Universal Analytics did attribution reasonably well. It wasn't perfect, but it gave you multi-channel funnels, model comparison tools, and conversion paths that made sense. Most advertisers understood it.
Then GA4 replaced it, and the industry collectively lost its measurement baseline.
The complaints about GA4 are well documented at this point. The interface is unintuitive. Reports that took two clicks in Universal now require custom explorations. Real-time data doesn't match what you see the next day. The learning curve drove most advertisers to either misconfigure it or stop using it as a source of truth altogether.
But even when GA4 is set up properly, there are real limitations that affect how useful it is for making ad spend decisions.
Most GA4 setups are broken
Before we get to the structural limitations, the practical reality is that most GA4 implementations we audit have significant issues:
Cross-domain tracking not configured. If a customer clicks an ad, lands on your marketing site, then clicks through to your shop subdomain to purchase, GA4 treats that as two separate sessions. The ad click gets attributed to nothing. The purchase gets attributed to a "referral" from your own website. This is one of the most common attribution errors we see, and it makes channel-level reporting essentially useless.
Events not mapped correctly. GA4's ecommerce tracking requires specific event names and parameters. Many implementations have the event names right but the parameters wrong — revenue coming through as zero, currency codes missing, item-level data absent. The conversion shows up in reports, but the revenue attributed to it is wrong or missing entirely.
Consent Mode integration is broken or missing. Consent Mode v2 requires both the consent banner and the Google tags to be configured correctly. If the banner fires consent signals before the tags load, or if the default consent state isn't set properly, you either lose data you should be collecting or collect data you shouldn't be. Both are common.
Enhanced Conversions not enabled. Enhanced Conversions sends hashed first-party data (email, phone) back to Google to improve match rates. It partially compensates for signal loss from cookie restrictions. It's been available for over two years now and we still see the majority of accounts without it enabled.
A properly configured GA4 — with server-side tagging, Enhanced Conversions, and Consent Mode v2 set up correctly — is already collecting first-party data and doing a reasonable job. The problem is that most accounts aren't there, and advertisers don't realise because they have nothing reliable to compare it against.
Where GA4 falls short, even when it's set up properly
Even with a professional GA4 implementation, there are genuine gaps that affect ad spend decisions.
The cross-device problem
GA4 tracks users primarily through browser cookies. When someone clicks an ad on their phone during lunch, researches on their work laptop that afternoon, and purchases on their tablet that evening, GA4 struggles to connect those touchpoints into a single journey.
Google's User-ID feature and Google Signals help bridge some of these gaps, but they rely on users being signed into Google across devices — which isn't universal. The result is that cross-device journeys are either misattributed or lost entirely. For ecommerce businesses where multi-device purchase journeys are common, this is a meaningful gap.
The attribution model mismatch with Google Ads
This is the most practical problem, and the one most advertisers don't realise exists.
GA4 uses its own data-driven attribution model. Google Ads uses a different one. They are not the same model, and they consistently produce different numbers — with GA4 almost always reporting fewer conversions and lower revenue than Google Ads.
This matters when GA4 is your primary conversion source for Smart Bidding. If GA4 attributes 80 conversions to a campaign but Google Ads would attribute 100, Smart Bidding is optimising on the lower number. It's underbidding on keywords that are actually profitable because the conversion data feeding it is understated relative to what Google's own system would report.
The irony is that by using GA4 as your conversion source to try and get "more accurate" data, you may actually be giving Smart Bidding a worse signal than Google's native tracking would provide.
Consent and signal loss
Even with Consent Mode v2 configured correctly, users who don't consent to cookies create a data gap. GA4 uses behavioural modelling to estimate conversions from non-consenting users, but that's modelled data — an estimate, not an observation.
With consent rates varying between 40-80% depending on market and implementation, the proportion of modelled vs observed data can be significant. The further you go down the chain of estimates, the less the numbers reflect what actually happened.
Cross-platform attribution: what GA4 does and doesn't solve
GA4 does handle cross-platform deduplication — if you have proper UTM tagging and custom dimensions set up, it can show you that a user arrived via a Meta ad, came back through organic search, and converted on a Google brand click, without double-counting.
There's also a widely held view in the industry that GA4's data-driven model tends to favour Google's own platforms when distributing credit. Google hasn't published the inner workings of the model, so this is difficult to verify empirically — but it's worth being aware of when you're evaluating channel performance across Google and non-Google sources.
The gap that GA4 genuinely can't close is, again, cross-device. If that Meta ad click was on mobile and the Google brand conversion was on desktop, GA4 likely sees those as two separate users. Proper UTM tagging doesn't help when the cookie identity changes with the device.
What first-party attribution adds
A properly configured GA4 gets you partway there. But there's a category of tool that goes further — dedicated first-party attribution platforms. These sit on your domain, run their own server-side tracking, and solve the specific gaps that GA4 can't.
Tools like Northbeam, Triple Whale, and well-configured server-side GTM setups with offline conversion imports can all achieve this. We've also built our own platform for our ecommerce clients because we wanted full control over the attribution model and data pipeline.
Here's what a dedicated attribution platform does differently:
Cross-device identity. GA4 relies on browser cookies and Google Signals to connect users across devices. A dedicated platform stores visitor identity server-side — when a visitor provides their email (newsletter signup, account creation, checkout), that identity persists independently of any browser. The mobile ad click and the desktop purchase get connected reliably, without depending on the user being signed into Google.
Verified purchase data. GA4 records conversions as browser events — a JavaScript tag fires when it detects a purchase. A dedicated platform verifies orders against your ecommerce platform's API directly. You're matching against the actual order in your backend — real revenue, real products, confirmed by your order management system. No reliance on whether a tag fired correctly or whether the user consented to cookies.
Feeding verified data back to Smart Bidding. This is where the real advantage sits. Google Ads has an offline conversion import feature — you send Google the click ID, the verified order value, and the timestamp. Smart Bidding then optimises on real, backend-confirmed sales rather than browser-fired events or modelled estimates.
If your browser-based tracking is under-reporting by 15-20% (which is common in a post-consent world), Smart Bidding is making decisions on incomplete data. Importing verified conversions closes that gap. The algorithm gets the full picture and bids accordingly.
The same principle applies to funnel events. Track the full funnel — product views, add to cart, begin checkout, purchase — and you're giving the algorithm four signals per customer journey instead of one. It learns faster and optimises more accurately.
The audience advantage
A dedicated attribution platform also strengthens remarketing audiences.
When you have verified purchase data with email addresses, you can build customer match audiences directly — all purchasers, high-value customers, repeat buyers. These aren't based on cookie pools that shrink every time a browser updates its privacy settings. They're based on real customer data matched against Google's user base.
You can also build audiences that GA4 struggles with at scale. Cart abandoners who added a product but didn't purchase within 30 days. Visitors who viewed specific product categories but never bought. These require server-side event tracking combined with a backend that can query across time windows and match against purchase data.
The quality of the seed list directly affects the quality of the similar audiences Google builds from it.
What this doesn't replace
None of this means you should rip out GA4. GA4 still has an important role — site analytics, user behaviour, content performance, and debugging are all things it handles well. With proper UTM tagging, it also gives you a reasonable cross-platform view of your marketing mix.
What a dedicated attribution platform replaces is GA4's role as the source of truth for conversion attribution and ROAS reporting — specifically when that data is feeding Smart Bidding. Those numbers should come from a system that verifies against your actual orders and feeds clean data back to the ad platforms for bidding.
Where to start
If you're managing ecommerce PPC and relying on GA4 for attribution:
Audit your GA4 setup. Check cross-domain tracking, ecommerce event mapping, Consent Mode integration, Enhanced Conversions, and server-side tagging. A properly configured GA4 is genuinely useful — and many of the issues advertisers blame on GA4 are actually configuration problems.
Compare GA4 to reality. Pull your GA4 purchase count for last month. Compare it to your actual order count from your ecommerce platform. Then compare GA4's Google Ads attributed revenue to what Google Ads itself reports. If either gap is more than 15%, your measurement foundation has a problem.
Evaluate what you actually need. If you're running Google only, your GA4 is properly configured, and the numbers are close to reality, GA4 may be sufficient for your needs. If you're running multi-platform campaigns, seeing significant cross-device purchase journeys, or spending enough that a 15-20% measurement gap translates into real money, first-party attribution is worth the investment.
The advertisers who will do best in 2026 and beyond are the ones who understand what their tools can and can't do — and fill the gaps where it matters most.