Google Analytics 4 setup: what to configure before your site goes live
By Carlos Alberto Cardona B.· · reviewed on 14 min read
A clean Google Analytics 4 setup comes down to four decisions you make before the first visitor arrives. Set the right time zone and currency, and filter out your own traffic. Keep data retention (how long data is kept) at 14 months, and link Search Console.

The essentials
- What it is
- In GA4, a conversion is a key event: the action you mark as important, such as a form submission.
- Why it matters to you
- A WhatsApp click is an intention, not a client. Compare it every month with the messages you actually get.
- What to do today
- What was measured wrong stays wrong: filter your own traffic and define key events before launch.
I'll be honest, I'm terrible at GA4. So I gave the Claude extension for Chrome the goals of a family business site I work on and asked it to set things up. Not everything came out right, but I got faster and learned to tell the AI more clearly what I want to measure.
Below is what I found when I reviewed that property, always with its date range. I don't know yet whether that work brings in clients: it needs more time, and patience.
It's for a service business that wants to know one thing: whether the site brings in inquiries. What follows is the setup order I now use, mistakes included. It exists because on a GA4 property I manage, 19 of 25 WhatsApp clicks turned out to be my own local tests.
What does GA4 measure, and what does it not?
GA4 is Google's free analytics tool. It measures what happens after someone arrives. Its own tag doesn't record what they searched for on Google: the query data comes from Search Console, through the link between the two products.
GA4 is event-based: everything a visitor does is an event. Google's documentation separates the events collected automatically from the custom ones you define. The ones that matter to a service business are usually yours to define.
It also doesn't measure people whose browser blocked the script, and on a small site that share can matter. Treat GA4 numbers as a floor, not a census.
If you learned on Universal Analytics, the previous version, it stopped processing data on July 1, 2023, even though many tutorials still describe it. Search Console and Bing Webmaster Tools tell you which searches found you. GA4 sees what happens next, button by button.
How to set up Google Analytics 4 in the order that keeps data clean
Set up Google Analytics 4 before the first visitor arrives, in this order. First the time zone and currency, then the internal traffic filter, data retention and the Search Console link. The first three can't be fixed retroactively, so the order matters more than any report you build afterwards:
Create the property with the right time zone and currency BEFORE any traffic arrives. Google warns that changing the time zone later doesn't rewrite historical data, so you end up with a break in every report.
Turn on internal traffic filtering and add your own IP. Leave the filter in “Testing” for a day and check it's catching you. Then set it to “Active”: from then on, the excluded data is gone for good.
Hostname filter: make the property count only your public domain. If your platform also serves the site on a preview subdomain, everything you test there mixes with the real data.
One key event per goal, not ten. For a service site, the form submission. Google calls them key events; each one should match something you can also count outside GA4.
Set data retention to 14 months; on a standard property the only other option is 2. The setting only affects explorations and funnel reports, but 2 months means you can't compare a quarter with the one before it there.
Link Search Console to the property: it puts the query data next to the behavior data. Since it brings in up to 16 months of Search Console history, it's the one step that isn't lost if you do it late.
If you installed GA4 by pasting the code and nothing else, your own tests are probably counting as visits. It happened to me, which is why I use this order.
| Setting | Can you fix it later? | What happens if it's late |
|---|---|---|
| Time zone | Not retroactively | Earlier data isn't rewritten: every report gets a break |
| Internal traffic filter | Not retroactively | Tests already counted stay counted |
| Custom dimensions | Not retroactively | GA4 doesn't fill them in for earlier events |
| Data retention | Not retroactively | At 2 months, explorations lose the previous quarter |
| Search Console link | Yes | It brings in up to 16 months of search history |
One more thing that can't be fixed afterwards: register custom dimensions before the first visitor, not after, because GA4 doesn't fill them in backwards. The first one I register is a stable ID for the pressed button, not its text. That way the series doesn't break when the wording changes.
From the browser · 5 min
Whether your contact event reaches GA4
Why it matters: If the event isn't recorded, your reports say zero contacts while people are writing to you. And every decision about the site gets made blind.
That the action you count as a contact reaches GA4, once, from your published domain. Because in Google Analytics 4 a click isn't a contact.
- 1.
Open GA4, Reports, Realtime.
- 2.
In another tab open your published site, not the preview, and submit the form with a test message.
- 3.
Go back to Realtime and look for your event's name in the “Event count by event name” card.
What comes out tells you which of these cases you're in.
It appears once.
It's recorded. Now compare it every month with the messages you receive.
It doesn't appear.
The tag isn't on your published domain, or an internal traffic filter that's already active is excluding you. Try again from your phone on mobile data, off your network.
It appears twice, or under two names.
Something is counting the same thing twice, such as automatic click measurement plus your own event. Your contacts would be inflated.
Ask your AI assistant: the prompt to copy
Open it in your browser's side panel (Claude, Gemini or ChatGPT), with your page open, and replace what's in brackets. The rules at the end make it quote what it saw and say “not verifiable” instead of making things up.
These are the events I saw in GA4's Realtime report after submitting my site's form: [each event's name and how many times it appeared].This is the event I marked as key: [name].Was it recorded once, never, or more than once? If more than once, tell me which other event could be counting the same thing.Use only what I pasted; if something is missing to decide, tell me what.Rules for this review:1. Quote the exact text you found. Don't paraphrase it.2. If something can't be seen from the browser, write “not verifiable from the browser.” Don't assume.3. Give me a table: element · what you found · fine / review / missing · why it matters in one sentence.4. Don't make up traffic, visits or rankings. Don't promise anything will rank higher on Google.5. If you can't open an address yourself, say so in the first line and work only with what I pasted. Don't describe a page you didn't see.
What this check does NOT prove: The event being recorded doesn't prove the message arrived: only your inbox or WhatsApp can tell you that. And Realtime shows minutes, not a month.
When it's no longer a do-it-yourself job: the event doesn't show up even from another device and you don't know where the tag is installed. I'll review your site for free and send it to you in writing.
How do you measure conversions in GA4?
To measure conversions, mark as a key event the action that equals an inquiry, such as a form submission. Then read those key events by acquisition channel: that shows which source brings you inquiries and which one only brings visits.
Google defines a key event as one that measures an action that's particularly important to your business. They used to be called conversions, and many tutorials still use that word.
This is how I read the property of a family business site, the one in the measurable SEO and paid search plan. It covers 30 days, Sep 18 to Oct 17, 2026, read through the GA4 Data API.
| Channel | Sessions | Engagement rate | Bounce rate | Key events | Avg. duration |
|---|---|---|---|---|---|
| Organic Search | 153 | 62.7% | 37.3% | 0 | 301 s |
| Paid Search | 128 | 66.4% | 33.6% | 9 | 104 s |
| Direct | 62 | 17.7% | 82.3% | 0 | 98 s |
Organic brought more sessions and visits three times longer, but no key events. In that window, people who came to read didn't message on WhatsApp in the same visit.
Paid search did add 9 key events, and on that property key events are mostly WhatsApp clicks. A click isn't a client yet, and with 9 events I draw no conclusions: under about 30, it's a signal to watch, not a finding.
Setting up GA4 events: how do you track WhatsApp clicks?
Record the click as its own event, with the button's ID, and mark it as a key event if WhatsApp is your main contact channel. But the click only opens another app: whether the person writes is something else, so compare it every month with the chats you receive.
That's what I did with an October 2026 campaign. I set the WhatsApp clicks from GA4 and Google Ads next to the conversations in the client's own log. The result is the measured funnel from search to lead.
On one property, 19 of the 25 WhatsApp clicks recorded in 90 days were my own local tests (GA4, hostname breakdown, Sep 16, 2026). Before reading any conversion number, break it down by hostname. If you name events, name them once and write the names down.
On the family site, one landing page fired two events per WhatsApp click (5 and 5 in GA4, Sep 18 to Oct 17, 2026). If both were key events, every inquiry would count twice. Before you mark a key event, check that each button fires only one.
If you also want to know how many visits each social network sends you, tag your links with UTM parameters. I show you how many visits Instagram sends you, using your bio link as the example.
And don't confuse clicks with visits: the same person can press three buttons. GA4 lets you count each key event once per session, which answers how many visits tried to reach you. Google labels that option as legacy and recommends counting every event, so if you change it, write the date next to your numbers.
With fewer than about 30 events in a group, I don't draw conclusions. That's my working rule, not a Google rule.
With dirty data, the costliest decision is about money: raising or cutting ad spend while looking at inquiries that were your own tests. What was measured wrong stays wrong, so the order matters from today. Deciding what to measure, and how, is part of the 90-day SEO strategy, and the first step is the free diagnostic.
How do you mark a contact form as a key event?
Have the site send its own event when the contact form is submitted successfully, not when someone merely starts typing. Google recommends naming it generate_lead; then, in Admin, open Events and mark it as a key event.
On the family site, GA4 counted 45 started forms in 30 days (form_start event, Sep 18 to Oct 17, 2026). They looked like 45 half-finished inquiries.
They were people using the site's calculator. GA4's enhanced measurement records form_start the first time someone interacts with a form, and the calculator counts as one. In that window there wasn't a single form-submitted event.
What is bounce rate in GA4?
In GA4, bounce rate is the opposite of engagement rate: according to Google, it's the percentage of sessions that weren't engaged. An engaged session lasts longer than 10 seconds, has a key event, or has 2 or more page views.
I find it useful, not as a grade but as a question. On the family site, Direct traffic bounced 82.3%, against 37.3% for Organic (GA4, Sep 18 to Oct 17, 2026). A number like that doesn't say what's happening; it says where to look.
Avinash Kaushik put it bluntly in 2014 on his blog, Occam's Razor: dashboards full of numbers and no conclusions “are data pukes.” What he asks for instead is the “so what” behind every number.
How is conversion rate calculated in GA4?
On a service site, conversion rate is the percentage of sessions that ended in at least one key event. GA4 calls it session key event rate: sessions with a key event divided by total sessions.
On the family site, paid search added 9 key events in 128 sessions (GA4, Sep 18 to Oct 17, 2026). Since one session can hold more than one, the rate was 7% at most. With so few events, it's a count, not a conclusion.
How do you import GA4 key events into Google Ads?
Link GA4 to Google Ads, turn on auto-tagging and mark the event as a key event in GA4. Then, per Google Ads Help, go to Goals, then Summary, click Create conversion action and pick the events from your GA4 property.
The real question is what you import. On the family site, Google Ads counted the WhatsApp click as a conversion, with a value of 0 (Google Ads, Oct 6 to 18, 2026). In 146 clicks there were 2 of those conversions: two WhatsApp clicks, not two clients.
And don't look for them in GA4's standard reports: Google notes that Google Ads conversions don't appear there. Each tool counts its own.
How do you know which channel brings clients before you spend more?
Measure one service at a time: when you set up Google Analytics 4, add one key event for each way people can contact you. Compare those events every month with the clients you actually closed, and only when both counts match, decide which channel to scale.
There's a line attributed to John Wanamaker, though no one has found where he said it. It says half the money spent on advertising is wasted, and the trouble is not knowing which half. Today you can look for that half, but only if you measure the inquiry, not the click.
That's where I am with the family site: the events are marked, and it needs more time to show whether they bring clients. It takes patience, and not deciding where to invest before the count is in.
Launching a new site? If you're building it with AI, GA4 comes in at step 8 of how to build a website with AI. If you'd rather I build it, SEO website design includes GA4 and Search Console in your name.
Why read it next to Search Console, not instead of it?
The two products answer different halves of the same question. Search Console says what Google showed and who clicked; GA4 says what happened next. Either one on its own produces a confident wrong conclusion.
A page with impressions and no clicks is a title and description problem. A page with clicks and no engagement is a content problem, or a speed problem, which is what Core Web Vitals measure. You can't tell those apart from inside one tool.
Don't expect Search Console clicks and GA4 sessions to match: they measure different things and never line up. The one unit they share is the landing page, so compare page with page.
Reading the sources separately is what I did in a measurable SEO and paid search plan: GA4 keeps organic and paid apart there. When each channel makes sense is in SEO vs SEM.
If you'd like me to read the Search Console side with you, with read access, that's part of the technical SEO audit.
Want to check it yourself?
Four minutes, no setup:
Open Realtime, then open your own site in another tab. If you still appear there once the filter is “Active”, check the IP address it uses: every number you have may include you.
Open Admin, then Data Settings, then Data Retention. If it says 2 months, change it now: whatever is already older than that is gone from explorations.
Open Reports, then Acquisition. If “Organic Search” is close to zero on a site that's indexed, check that your tag is on every page, not just the home page.
Watch it explained, by the people who built it
Three short videos from Google's own channels. They're the same source the links above point to, in video form:
Reports in Google Analytics · Google Analytics, 16 minutes. A tour of the interface, which is the part no article replaces.
Create and manage events · Google Analytics, 4 minutes. Official Analytics Academy lesson on what an event is.
Key Events and Conversions · Google Analytics, 4 minutes. Why “conversions” became “key events”, which matters if you follow an older tutorial.
Frequently asked questions
GA4 measures what people do on your site; Search Console, which searches brought them from Google and how often Google showed you without a click. Query data reaches Analytics through the Search Console link, so you read them together.
The time zone, the currency and the internal traffic filter, before any traffic arrives. None of those can be fixed retroactively.
You need the Administrator role: in Admin, open Access management for the account or the property. Click “+” and “Add users”, enter their Google account email and pick a role; if they only need to read reports, Viewer is enough.
No. The click opens another app; whether the person writes is something else. Record it as a click and compare it every month with the messages you receive.
It's what GA4 now calls a conversion: an event that measures an action that's particularly important to your business. On a service site, the form submission or the WhatsApp click. You mark it in Admin, under Events.
With the free add-on “GA4 Reports Builder for Google Analytics”, from the Google Workspace Marketplace. According to its listing, it lets you create and run reports for your GA4 properties in Google Sheets. Filter out your own traffic first: whatever you export carries the same data you see in GA4.
In Admin, under Product Links, click BigQuery Links. You need the Editor role or above on the property, and Owner access to the BigQuery project. According to Google, a standard property exports up to 1 million events in each daily export.
This guide is for service sites, not stores. A store needs the ecommerce events Google documents, such as view_item, add_to_cart, begin_checkout and purchase. The order in this guide still applies: time zone, filter and host guard before launch.
Sources
Article reviewed on October 31, 2026. Each date is the source's last published update.
See the 23 sources
- 01Reports in Google Analytics · Google Analytics · video
- 02Create and manage events in Google Analytics · Google Analytics · video
- 03Key Events and Conversions in Google Analytics · Google Analytics · video
- 04About events · Google Analytics Help
- 05Set up a property · Google Analytics Help
- 06Filter out internal traffic · Google Analytics Help
- 07Data retention · Google Analytics Help
- 08Search Console integration · Google Analytics Help
- 09Change the counting method of key events · Google Analytics Help
- 10Key events · Google Analytics Help
- 11Add, edit, and delete users and user groups · Google Analytics Help
- 12Google Analytics 4 has replaced Universal Analytics · Google Analytics Help
- 13[GA4] Engagement rate and bounce rate · Google Analytics Help
- 14Analytics dimensions and metrics · Google Analytics Help
- 15Enhanced measurement events · Google Analytics Help
- 16[GA4] Recommended events · Google Analytics Help
- 17Create conversions from Google Analytics events in Google Ads · Google Ads Help
- 18Mark events as key events · Google Analytics Help
- 19Digital Dashboards: Strategic & Tactical: Best Practices, Tips, Examples · Avinash Kaushik · Occam's Razor, Jul 15, 2014
- 20One-Half the Money I Spend for Advertising Is Wasted… · Quote Investigator · Apr 11, 2022
- 21GA4 Reports Builder for Google Analytics · Google Workspace Marketplace
- 22Set up BigQuery Export · Google Analytics Help
- 23Measure ecommerce · Google for Developers
About the author. Carlos Alberto Cardona B., SEO specialist in Bogotá. I write what I measure, with every number next to its instrument and date, and I verify what AI executes myself.
