The caveat first: it undercounts
Since consent banners became the norm, GA4 records only visitors who accept analytics cookies. How many decline varies by site and by banner design, but it is a large share. On a retail site we run ourselves, one month showed 337 clicks from Google Shopping in Google’s own Merchant Centre and 54 sessions from that channel in GA4. Same visitors, one sixth recorded.
So GA4 is not the place to learn how many people visited. It is the place to learn what the ones who consented did, and to watch trends, on the assumption that consenters behave roughly like everyone else. For the true count of Google search visitors, Search Console counts every click regardless of consent.
Set up once, or none of this works
- Mark the actions that matter as key events (GA4’s name for conversions): form submitted, phone number tapped, booking made, purchase. Without these, GA4 measures traffic and nothing that pays.
- Link Search Console and, if you advertise, Google Ads, so their data appears in GA4 and conversions can be imported.
- Exclude your own visits with an internal traffic filter, or your testing inflates everything.
- Set data retention to 14 months, the maximum on the free tier; the default is two.
The GA4 setup guide on this site walks through each of these. Ten minutes of reports a month is only worth it after this hour of setup.
The five reports
| Report | Where | Answers | Look for |
|---|---|---|---|
| Traffic acquisition | Reports → Acquisition → Traffic acquisition | Where did consenting visitors come from: organic search, direct, paid, social, referral? | Which channel produces key events, not just sessions |
| Pages and screens | Reports → Engagement → Pages and screens | Which pages were viewed, for how long? | Pages with views and no engagement: wrong visitors or slow load |
| Key events | Reports → Engagement → Key events (or Conversions) | How many forms, calls, purchases, and from where? | The trend by month, and by channel and landing page |
| Landing page | Reports → Engagement → Landing page | Which pages bring people in, and do those people convert? | High entrances and low key events: a page that attracts but does not persuade |
| Tech: device and browser | Reports → Tech → Tech details | Phone or desktop? Which browsers? | A conversion rate much lower on mobile means a mobile problem |
Source: Google Analytics 4 standard reports as of September 2026; menu names vary slightly by property configuration.
Reading it without being misled
- Full calendar month against the same month before. Never the default 28 days, never seven.
- Read key events before sessions. Sessions are input; key events are output.
- Bounce rate in GA4 is the inverse of engagement rate (under ten seconds, no key event, one page). It is hidden by default for a reason; look at engagement and key events instead.
- Direct traffic is partly untagged traffic: email links, apps, and visitors whose source GA4 could not see. Do not read “direct” as “typed the address”.
- Small numbers swing. Ten key events a month against twelve is noise. Judge on quarters.
- When a number changes suddenly, check tracking before checking the business: a tag broke, a banner changed, a filter was added.
The ten minutes a month
- Key events: total this month against last, and against the same month last year.
- Key events by channel: is the mix changing?
- Landing pages: any page with rising entrances and no key events?
- Device: is mobile converting far below desktop?
- Then Search Console for the true search numbers, because GA4 cannot give them.
