The big idea: Analytics tells you what people actually do, not what you hope they do. Measure the actions that matter to the business, and let the data change your mind.
See it in action
The Case of the Missing Sales Lab
Sales at an outdoor store dropped fast. Switch dashboard views like a GA4 analyst, log the clues, then solve the case.
Trailhead Outfitters is a fictional store. All numbers are made up for practice.
Weeks 1-2 (before)Weeks 3-4 (after the site update)
Key concepts
- Event-based tracking: GA4 records actions as events, such as page views, clicks, scrolls, video plays and purchases.
- Key events: Mark the actions that matter most to the business, like a sign-up or purchase, as key events so you can track them.
- Acquisition reports: See where visitors come from (search, social, email, referral, paid) and which sources bring people who take action.
- Engagement: Engaged sessions and engagement rate show whether people actually spend time with your content.
- UTM parameters: Add tags to campaign links so you can tell which post, email or ad sent each visitor.
Learn it in depth
The deep dive: GA4 analytics
GA4 in plain words first, then the detail. Open the “Go deeper” boxes when you want more.
1The model: everything is an event
GA4 records actions. A page view is an action. So is a scroll, a video play or a purchase. Each action is an event, and events carry details called parameters.
session_start event and ends after 30 minutes of no activity (you can change the timeout).purchase or generate_lead. GA4 used to call these conversions. Google Ads still says conversions.Go deeper: the four kinds of events
- Automatically collected: GA4 sends these with no setup, like
first_visit,session_startanduser_engagement. - Enhanced measurement: a switch in the data stream settings. Adds page views, scrolls (90% of the page), outbound clicks, site search, file downloads, form interactions and embedded YouTube video plays.
- Recommended: names Google has already defined, like
add_to_cart,begin_checkout,purchase,sign_upandgenerate_lead. Use these names when they fit: ecommerce reports depend on them. - Custom: your own names for anything else, like
quiz_complete. Use lowercase with underscores and keep names consistent.
Custom parameters only show up in standard reports after you register them as custom dimensions or metrics. Register them early: they do not fill in backward.
2Core reports and the question each answers
Start with the question, then pick the report. Menus can be customized, so names and placement vary by property.
| Report | The question it answers |
|---|---|
| Realtime | Is tracking working right now? Did the email we just sent bring people in? |
| User acquisition | How did new users first find us? |
| Traffic acquisition | Where did each visit come from, and which sources lead to key events? |
| Pages and screens | Which content gets viewed, and for how long? |
| Landing page | Where do visits start, and do people stay and act? |
| Events and key events | What are people doing, and how often do the important actions happen? |
| Ecommerce purchases | Which products get viewed, carted and bought, and which get viewed but not bought? |
| Retention | Do people come back after their first visit? |
| Tech (device, browser) | Is one device or browser performing worse than the others? |
| Demographics | Where are our users, and what language do they use? |
Go deeper: reports vs explorations
Reports are fast, pre-built and shared with everyone on the property. Explorations are custom analyses you build yourself (see section 6). Use reports for the weekly check and explorations to dig into a question.
Add a comparison (for example, mobile vs desktop) to almost any report. It is the quickest way to spot a problem hiding inside a total, like the broken mobile checkout in the lab above.
3Where traffic comes from
GA4 labels every visit with three main tags. Think of them as where, how and why.
Default channels you will see
Go deeper: first user, session and attributed scopes
- First user source (User acquisition report): how the person found you the very first time. It never changes for that user.
- Session source (Traffic acquisition report): where this specific visit came from.
- Plain “Source” in key event reports: the source GA4 gives credit for that key event, using the property’s attribution model (data-driven by default).
Same person, three answers. A shopper who found you through Instagram, came back from an email, then bought after a Google search has a first user source of instagram and a final session source of google. Know which scope you are reading before you say a channel “works.”
Google Ads traffic is tagged automatically with auto-tagging (the gclid), so you do not need UTMs on Google Ads links. You can also build custom channel groups when the default rules do not match how your team plans.
4UTM builder
UTM parameters are tags you add to the end of a link. They tell GA4 exactly which email, post or ad sent the visit. The worked example below is filled in for a fictional store. Change any field and watch the link update.
Your tagged link
Go deeper: UTM rules teams live by
- Lowercase everything. GA4 treats “Email” and “email” as two different values.
- No spaces. Use underscores or hyphens. Spaces turn into %20 and make reports messy.
- Keep a shared naming sheet. One row per link: date, URL, source, medium, campaign, content, owner.
- Never tag internal links (links from your site to your site). The UTM overwrites the real source of the visit.
- Use utm_content for A/B tests, for example hero_button vs footer_link, so you can see which link in the same email worked.
5Engagement rate vs bounce rate
GA4 asks a better question than “did they leave?” It asks “did they engage?”
- Lasted longer than 10 seconds (the default; you can raise it)
- Triggered a key event
- Had 2 or more page or screen views
Engagement rate = engaged sessions ÷ sessions
Bounce rate = 100% − engagement rate
Go deeper: why old bounce rates do not compare
In the old Universal Analytics, a bounce was a single-page visit, even if the person read for five minutes. In GA4, that same visit counts as engaged, so it is not a bounce. Do not compare a GA4 bounce rate to an old benchmark. Compare GA4 to GA4: this month vs last month, this page vs that page.
Average engagement time counts only the time the page was in focus on screen. It is a more honest measure of attention than the old “time on page.”
6Explorations: funnel and path
Explorations are your workbench. Two you will use most:
Go deeper: settings that change the answer
- Closed vs open funnel. Closed: people must enter at step 1. Open: people can join at any step. Checkout funnels are usually closed.
- Directly vs indirectly followed by. “Directly” means the next event must be the next step. “Indirectly” allows other actions in between. Most real shoppers wander, so start with indirectly.
- Trended funnel shows each step over time, which is how you spot the day something broke.
- Explorations are private to the person who built them until shared, and they only reach back as far as the property’s data retention setting.
7Privacy and data gaps, in plain words
GA4 never shows “every visitor.” These four things explain most of the gap.
Go deeper: the details
- Consent signals: consent mode uses four signals:
analytics_storage,ad_storage,ad_user_dataandad_personalization. Google requires consent mode for using its ad features with visitors in the European Economic Area. - Basic vs advanced consent mode: in basic, tags send nothing until the visitor agrees. In advanced, tags send cookieless pings when the visitor declines, and GA4 can model (estimate) the missing behavior if the property has enough traffic.
- Thresholds show up most often when Google signals is on or when you use demographics with small audiences.
- Sampling: on standard (free) properties, explorations sample when a query goes past about 10 million events. Shorten the date range or use standard reports to avoid it.
- “(other)” rows appear when a report has too many unique values, like thousands of unique page URLs.
- Data retention: explorations can only use data back to the retention limit. Standard properties default to 2 months. Change it to 14 months on day one.
8Connecting to Looker Studio
Looker Studio is Google’s free dashboard tool. It connects to GA4 so your boss sees one clean page instead of ten reports.
- Create a report and add the Google Analytics connector.
- Pick the account and GA4 property. You need access to the property yourself.
- Build from the question. A scorecard for key events, a time series for the trend, a table by channel.
- Add a date range control and filters (device, channel) so viewers can explore on their own.
- Share with view access and schedule an email delivery for the weekly meeting.
Go deeper: why the dashboard and GA4 might not match
- The connector pulls through the GA4 Data API, which has usage quotas. Heavy dashboards with many charts can hit them and show errors. Fewer charts per page helps.
- Thresholds and “(other)” rows apply here too, so small numbers can differ from what you see in GA4.
- Blend GA4 with other sources, like Google Ads or Search Console, to put spend, clicks and results side by side.
- For large or long-term analysis, GA4 can export raw event data to BigQuery (free to set up, with limits on standard properties). That is where data analysts take over.
9A 30-minute weekly analytics routine
Same day, same order, every week. Check items off as you practice.
Try it
Open a GA4 demo account or a site you have access to. Find the top three traffic sources from the last 30 days, then find which one sends visitors who complete a key event. Write one sentence on where you’d put more effort, and why.
Tips and best practices
- Write the measurement plan before you touch GA4. List the three to five business actions that matter, like purchase or generate_lead, and mark only those as key events. If everything is a key event, nothing is.
- Set data retention to 14 months on day one. Standard properties default to 2 months, and explorations cannot reach back further than the retention setting. You cannot recover the lost history later.
- Tag every link you control with UTMs, in lowercase. Use source, medium and campaign on every email, social post and partner link, and keep one shared naming sheet. Never tag links from your own site to your own site.
- Watch Direct and Unassigned as a tracking alarm. When either one grows, it usually means untagged or messy links, not a new audience. Fix the tags before you report on channels.
- Segment before you conclude. Break every big change down by device, channel and landing page. A total that drops 30% often hides one broken step on one device.
- Check tracking after every site change. After a release, test key events yourself on a phone and a laptop and confirm them in Realtime. Add an annotation with the release date so future analysts know what changed.
- Use recommended event names when they fit. Names like add_to_cart, begin_checkout and purchase unlock GA4’s ecommerce reports. Register custom parameters as custom dimensions early, because they do not fill in backward.
- End every analysis with a decision. Write three sentences: what happened, why you think so, and what the team should do next. A chart without a recommendation is a report nobody acts on.
Quick check
Think through your answer, then select a question to reveal one.
What is a key event in GA4?
An action that matters to the business, like a purchase or sign-up, that you’ve marked so GA4 tracks it as a goal.
Why add UTM parameters to a link?
So your reports show exactly which campaign, post or email brought each visitor.
Go deeper
- Google Analytics Academy and the GA4 certification (Google Skillshop)