on using heatmaps for user insight, which fills the gap left by the retired behaviour flow reports.

Frequently Asked Questions

When Did GA4 Switch Over?

Standard Universal Analytics properties stopped processing new data on July 1, 2023, and Analytics 360 properties followed on July 1, 2024. The old interface and API were shut off and historical data deleted in July 2024, so GA4 has been the only option for new collection since then.

What Is GA4 Analytics?

Google Analytics 4 is the event-based measurement platform that replaced Universal Analytics, first released as App + Web in 2019 and made generally available in October 2020. It records every interaction as an event with parameters, covers web and app traffic in one property, and includes free BigQuery export plus machine-learning-based modelling for gaps in consented data.

What Is the Difference Between GA4 and GSC?

GA4 measures what people do after they land on your site, while Google Search Console reports what happens in Google Search before the click, including impressions, average position and query data. Search Console is the only source of keyword-level data, so most teams link the two and read them side by side.

What Is the Difference Between GA4 and GTM?

Google Tag Manager is a tag deployment tool and GA4 is the analytics platform that receives the data. GTM holds no reports of its own; it fires the GA4 tag, pushes custom event tracking, and manages other scripts such as ad pixels from one container without developer edits.

Get a Second Opinion on Your GA4 Setup

If your property was migrated by the setup assistant and nobody has touched the configuration since, the numbers you are reporting on are probably measuring the wrong things politely. SEO Quirk audits GA4 properties, cleans up key events and reconnects ads data so the reporting matches how your business actually earns money, and you can read more about working with a local agency before you get in touch.