When someone decides to measure their website, the same questions come up. Here are concrete answers, straight to the point and with the nuances that only appear after reviewing real accounts.
What tool do I need to get started?
With two sources you cover 90% of the basics: a web analytics tool (Google Analytics 4 or a privacy-focused alternative like Plausible or Matomo) and Google Search Console. The first tells you what happens inside the site; the second, how Google sees you. If your website is small and you don’t sell online, Search Console can be even more useful at first: it shows queries, impressions, and CTR without bot noise.
Important nuance: GA4 changed the data model compared to Universal Analytics. Bounces no longer exist as a classic metric and many figures won’t match those from two years ago. It’s not an error, it’s a different model.
Why do GA4 and Search Console give different figures?
Because they measure different things. Search Console counts impressions and clicks from Google’s SERP; GA4 counts sessions that reach the site and execute the script. In between there are filters, blockers, cookie consent, and redirects. It’s common to see a 10-30% difference in clicks, and it doesn’t always indicate a problem. According to Google’s official documentation, both tools use different sampling and attribution methodologies.
Which metrics should I look at first?
- Sessions and users by channel, to see where people come from.
- Conversion rate by landing page, not overall.
- Landing pages with high traffic and low engagement time.
- Queries with high impressions and low CTR in Search Console: there are titles to improve there.
A common mistake: obsessing over total visits and not segmenting. A traffic spike from a country that isn’t your market can inflate the figure and contribute nothing.
How do I detect fake traffic or bots?
Typical signs: mass sessions with 0-second duration, improbable languages, exactly 1 pageview per session, and sudden jumps without a campaign. In GA4 you can create an audience with those conditions and exclude it from reports. There’s no perfect filter: sophisticated bots simulate behavior. If the business depends on analytics, it’s worth complementing with server logs to see which IPs and user-agents hit your website.
Is it worth installing a heatmap?
Yes, but with realistic expectations. Tools like Hotjar or Microsoft Clarity show clicks, scroll, and recordings. They’re useful for detecting specific usability problems (a button nobody clicks, a form that’s abandoned at the phone number field). They don’t replace quantitative analytics nor allow extrapolating statistical conclusions. And be careful with privacy: recordings can capture personal data if they aren’t masked.
How often should I review the data?
A weekly check of Search Console to detect drops and a deeper monthly review of channels and conversions. Daily only makes sense during launches or active campaigns. Analytics without a prior question becomes an expensive distraction.
What can’t analytics tell me?
It won’t tell you why a user left if there are no events configured, nor the real impact of a press mention if that visit arrives as direct traffic. Nor does it reliably attribute long journeys across multiple devices. Attribution is a model, not an absolute truth; it’s best read as a useful approximation for deciding, not as an accounting record.