Analytics and Reporting
Most analytics setups answer questions nobody asked, and miss the two that the business actually needs. Ivan Sp implements, troubleshoots, and reports on the tracking that decides whether any other work on a site can be proven to have worked. Analytics without a clear baseline cannot show improvement, and a report without plain-language context leaves a client guessing at what a number actually means.
Four Areas This Practice Covers
- GA4 and Google Tag Manager implementation
- Google Tag Manager troubleshooting
- Conversion and funnel tracking
- SEO reporting
Scoped Deliberately, Not Maximized
I scope every implementation to the questions a client actually needs answered, rather than to the most granular tracking technically possible. A tracked event nobody analyzes carries a real ongoing maintenance cost, since every site change risks breaking it. Right-sized tracking survives, gets used, and stays worth the effort of maintaining.
From Setup Through Ongoing Reporting
Implementation establishes the baseline. Troubleshooting keeps existing tracking honest when something breaks. Funnel tracking measures specific conversion paths in detail. Monthly reporting turns the accumulated data into plain-language findings a client without a search background can actually act on.
Where the Standard Analytics Advice Breaks Down
- The usual advice
- Track everything. Data you did not capture is data you can never analyse later.
- What holds up in practice
- Every event is a thing that can break silently, and a container with 47 events takes hours to debug rather than minutes. The binding constraint is attention, not storage.
Imagine a property configured with 47 custom events and 18 marked as conversions. The setup is technically impressive and complete.
Two reports get read each month. When a booking stops being attributed, finding the broken tag among 47 takes most of a day.
I write the business questions first and build the smallest setup that answers them. Nine events that are trusted beat 47 that are merely present.
An overbuilt property is not free. Every extra event is maintenance, another surface for consent questions, and one more place a silent failure can hide.
Four Things Commonly Assumed About Analytics
How many of the reports in your analytics account did anyone read last month?
I meet these four assumptions on most measurement engagements.
| Commonly assumed | What actually holds |
|---|---|
| Track everything now, analyse it later. | Every event is a thing that can break silently. A container with 47 events takes hours to debug rather than minutes. |
| GA4 numbers should match the old analytics numbers. | GA4 counts events and the previous model counted sessions. The two define a visit differently, so the totals will not reconcile. |
| Default settings are sensible defaults. | The default data retention is two months, which quietly prevents any year-on-year comparison. |
| A conversion is any useful interaction. | Marking eighteen key events makes none of them meaningful. Key events should represent real business value. |
| Real-time data proves tracking works. | Real-time proves a page view fires. Attribution faults are invisible there and are the more expensive problem. |
An analytics setup nobody reads is a monthly maintenance cost with no return. Nine trusted events beat forty-seven that are merely present.
The Measurement Stack
- GA4 model
- Event-based rather than session-based, which is the structural change from Universal Analytics.
- Universal Analytics
- Stopped processing new data in standard properties on 1 July 2023.
- Automatic events
- page_view, session_start, first_visit, and user_engagement, collected without configuration.
- Enhanced measurement
- Scroll, outbound click, site search, video engagement, file download, and form interaction.
- Default session timeout
- 30 minutes of inactivity, configurable in the property settings.
- Data retention
- Two months by default, extendable to fourteen for event and user data.
- Deployment layer
- Google Tag Manager, using containers, tags, triggers, variables, and the data layer.
- Search data
- Google Search Console, which retains roughly 16 months of query performance.
- Export path
- The GA4 BigQuery export, which removes both sampling and interface row limits.
A Worked Example: Overbuilt Against Scoped
Consider two implementations built for the same business in the same week. The comparison below is illustrative, written to show what scoping changes. No client engagement is described here.
| Measure | Maximised setup | Scoped setup |
|---|---|---|
| Custom events configured | 47 | 9 |
| Events tied to a business question | 9 | 9 |
| Key events marked | 18 | 4 |
| Reports actually read each month | 2 | 5 |
| Third-party booking attribution working | No | Yes |
| Time to debug a broken tag | Hours | Minutes |
Ivan Sp scopes to the questions the business needs answered. Both setups captured the same nine useful events, and one buried them under thirty-eight more.
An overbuilt analytics setup is a dashboard with forty gauges in a car with one driver. Every gauge works, and nobody can find the fuel level.
The booking attribution row is the real cost. Effort spent on events nobody reads is effort not spent closing the gap that mattered.
How an Analytics Engagement Runs
I start from the questions rather than from the tool, which is what keeps the setup small.
- Write down the questions the business needs answered, before opening any analytics interface.
- Map each question to the single simplest event that would answer it.
- Deploy the GA4 tag through Google Tag Manager rather than hard-coding it into the theme.
- Enable enhanced measurement, then disable any part of it that produces noise for this site.
- Configure only the custom events the mapped questions require.
- Mark as key events only those actions that represent real business value.
- Close third-party attribution gaps, such as bookings handled by an external scheduling tool.
- Extend data retention to fourteen months, since the default two months prevents year-on-year comparison.
- Verify every tag in Tag Manager preview mode and in GA4 DebugView before going live.
- Enable the BigQuery export where reporting needs to join sources or exceed interface limits.
How Analytics Work Is Scoped
I scope measurement by the number of questions, not by the number of events available.
- What arrives
- A measurement plan, a configured GA4 property, a Tag Manager container, and verification evidence.
- Typical duration
- Two to four weeks for a standard site, longer where third-party tools need bridging.
- What the client supplies
- GA4 and Tag Manager access, site access, and the list of business questions.
- Included
- Measurement planning, tag deployment, event configuration, verification, and documentation.
- Not included
- Server-side tagging infrastructure, and consent management platform licensing.
- What changes the estimate
- Ecommerce tracking, several domains, or booking flows on external platforms.
- How it is quoted
- After a short scoping call, once the question list is agreed.
Tools and Primary Sources
The references below define the platform behaviour named above. Each opens in a new tab.
- GA4 enhanced measurement lists exactly which interactions are collected without custom configuration.
- GA4 data retention settings documents the two and fourteen month options and what each affects.
- Google Tag Manager is the deployment layer for tags, triggers, and the data layer.
- GA4 BigQuery export documents the raw event export that removes interface row limits.
- Search Console Performance report supplies the query data GA4 does not carry.
Check Your Own Setup in Ten Minutes
I suggest three checks that expose the most common analytics faults.
- Open GA4, then Admin, then Data Retention, and check whether it still says two months.
- Open Admin, then Events, and count how many are marked as key events.
- Ask whether each key event represents real business value or merely an interaction.
- Open Reports, then Acquisition, and check whether your booking or form tool appears as a referral.
Two-month retention prevents any year-on-year comparison and takes one click to change. A booking tool showing as a referral means the conversion is credited to the wrong source.
Common Questions About Analytics and Reporting
Do I need GA4 and GTM if my site already has some tracking installed?
An audit of what already exists comes first. Existing tracking sometimes needs correction rather than a full rebuild, and the right starting point depends on what the current setup is actually capturing correctly.
How does this relate to the AI Agents reporting pipelines?
This practice covers reporting delivered as an ongoing service to a client. The engineering behind large-scale, BigQuery-based reporting pipelines is covered separately under custom SEO reporting pipelines.
Can analytics work happen before a site has meaningful traffic?
Implementation should happen as early as possible, regardless of traffic level, since a baseline captured after growth begins cannot measure that growth. Reporting becomes more useful once there is enough data to show a real trend.
Why does GA4 report different numbers from the previous analytics setup?
GA4 counts events rather than sessions, and the two models define a visit differently. Comparing GA4 totals against Universal Analytics totals is comparing two different definitions, so trends within GA4 are more reliable than cross-platform comparisons.
Discuss Analytics and Reporting Work
Analytics work gets scoped around what currently exists and what questions actually matter to the business. The first call is a free 30-minute consultation with no pitch attached.