When Google announced it was sunsetting Universal Analytics in favor of Google Analytics 4, the pitch was compelling: a privacy-first, event-driven platform built for a cookieless world. Cross-device tracking. Better machine learning. Future-proof infrastructure. Marketers were told to migrate, adapt, and trust the process.
Then they actually used it.
The backlash was immediate and, frankly, earned. Search Engine Journal captured the professional consensus early: GA4 was “not ready for prime time,” “unintuitive,” and — in the words of one practitioner quoted by Wide Angle Analytics — “downgrading from a Tesla to a tricycle.” Those aren’t the words of people resistant to change. That’s the sound of professionals discovering that a tool billed as an upgrade had stripped away capabilities they depended on.
Years later, with Universal Analytics gone and GA4 entrenched, it’s worth asking an honest question: did GA4 actually fix the problems it promised to fix? And more importantly, what problems did it introduce?
The Migration Was Painful, and the Wounds Haven’t Fully Healed
A mid-2023 Search Engine Land poll of marketers, reported by Nicola Agius, surfaced a consistent set of complaints: data discrepancies between GA4 and other platforms, missing features that UA handled well, and no lightweight option for small businesses that don’t need enterprise-grade complexity. Google never built a “GA4 Lite” — smaller businesses got the same complex platform as enterprise users, with all the overhead and none of the handholding.
The frustration wasn’t cosmetic. Reporting structures changed fundamentally. Metrics were renamed, redefined, or eliminated entirely. Historical UA data didn’t migrate, so businesses that relied on multi-year trend analysis suddenly had a hard cutoff in their data history. That’s not a transition. That’s a reset.
Your Data Is Already Old When You See It
In an environment where programmatic ad platforms make bidding decisions in milliseconds and campaign managers need to kill underperforming ads before the day is over, data latency is a competitive disadvantage. Celebrus documented in 2024 that GA4 data freshness runs 12 to 48 hours behind real-world activity. Universal Analytics delivered data within roughly 4 hours.
Think about what that gap means in practice. You’re checking GA4 on Tuesday morning to decide whether to scale a campaign or pull budget. The data you’re looking at reflects Monday, at best. If something broke Sunday night — a landing page error, a UTM misconfiguration, a conversion tag firing on the wrong event — you’ve been spending money against bad intelligence for over a day before you even know there’s a problem.
Real-time ad platforms deserve real-time measurement. The two are operating on completely different clocks.
Fourteen Months Is Not a Business Horizon
GA4’s default retention limit for individual event-level data is 14 months. Go back further than that and the data simply isn’t there, unless you’ve connected GA4 to BigQuery and are paying to warehouse it yourself.
Celebrus flagged this in their 2024 analysis. Root and Branch Group, in a January 2025 update, put it plainly: GA4 seems “built more as a tool for collecting data,” with the actual analysis expected to happen somewhere else.
For plenty of businesses, that’s an expensive inconvenience. For seasonal businesses — retail, hospitality, tax services, anyone with a meaningful holiday or cyclical spike — 14 months isn’t even two full cycles. You can’t do a real year-over-year comparison when your data disappears partway through the comparison window. And if you don’t have a data engineer comfortable with BigQuery, the solution on paper becomes a roadblock in practice.
Sampling Distorts the Reports You Actually Care About
Data sampling is what happens when a reporting tool has too much data to process and starts making statistical inferences instead of counting actual events. GA4 triggers sampling at 10 million events per query for standard reports.
That sounds like a lot until you realize what it means at scale. Matomo’s October 2024 analysis noted that complex reports — filtered segments, multi-step funnels, cross-channel comparisons — hit that threshold fast. Every session generates dozens of events. Heavy traffic sites aren’t the only ones affected. A moderately trafficked e-commerce site running detailed funnel analysis can hit sampling regularly.
When sampling kicks in, GA4 is no longer reporting what happened. It’s estimating what happened, based on a subset of your data. Matomo’s analysis shows that complex reports can drop to below 50% sampling rates, with error margins as high as 30%. Platforms that avoid sampling entirely give you the full picture. That’s the difference between a campaign that looks neutral and one that’s clearly failing.
When GA4 Can’t Track a Conversion, It Guesses
This one deserves more attention than it typically gets. GA4 uses a technique called modeled conversions to fill in gaps created by ad blockers, browser privacy restrictions, and users who decline consent. When it can’t observe a conversion directly, it estimates whether one probably occurred, based on patterns in the data it can see.
WebFX’s Nolan Barger put it directly in a March 2026 analysis: “relying on data modeling to cover over a third of your dataset is a bold move.” That’s a third of your conversion data potentially being estimated rather than observed. And GA4 doesn’t always make it obvious which conversions in a given report are real and which are modeled — the two sit side by side in your reports as if they’re equivalent.
Modeling isn’t inherently dishonest. It’s a reasonable response to a fragmented tracking environment. But treating estimated conversions the same as observed ones, without surfacing the distinction clearly, creates a false sense of precision in your reporting.
Google Removed the Attribution Models You Actually Understood
Attribution was already complicated. Google made it more complicated in 2023 by removing First Click, Linear, Time Decay, and Position-Based attribution models from GA4. Search Engine Land covered the removal. What remained: Data-Driven attribution and Last Click.
Last Click is simple to understand but notoriously bad at capturing how customers actually make decisions. Data-Driven is a machine learning model that distributes credit based on patterns Google’s system identifies — but it’s a black box. You cannot see the logic. You can’t audit it. You’re trusting that Google’s algorithm is distributing credit fairly across your channels.
Silverback Labs documented a 40-plus percent discrepancy between the two remaining models in 2025. Two attribution models, one comparison, and a gap of more than 40 percent in how they allocate credit for conversions. “Black box anything is a red flag” is how they put it. Hard to argue.
The Reports That Should Work Don’t Always Work
Zima Media’s 2024 reporting roundup catalogued specific GA4 failure modes that practitioners kept running into: a 90-day cap on conversion windows (limiting for products with longer consideration cycles), Lifetime Value reports that also cap at 90 days despite the label implying something more comprehensive, and broken revenue attribution for installment payment structures where revenue comes in over time rather than at a single point.
Combine that with Celebrus’s observation that the historical data wall creates a steep learning curve just to get basic comparisons working, and you have a reporting layer that fights you when you need it most.
Compliance Is Not a Solved Problem
For any business operating in regulated industries — healthcare, finance, legal services, insurance — GA4’s compliance posture matters. Celebrus’s 2024 analysis noted that GA4 lacks out-of-the-box GDPR and HIPAA compliance. Getting to a compliant configuration requires technical work, third-party consent management platforms, and ongoing maintenance as regulations evolve.
Universal Analytics, with proper configuration, gave regulated businesses a cleaner path to compliance. GA4 made it harder, not easier, at a moment when privacy regulation is expanding, not contracting.
What GA4 Actually Is
GA4 is a free, event-based analytics platform designed to feed data into Google’s advertising ecosystem. That’s not a cynical read — that’s what it was built to do. It tracks user behavior across web and app surfaces, feeds signals into Google Ads for bidding optimization, and provides enough reporting for businesses whose analytical needs are relatively modest.
That’s genuinely useful. For what it is.
The problem is what it isn’t. It isn’t a complete measurement solution for businesses that need real-time data, long historical records, unsampled reports, transparent attribution, and compliance-ready infrastructure out of the box. It wasn’t designed to carry that weight, and the signs are everywhere in the product.
Most businesses didn’t choose GA4 because it was the best tool for their needs. They chose it because it was free and because Universal Analytics was taken away. Those aren’t the same thing as a rational product decision.
The harder question isn’t whether GA4 is broken — it’s why so many businesses are still treating it as the definitive source of truth for their marketing measurement when the evidence suggests it was never designed for that role.