Google Analytics 4 arrived and quietly deleted the mental model most business owners had spent a decade building. Bounce rate vanished. Goals became something called key events. The reports moved, the numbers stopped matching, and the thing that used to answer "how many people visited my site last month" in four seconds now takes a tour of three menus. Plenty of owners logged in once, decided it was not for them, and never went back. That is a fair reaction to a badly explained product, but it also means nobody knows which marketing is producing enquiries.
This guide covers what genuinely changed and why, the reports that repay the effort, the metrics people misread most often, and how to measure the three or four things a small business cares about. We will also be honest about the parts that are overkill for a six-person company.
Key takeaways
- GA4 records events rather than pageviews, which is why the reports look unfamiliar and the numbers no longer match the old Analytics.
- Engagement rate replaced bounce rate, and it measures the opposite thing: the share of visits that showed interest rather than the share that did not.
- Out of dozens of reports, four are worth your attention and the rest can be ignored.
- Nothing important is measured until you configure it, so form submissions, phone taps and quote requests need setting up by hand.
- UTM campaign tags on the links you share are the only reliable way to know which marketing produced an enquiry.
- Consent banners mean your data is always incomplete, so treat GA4 as a trend tool rather than an accounting ledger.
- A handful of correctly configured events beats a dashboard nobody opens.
Who this guide is for
This is for the owner or manager of a small business who has a website, has GA4 installed, and cannot get a straight answer out of it. You do not need to be technical. You do need to spend about ninety minutes in the settings once, because GA4 out of the box measures almost nothing that matters to you.
It will be most useful if you run a service business, a local trade, a clinic, a studio or a small shop, where the goal is enquiries rather than thousands of transactions a day. If you run a large ecommerce operation with a data team, someone is already doing this properly.
One caveat first. GA4 tells you what happened on your website. It cannot tell you why, and it cannot value the lead who rang after seeing your van. Treat it as one instrument on the dashboard and you will make sensible decisions. Treat it as the truth and you will cut the marketing that quietly works.
Why GA4 rebuilt itself around events
Universal Analytics was built for the web as it looked in 2012. A visit was a session, a session contained pageviews, and everything else was bolted awkwardly to the side. That held up while people moved between pages on a desktop computer.
Then behaviour changed. Visitors arrive on phones, tap things, scroll halfway, switch to an app, come back, submit a form that never loads a new page. A model built on pageviews cannot describe that without a lot of bodging, so GA4 was rebuilt around one idea: every interaction is the same type of thing, an event, with parameters attached.
Once that clicks, most of the confusion goes. A page being viewed is an event called page_view. Somebody scrolling to the bottom is an event called scroll. Downloads, video plays, form submissions and the start of a session are all events in the same list.
What GA4 measures on its own
Some events arrive the moment the tag is installed, including page_view, session_start, first_visit and user_engagement. That is why a brand new property shows you traffic on day one without you doing anything.
On top sits Enhanced Measurement, on by default. It adds scroll tracking at ninety per cent page depth, clicks to other websites, site search, video engagement, file downloads and some form interactions. Check it is switched on under Admin, then your data stream, because developers occasionally turn parts of it off.
What it does not measure is the thing you care about most: somebody contacting you. Its form tracking is unreliable on the JavaScript-driven forms most sites now use, and a tap on a phone number is not counted at all.
Why your numbers stopped matching
Compare the old Analytics to GA4 for the same month and the figures will not line up. Sessions are counted differently, they no longer restart at midnight or when the campaign source changes, and users are estimated rather than counted exactly.
So comparing 2022 numbers to 2025 numbers is close to meaningless. Draw a line at the changeover and compare forward from there.
Users, sessions and engaged sessions, in plain English
Three metrics sit at the top of nearly every report and they get confused constantly. Straightening them out takes two minutes and saves a lot of bad conclusions.
A user is a device and browser combination that GA4 believes to be one person. Somebody who visits on their phone in the morning and their work laptop that afternoon is usually two users. Useful as an estimate, poor as an exact count.
A session is a visit. It starts when someone arrives and ends after thirty minutes of inactivity by default. One person can have many sessions, so somebody who checks your prices at lunch and returns at nine that evening is two sessions and one user.
What makes a session engaged
An engaged session is GA4 separating people who looked at your site from people who arrived and left. A session counts as engaged if any one of three things happens: it lasts longer than ten seconds, it includes two or more page views, or it triggers a key event.
That ten second threshold can be raised to sixty in the settings. If your pages are long reads, ten seconds is a low bar that flatters the figure. If your site answers a question in one paragraph, ten seconds may be a good visit.
Engagement rate, and what happened to bounce rate
Bounce rate measured single-page visits with no interaction, and it was probably the most misunderstood number in marketing. Somebody lands on your contact page, reads the number, rings you: that was a bounce, and a lead.
GA4 flipped it. Engagement rate is the percentage of sessions that were engaged, so higher is better. Bounce rate still exists as a metric, but it is now simply the inverse. Same information the right way round, phrased so that improving it means improving something real.
| Metric | What it actually means | What people wrongly assume |
|---|---|---|
| Users | Estimated distinct browsers and devices | An exact headcount of people |
| Sessions | Visits, ending after 30 minutes of inactivity | The same thing as users |
| Engaged sessions | Visits over ten seconds, or with two page views, or a key event | Visits from people who were genuinely interested |
| Engagement rate | Engaged sessions divided by all sessions | A quality score for your content |
| Views | Total pages viewed, including repeats by one person | The number of people who saw the page |
| Event count | Every recorded interaction, including automatic ones | Meaningful actions taken by customers |
| Key events | Only the events you deliberately marked as important | Confirmed sales or enquiries |
The handful of reports worth actually opening
The left-hand menu contains far more than a small business needs, most of it built for apps, ecommerce and enterprise cases you will never touch. Retention, Demographics, Tech, most of Advertising and all of Explore can be ignored without missing anything.
What follows earns its place. Open these once a month and nothing else, and you are ahead of nearly every small business owner we speak to.
Form one habit immediately: change the date range before reading anything. GA4 defaults to a short window, and a fortnight of data for a small site is mostly noise. Use twenty-eight days.
Traffic acquisition and landing pages
Traffic acquisition, under Acquisition, is the most useful report in the product. It tells you which channel each visit came from: organic search, direct, paid search, referral, social, email. The grouping logic deserves its own section, which is next.
Landing pages, under Engagement, tells you which page people arrived on. That is different from your most-viewed page and often more revealing. If one blog post brings in a third of your arrivals, it deserves a proper call to action on it.
Pages and screens
Pages and screens shows what people looked at once they were on the site. Sort by views, then read the engagement time column beside it, because high views with low engagement time usually means a navigation step rather than a destination.
The genuinely useful part is the bottom of the list. Pages with almost no views are buried, invisible to search, or serving no purpose.
Realtime, and how to use it properly
Realtime shows activity in the last thirty minutes. It is compelling to watch and watching it is almost always a waste of time. Two people are on your site right now, which is nice, and tells you nothing you can act on.
It becomes valuable as a testing tool. Set up a new key event, open Realtime on a laptop, then go through the action on your phone: fill in the form, tap the number, submit the quote request. If the event appears within a few seconds the tracking works, and if not you have found out in two minutes rather than two months.
For fiddlier problems there is DebugView, in Admin. It streams every event from a device you have put into debug mode, usually via Google Tag Manager's preview mode. That beats guessing why an event will not fire.
How GA4 groups your traffic into channels
Traffic acquisition sorts every visit into a channel group: Direct, Organic Search, Paid Search, Organic Social, Paid Social, Email, Referral, Display and a few others. GA4 decides which group a visit belongs to from where it came and, crucially, from any campaign tags on the link.
Organic Search means somebody found you through a search engine without clicking an ad. Referral means another website linked to you. Direct is the awkward one, because it is not really a channel. It is the bin GA4 uses when it cannot tell where a visit came from.
That bin fills faster than people realise. Links in text messages, links inside PDFs, desktop email clients, QR codes and apps that strip the referrer all land in Direct. A large Direct slice is rarely brand recall. It is usually untagged links.
Session channel versus first user channel
GA4 offers two versions of most acquisition dimensions and explains the difference badly. Session default channel group answers "where did this visit come from". First user default channel group answers "where did this person originally find us, ever".
To judge what your marketing did last month, use the session version. To see which channel introduces people who later become customers, use the first user version. Mixing them produces numbers that look wrong for no obvious reason.
When the grouping gets it wrong
Channel grouping is rule-based, and rules can be fooled. A newsletter sent through some platforms arrives as Referral rather than Email, a link posted in a private group arrives as Direct, and paid traffic tagged carelessly can sit in Organic Search quietly flattering your SEO figures. The fix is not to fight the rules but to tag your own links, which is the next section.
Key events: measuring what a small business actually cares about
Google renamed conversions to key events, which confused everyone who had only just learned the old word. The concept is unchanged: a key event is an event you have flagged as mattering, so GA4 counts it separately and lets you compare channels by it.
This is where GA4 becomes worth the effort, and it is the step almost everybody skips. Without key events you have a traffic counter. With three or four you can see which marketing produced enquiries rather than visits.
For a small business the list is short: a contact form submission, a tap on the phone number, a quote or booking request, possibly a price list download. Resist marking twelve things as key events, because when everything is important nothing is.
Form submissions and quote requests
The reliable approach is to send people to a dedicated thank-you page after submitting, then fire an event on that page. It works regardless of how the form was built, survives redesigns, and you can verify it by filling in the form yourself.
If your form submits without changing the page, which is common now, you need either your form plugin's own tracking option or a small Tag Manager setup that listens for the submission. Most decent form tools have a setting for this, and if yours does not, that is a fair reason to change tools.
Phone taps and email clicks
On a phone, tapping the number is the most common way a local business gets contacted, and GA4 does not record it by default. The link is a tel: link rather than a link to another website, so Enhanced Measurement's outbound click tracking ignores it entirely.
Setting it up means adding a click trigger for links starting with tel: and firing an event when one is clicked. It is a fifteen minute job in Tag Manager, and the highest-value tracking change most local businesses can make. The same applies to mailto: links.
Marking an event as a key event
Once the event exists and you can see it arriving, go to Admin, then Events, find it in the list and switch on the key event toggle. It only affects data collected from that point on, so do it early.
One catch: GA4 counts every occurrence, so a form submitted twice records two key events. Sanity check against your inbox rather than trusting the number blindly.
UTM tags: knowing which marketing produced the enquiry
A UTM tag is a bit of text added to the end of a link that tells GA4 where the click came from. They are not technical, they are not something you install, and they cost nothing. They are extra words on the end of a web address.
Without them a large chunk of your marketing is invisible. The link in your newsletter, the link on your Google Business Profile, the QR code on a flyer, the link pasted into a Facebook group: several land in Direct or Referral with no way of telling them apart. With tags, each arrives labelled.
The mechanics are simple. Add a question mark to the end of your web address, then the parameters, separated by ampersands. The page loads as normal and the visitor sees nothing unusual apart from a longer address.
The parameters worth using
- utm_source. Where the click came from as a named place: newsletter, facebook, gbp, flyer.
- utm_medium. The type of channel: email, social, cpc, print, referral. GA4 uses this to decide the channel group, so stick to the standard words.
- utm_campaign. The specific piece of marketing: spring-offer, march-newsletter, new-showroom.
- utm_content. Optional, for telling apart two versions of the same thing: header-button versus footer-link.
- utm_term. Rarely needed outside paid search, where it holds the keyword. Ignore it.
A naming convention that survives contact with reality
The failure mode is not forgetting to tag. It is tagging inconsistently, so six months later your reports show facebook, Facebook, FB and facebook.com as four separate sources. GA4 treats those as different things, because they are different strings.
Pick rules and write them down somewhere you will find again. Everything lowercase, hyphens instead of spaces, a short list of allowed mediums and no improvising. Keep a spreadsheet of every tagged link with a column for where it was used, because in a year you will not remember what spring-2 meant.
One firm rule: never put UTM tags on links between pages of your own site. That starts a new session and wipes the original source, so an enquiry that genuinely came from search gets credited to your own internal link. It corrupts everything downstream of it.
Tagged links are the closest thing analytics has to a free lunch. Ten minutes of naming discipline turns "we think the newsletter works" into a number you can point at.
Why your numbers will never be perfect
Every analytics setup undercounts and GA4 is no exception. That is not a reason to ignore it, but it is a reason to stop treating the figures as an audited account. Knowing where the gap comes from stops you chasing ghosts.
The biggest single cause in the UK is consent. Analytics cookies need permission, so your cookie banner sits between every visitor and your data. Everybody who declines, ignores it, or leaves before answering is either missing entirely or only partially represented.
The rest comes from ad blockers, privacy-focused browsers, private windows and corporate networks that strip tracking scripts. None of it is fixable by you and none of it is evenly spread, so two businesses with identical traffic can report different numbers.
What consent mode changes
If your banner simply blocks the Analytics tag until someone accepts, everyone who declines is invisible. If it uses Google's consent mode, the tag still loads and sends signals without cookies when consent is refused, and Google fills some of the gap with modelling.
The second approach reports more completely. It is also more work to implement and has to be done honestly, because consent is a legal matter rather than a tracking preference. If you do not know which one you are on, ask whoever built the site.
How to work with incomplete data
Use GA4 for direction rather than magnitude. Traffic up forty per cent month on month is a real signal even if the number is undercounted, because the undercounting stays roughly consistent between months. That is what makes trends trustworthy while totals are not.
Cross-check against things you can count exactly. Your inbox knows how many enquiries arrived and your phone system knows how many calls came in. If GA4 shows twelve form submissions and your inbox holds nineteen, you know roughly how much it is undercounting by.
GA4 is more than most small businesses need
Worth saying plainly: GA4 was built for organisations with analysts, and it shows. Audiences, custom dimensions, funnel explorations, attribution modelling, BigQuery export. All powerful, all irrelevant to a business with a five-page site and a contact form.
You will not get more value from the advanced features. You will get it from three correct key events, tagged links, and the discipline to check four reports monthly. We have seen businesses spend weeks on dashboards nobody opens while the phone number on their site stays untracked.
Common mistakes
Almost every account we look at contains several of these, and each is a short fix rather than a project. None need a developer, and most take twenty minutes.
The pattern behind nearly all of them is the same. GA4 was installed, nobody configured it, and the reports have been quietly wrong ever since. Wrong numbers you trust are worse than no numbers, because they justify cutting the wrong things.
- Leaving data retention at two months. Change it to fourteen in Admin under Data settings, on day one.
- Never setting up key events. Traffic without enquiries measured tells you about activity, not results.
- Not filtering out your own visits. If you and your staff browse the site daily you are in the data, reading your own behaviour back to yourself.
- Tagging internal links with UTMs. It resets the session and destroys the original source.
- Inconsistent UTM naming. Four spellings of one source turn a clean report into a puzzle.
- Comparing GA4 to old Universal Analytics numbers. Different counting methods, so the comparison means nothing.
- Trusting tracking without testing it. Submit the form and tap the number yourself, then check Realtime.
- Checking it daily. Small site traffic is noisy. Weekly at most, monthly for most decisions.
- Mixing session and first user dimensions. Two reasonable numbers answering different questions.
Work down that list once and you will have fixed more than most agencies bother to check on a new account. It is dull, and it is the difference between a reporting tool and a decorative one.
A realistic plan for your first fortnight
You do not need to do all of this at once. Work in order, because each step makes the next more useful, and stop when you have enough to make decisions with.
Two steps below are timing-sensitive rather than difficult. Data retention and key events only apply to data collected after you switch them on, so doing them first is the difference between having a year of history next summer and having two months.
- Day one, five minutes. Admin, Data settings, Data retention. Change it from two months to fourteen.
- Day one, ten minutes. Check Enhanced Measurement is on for your web data stream, then load your own site and confirm you appear in Realtime.
- Day one, ten minutes. Set up internal traffic filtering so your own visits and your staff's are excluded.
- Day two, an hour. Set up your key events: form submission, phone tap, quote or booking request. Test each from a real phone.
- Day two, five minutes. Mark those events as key events in Admin, Events.
- Day three, thirty minutes. Write down your UTM naming rules, then tag the links you use most: your Google Business Profile link, your email footer, your social profiles.
- Day three, ten minutes. Link GA4 to Google Search Console so search query data sits alongside behaviour data.
- Week two. Leave it alone. Two days of data tells you nothing and drawing conclusions from it is worse than not looking.
- End of week two, twenty minutes. Open the four reports. Write down one thing you learned and one thing you will change.
Then set a recurring reminder for the same twenty minute review once a month. The habit matters more than the depth, because the value comes from noticing a change early rather than from understanding every metric on the screen.
Frequently asked questions
Why do my GA4 numbers not match my old Analytics numbers?
What replaced bounce rate in GA4?
Does GA4 track phone number taps automatically?
What is the difference between a conversion and a key event?
Why is so much of my traffic showing as Direct?
Do I need UTM tags if I only do a bit of marketing?
How much data does my cookie banner cost me?
Is GA4 worth it for a small business at all?
Where this fits
GA4 is one of three free Google tools a small business website should have working together. Analytics tells you what people did on your site, Search Console tells you what they searched to find it, and the Business Profile covers what happened before they reached the site at all.
For the wider view of which tools matter and which you can ignore, our guide to the Google tools every small business needs covers the set and how they connect. Measurement works best when the site itself is built to produce enquiries, which is what our web design work aims at and what our SEO work feeds.
If you would rather someone set the key events up properly and hand you a one-page monthly summary instead of a dashboard, get in touch. We will tell you honestly whether it is worth doing for a business your size.
