Analysing Web Traffic in Google Analytics 4: From Vanity Metrics to Real Insight

Most teams that own a GA4 property are not short of numbers. Sessions are up, users are down, engagement rate moved two points, and nobody in the room can say what to do about any of it. That is the real problem with web traffic reporting in Google Analytics: the default reports answer “how much” when the business question is almost always “which traffic is worth more of our budget”. The fix is not staring harder at the same dashboard. It is changing which parts of GA4 you read, and in what order.

Key Takeaways

• Want faster measurement clarity? Start with digital web analytics implementation and tracking discipline.
• Need reporting your team can act on? See data analysis for dashboards, reporting and decision-ready summaries.
• If the traffic itself is low quality, targeting and on-page signals are the lever. Pair SEO consulting with your GA4 analysis.

What GA4 traffic actually means: sessions, users, engaged sessions

Start with the core confusion. In Google Analytics 4, everything is an event first, and metrics are rolled up from those events afterwards. That is why the numbers look different from Universal Analytics: it is not a discrepancy to be reconciled, it is a different measurement model.

Three metrics you will see everywhere:

• Users: unique users over the selected date range, based on the identifiers available and the reporting identity you have configured.
• Sessions: still reported, but treat it as a reporting construct rather than the outcome.
• Engaged sessions: sessions that met GA4’s engagement criteria, which by default means lasting longer than ten seconds, firing a key event, or generating at least two page views.

From those, GA4 derives engagement rate: engaged sessions divided by sessions. This is the part that trips teams up in a useful way. Where Universal Analytics let you conflate volume with quality, GA4 forces you to keep them apart.

Three reasons your GA4 numbers will not match the ad platforms:

• Attribution logic: GA4’s attribution model and lookback window do not match a platform’s click attribution window.
• Measurement scope: ad platforms report on clicks and conversions tied to their own accounts, GA4 reports on what happened in your property.
• Event definitions: a “conversion” in one tool is not automatically the same thing as the key event you configured in GA4.

The GA4 report to open first: traffic acquisition, channel and landing page

When you are trying to make sense of a traffic report, do not start with “top pages”. Start with the acquisition story, because that tells you where visits came from and what kind of journey they are starting.

The report paths worth memorising:

• Traffic acquisition: Reports → Acquisition → Traffic acquisition
• Channel breakdown: use the Session default channel group dimension in that same table
• Landing pages: Reports → Engagement → Landing page
• Source and medium: in the acquisition view, swap the primary dimension to Session source / medium

The distinction that matters:

• Session default channel group applies GA4’s own channel rules. It shifts when your UTMs are missing or malformed, or when a referral condition is met.
• Session source / medium shows the raw attribution inputs GA4 recorded from your tagged URLs. This is the “what should have happened” lens.
• The Landing page report shows where people actually arrived, which is your first real intent signal.

Read those three lenses together and a vague problem becomes a specific one: did we lose campaigns, lose tagging, or lose landing page performance? Three very different fixes, and only one of them belongs to the media team.

Segment by intent, not by volume

Here is the shift that tends to stick with B2B teams. Instead of asking which channel brought the most sessions, ask which traffic streams produce the behaviours that precede your outcomes.

The dimensions that carry intent:

• Landing page: group entry pages by topic, funnel stage or product area rather than looking at them one URL at a time.
• Campaign tagging: use your UTMs to map traffic to demand types, for example product page, case study, webinar, comparison.
• Geography and device: useful as a supporting lens only. Intent almost always comes from the page and the message match, not the hardware.

Building this in GA4 without creating a mess:

1. Pick five to fifteen landing pages that genuinely represent your funnel stages, top through bottom.
2. Define the intent actions as key events: requested a demo, booked a call, submitted the contact form, opened a pricing enquiry.
3. Compare engagement rate, key event rate and conversion paths across those landing page cohorts.

You do not need every possible segment in one report. You need a handful of repeatable slices that answer the same business question every week.

Judge traffic quality with engagement rate and key events, not bounce rate

Bounce rate was a blunt instrument: a single-page visit that answered the user’s question perfectly counted the same as an instant back-click. GA4 gives you better signals, so use them.

What to measure instead:

• Engagement rate by channel, campaign and landing page
• Key events by source and medium, so you can see which traffic converts rather than which traffic arrives
• Conversion paths, using Explorations to see the sequence between arrival and outcome

Where to find each one:

• In the acquisition views, add engagement rate as a column and read it against sessions on the same timeline.
• In Explorations, build a funnel or path exploration using your own event names.
• In the Landing page report, look for entry pages that produce engaged behaviour rather than page views.

When traffic looks worse, you should be able to say which of three things happened: fewer people came, the same people engaged less, or the conversion path itself broke. That is analysis you can act on before the next budget meeting.

Data-integrity traps that break your GA4 traffic numbers

If the numbers feel unreliable, the cause is usually upstream. GA4 is accurate when the inputs are consistent, and it will not repair bad inputs on your behalf.

The traps worth checking, roughly in order of how often they bite:

• Unfiltered internal traffic: staff visits, QA runs and staging traffic inflate everything. Define internal traffic rules in Admin → Data Streams → Configure tag settings, then make sure the data filter is actually set to Active rather than Testing.
• Missing UTMs: untagged campaign links collapse into direct traffic or land in the wrong channel group, which makes paid and email look like they underperform.
• (not set): this shows up when the field GA4 needs was never captured. Treat a rising (not set) share as an alarm about collection health, not as a row to filter out of the report.
• Self-referrals: a misconfigured referral exclusion list makes your own domains appear as referral sources and starts new sessions mid-journey.
• Cross-domain gaps: if users move between domains without the linker working, one journey fragments into two sessions and the conversion gets credited to the wrong source.
• Consent-mode modelling: when consent rates or banner behaviour change, what GA4 records changes with them. Expect reporting shifts even in a week when marketing did nothing differently.

What to do when direct traffic jumps:

• Compare UTM coverage across the same date ranges and look for a tagging regression, usually from a new campaign or a new tool.
• Check the channel breakdown so you can see whether traffic has been reclassified rather than lost.
• Confirm internal filtering is still active and that no testing tool started firing events into the production property.

These checks turn “GA4 looks wrong” into “GA4 is telling us which part of our data pipeline broke”, which is a much shorter conversation.

Build explorations that answer a business question

Standard reports show what happened. Explorations let you ask why, at least from the angles you choose. The trap is building explorations that look impressive and answer nothing.

Three templates that earn their keep:

1. Funnel: entry page to conversion
   • Start with a landing page cohort, step to the first key event, then to the conversion event.
   • Compare two time windows, before and after a campaign or site change, and see which step lost people.
2. Path exploration: which journeys actually matter
   • Work backwards from a key event to see the sequences that lead into it.
   • Filter by channel group or by campaign name so you are comparing like with like.
3. Cohort comparison: engaged traffic versus outcomes
   • Bucket users by early engagement behaviour, engaged session or not, and track how many reach a key event.
   • This keeps the analysis anchored to intent rather than to arrival counts.

On choosing dimensions that will not confuse stakeholders: use source and medium together with campaign for campaign performance, use the channel group only after you have validated tagging coverage, and lead with landing page whenever the question is about on-site intent.

Worked example: diagnosing a traffic drop without guessing

Say someone reports that sessions are down this week. Do not start with sessions. Start with data integrity, then acquisition, then quality, then the path.

Step 1, validate the measurement basics. Using identical date ranges, check whether (not set) is spiking, which points at missing fields. Compare the direct traffic share against the previous period, because a jump there usually means UTMs or referral handling. Confirm internal traffic filtering is still in place.

Step 2, isolate where the drop lives. Open Reports → Acquisition → Traffic acquisition and read the drop by session default channel group. Then swap the primary dimension to session source / medium and read it again. If a single channel group falls while source and medium look stable, you are probably looking at a reclassification caused by a tagging change, not lost traffic. If several source and medium rows fall together, look at campaign delivery or at whether the landing pages were reachable.

Step 3, test whether quality dropped too. Go to Reports → Engagement → Landing page and compare engagement rate week over week for the same entry pages. Then check whether key event rates moved with it.

Step 4, pinpoint the break. Build an exploration funnel from that landing page cohort to the first key event and on to the conversion, then compare the two date ranges side by side.

What you do with the answer: if the landing pages are unchanged but engagement rate fell, look for a message mismatch or a page load problem. If engagement rate held steady but key events fell, look at the form, the conversion step, or a tracking change that quietly stopped an event from firing. Those are three different teams and three different fixes, which is exactly why the diagnosis order matters.

A practical weekly and monthly review routine

Reporting stays useful when it is a habit rather than an investigation. A routine that combines measurement health, traffic quality and outcomes will catch most problems while they are still small.

Every week, thirty to forty-five minutes:

• Acquisition scan: review traffic acquisition by channel group and note any shift large enough to explain to someone else.
• Tagging health: check movement in (not set) and any sudden rise in direct traffic.
• Quality signal: compare engagement rate for your top landing pages against the previous week.
• Outcome check: confirm key events are still firing and the conversion path is intact.

Every month, one to two hours:

• Intent review: revisit the five to fifteen landing pages and see which groups generate engaged sessions and key events.
• UTM hygiene: audit naming consistency across campaigns and fix whatever is causing misclassification. Lowercase everything, and pick one word for each medium.
• Cross-domain and consent check: validate cross-domain journeys and review what consent-mode modelling is doing to key event recording.
• Decision log: write down the business question, the dimensions you used, and the decision you took. Next quarter you will want to know why you moved the budget.

Keep that rhythm and analytics becomes an operating system for the team rather than a monthly spreadsheet scramble.

How our team supports GA4 traffic analysis

Teams get stuck between “we have the data” and “we don’t trust it”. Our work is closing that gap, so the reporting actually supports the next decision.

Where we usually start:

• Digital web analytics: implementation, tracking design and attribution modelling.
• Data analysis: turning sessions and users into engagement, key events and conversion path comparisons.
• SEO consulting: aligning landing pages and messaging with the intent your acquisition actually brings in.

That combination matters because a traffic problem is rarely only a measurement problem. It is usually some mix of tagging, traffic quality and landing page intent.

Conclusion

If your web traffic reporting in Google Analytics feels like vanity, that is usually not GA4’s fault. Read the acquisition story before the page list, trust engagement signals over bounce rate, and measure outcomes through key events and conversion paths. Once the underlying tagging is clean and the channel groups are defensible, the arguments about session counts stop and the decisions start.

Frequently Asked Questions

What should I look at first in a GA4 traffic acquisition report?

Start at Reports → Acquisition → Traffic acquisition, read it once by session default channel group and once by session source / medium. Then pair it with the Landing page report so you can connect each traffic stream to real intent and to engagement rate.

Why do my GA4 sessions and users not match Universal Analytics?

GA4 rolls events up into metrics using different rules, so the totals will never line up perfectly. Rather than trying to force a match, anchor your reporting on engagement and key events, which are comparable over time within GA4 itself.

How do I diagnose a drop in traffic in Google Analytics 4 without guessing?

Check (not set), any spike in direct traffic, and your UTM coverage first, since those explain a surprising share of apparent drops. Then isolate the fall by channel group and by source and medium, and confirm whether engagement rate and key events moved with it in the Landing page report and in Explorations.

Is engagement rate a better quality signal than bounce rate in GA4?

In most cases, yes. Engagement rate reflects whether someone did something meaningful rather than whether they viewed a second page, and it connects directly to your key events. Bounce rate in GA4 is simply the inverse of engagement rate, so it adds nothing new.

What causes “(not set)” traffic in GA4?

It appears when the field GA4 needs for a dimension was never captured, most often because of missing or broken UTM parameters. Referral handling and consent changes can also produce it, so check collection health before drawing any conclusion from the report.

How should I use UTMs for traffic analysis?

Tag consistently so campaigns map reliably into channel groups and source / medium, and standardise on lowercase values with one agreed term per medium. Then judge each campaign on engagement rate and key events rather than on the volume it delivered.

What is the difference between session default channel group and source / medium?

Session default channel group applies GA4’s own grouping rules, while source / medium shows the raw attribution inputs recorded from your tagged links. When the acquisition story looks odd, check both, because missing tags push traffic into buckets nobody expects.