Resources
GA4 Tracking That Survives 2026
GA4 is a good free tool being misused as an exact measurement. Its limitations are structural rather than bugs — knowing them prevents most bad marketing decisions before they are made.
What are GA4's real limitations?
GA4 replaced Universal Analytics with a model that is genuinely different in structure, not just in interface. The limitations below are consequences of that design. Knowing them prevents you drawing conclusions the data cannot support.
| Limitation | What it means in practice | How to work with it |
|---|---|---|
| No pageviews in the admin | Page views appear in reports but not in realtime or explorations as before | Use explorations and the events report; it is a change of habit, not a defect |
| Event data retention | Standard is 14 months; beyond that events are lost | Export to BigQuery if you need long-term event-level history |
| Modelled conversions | Where signal is lost, conversions are estimated rather than observed | Compare modelled and observed separately; never blend them silently |
| Ad blockers and ITP | A meaningful share of conversions never fire a tag | Server-side tagging, or accept the loss and account for it |
| Consent mode thins data | Denied-consent sessions produce cookieless pings with different behavior | Never compare raw figures across the consent change date |
| Attribution is modelled | Multi-touch attribution is a model, not an observation | Treat platform attribution as directional, never as ground truth |
| Cross-domain complexity | Tracking across related domains requires deliberate setup | Set up cross-domain measurement or self-referrals corrupt everything |
| Sampling can apply | Some reports sample larger datasets | Check for sampling notices before drawing conclusions from a report |
What is the measurement layer?
Analytics has five layers, and problems are usually diagnosed at one and blamed on another. Knowing which layer has failed saves enormous time.
Collection
How data travels — client-side to Google, or through your own server. Vulnerable to blockers, latency and sampling.Processing
Deduplication, consent application, identity resolution, modeling. This is where most 'the numbers look wrong' problems actually live.Reporting
Dashboards, explorations and alerts. A correct model can produce useless output if nobody reads it.Decision
Whether anyone changes what they do based on it. The layer most often missing, and the one that determines whether analytics is worth its cost.What should I actually track?
Track the minimum set that supports decisions you actually make, plus the technically necessary events. Teams routinely instrument far more than they use, which adds collection cost and complexity for nothing.
| Business model | Essential events | The metric that matters |
|---|---|---|
| Content / media | Scroll depth, engaged session, newsletter signup, outbound click | Subscriber growth and assisted conversions |
| Lead generation (B2B) | Form submit, demo booking, content download, scroll to CTA | Qualified pipeline, not raw form fills |
| SaaS / product | Signup, activation event, trial start, invite sent, upgrade | Trial-to-paid and activation rate |
| Ecommerce | View item, add to cart, checkout start, purchase, refund | Contribution margin per session and repeat rate |
| Local service | Call click, direction request, GBP action, form submit | Booked jobs, tracked to source |
| Professional services | Contact, consultation booking, document download | Consultations booked and their value |
How do I track calls and offline conversions?
For most service businesses the phone call is the primary conversion, and it is invisible to GA4 unless you explicitly connect it. This is the most common blind spot in local and B2B marketing, and it means reported conversion rates are frequently wrong by a large factor.
Three approaches, in increasing order of effort.
- Dynamic number insertion. Display a tracking number in place of your real one in ads and on the site; calls to it are attributed to source. Handle staff numbers and number recycling, and treat attribution as directional.
- Offline Conversion import. Upload closed deals, booked jobs or completed quotes back into GA4 with the original source. This is the highest-value measurement work available to most B2B businesses and it is routinely not done.
- Ask at the point of conversion. "How did you hear about us?" captures self-reported attribution. Imperfect and inconsistent, but free, and it gives a rough sense that the rest of the data can then be calibrated against.
Without at least one of these, any return-on-investment figure you report is a model output derived from an incomplete dataset, and should be labeled as such.
How does consent mode v2 change measurement?
Consent Mode v2 requires specific parameters to be declared before tags fire, and it changes what you can measure rather than simply whether you measure it. With consent denied you receive modelled data — cookieless pings that Google estimates rather than observes.
This creates a specific, manageable problem: consented and non-consented traffic are different populations, and blending them distorts every rate you calculate. Conversion rates, engagement rates and averages all shift if modelled sessions are mixed with observed ones.
This site itself sets no cookies and runs no analytics, which is why the only data held is what you type into the contact form.
How do I build reports people actually use?
Most dashboards are built and then ignored, usually because they report what was easy to measure rather than what drives a decision. The fix is to start from the decision.
- Start from the decision, not the data. Write the decision first: "do we increase budget or not?" Then work back to the number that answers it.
- Report revenue where you can. Any report ending in sessions rather than money is a vanity report in waiting.
- Keep it to one page. A report nobody reads has negative value, because it signals that measurement is not taken seriously.
- Show the change and the cause. What moved, and the most likely reason. Not what you did.
- Annotate discontinuities. Consent changes, site releases, tracking changes, campaign starts. An unexplained spike is uninterpretable six weeks later.
- Review monthly, not daily. Daily reporting on weekly-moving metrics produces noise and retrains people to ignore real changes.
A useful test: if a report cannot change a decision, it should not exist. That single filter removes most reporting overhead without losing anything actionable.
Frequently asked questions
What are the main problems with GA4?
No pageviews in the admin, 14-month event data retention, modelled conversions where signal is lost, ad blockers preventing tags from firing, consent mode thinning data, and attribution that is a model rather than an observation. None of these make GA4 unusable — it is a good free tool — but all of them mean it should not be treated as exact measurement.
Why did my GA4 numbers change overnight?
The usual causes are a consent banner going live, a tag deployment, a site release, a tracking change or ad blockers affecting a different share of traffic. Check whether it was collection or real traffic by comparing GA4 against Search Console. Also check whether a tag stopped firing, which produces exactly this symptom and is often the simplest explanation.
Should I use server-side tagging?
It depends on your situation. It helps when ad blockers remove a meaningful share of your traffic, when you are EU-focused with strict data requirements, or in sensitive categories where reducing the data footprint matters. It is not worth the operational cost for a small site with limited technical resource. If blockers are genuinely removing traffic, that is a strong argument; otherwise fix attribution and consent configuration first, which is cheaper and more often the real problem.
Do I need server-side tagging for GDPR compliance?
No. Consent Mode v2 with proper configuration addresses the compliance question. Server-side tagging addresses data control and blocker resistance, which are related but separate concerns. You can comply without it, and you can implement it and still be non-compliant if consent is mishandled — particularly if server-side events fire regardless of consent state.
How do I measure phone calls in GA4?
Use call tracking with dynamic number insertion, then attribute the call to its source. Better still, import offline conversions — booked jobs, closed deals, completed quotes — back into GA4 against the original source. For service businesses this is frequently the highest-value measurement work available, and it is almost never done.
Is open rate still a useful email metric?
No, not reliably. Apple Mail Privacy Protection preloads tracking images, inflating open rates for people who never opened anything. Judge email performance on click rate, donation or purchase rate per recipient, and revenue per send. Open rate has become a vanity metric that mostly measures email client behavior.
Should I trust GA4 attribution or Search Console?
Both, for different things. GA4 is better for on-site behavior and conversion volumes. Search Console is the ground truth for clicks from search. Where they disagree about organic sessions, GA4 is usually undercounting because of blockers and consent, not overcounting. Use Search Console to know how many clicks arrived, and GA4 to know what happened next.
Do I really need analytics if I have a small site?
Yes, but a simpler one. GA4 free is more than adequate for most small sites. What matters is that you track the few events tied to decisions you make, and import offline conversions if your revenue comes from phone calls or in-person sales. The expensive part is not the tool; it is maintaining data quality.
Keep reading
Want this applied to your business?
Stop guessing what to prioritize. Get a straight answer on where your marketing is leaking money, and what to fix first.
Book a free 30-minute audit