Google Analytics terms, explained.

These are the Google Analytics (GA4) and Google Tag Manager terms you meet in your monthly report or when we ask for access. Each one is explained simply, with an example from a small business.

01. Setting up and linking

Account and property.

In Google Analytics (GA4), an account is the top-level folder, usually named after your business. Inside it sits one or more properties, and a property is the container that holds the data for one website or app.

It matters because access, settings and data all hang off the property. If a report looks empty, the first question is whether you are looking at the right property.

For example, a bakery with one website and a separate ordering app would normally have one account and two properties, so the website figures never mix with the app figures.

Data stream.

A data stream is the feed of visits from one source of data into a property, such as your website. Each website you want to measure gets its own web data stream.

If a stream is missing or pointed at the wrong address, nothing is recorded, however tidy the rest of the setup looks. We check the stream first when figures look wrong.

Measurement ID.

The measurement ID is the short code, starting with G-, that identifies one web data stream. It is how the code on your website knows which property to send visits to.

When we ask for it, it is not a password, and sharing it does not give anyone access to your data. It does need to match the one actually installed on your site, because a mismatch sends your visits to a property nobody is reading.

For example, if a previous developer installed a tag with one measurement ID and your reports are in a property with another, the reports will show no visitors at all even though the website is busy.

Google tag.

The Google tag is the single piece of code on your pages that sends data to Google Analytics, and to Google Ads if you use it. It is installed once, either directly in the website code or through Google Tag Manager.

It matters because the same tag can be installed twice by accident, for example by an old plugin and a new container. That counts each visit twice and inflates every figure.

Enhanced measurement.

Enhanced measurement is a set of switches in GA4 that record common actions on your website without any extra setup, such as scrolling down a page, outbound link clicks, file downloads, site search, video plays and form interactions.

It gives you useful extra detail for free, but it records what it can see rather than what matters to your business. A tap on your phone number is not in the list, and an automatic form event may not match what your business counts as an enquiry, so we check both.

For example, a PDF price list that is downloaded 85 times in a month will show up on its own, but taps on your phone number will not unless someone has set that up.

User access roles.

GA4 gives each person one of five roles. Viewer can look at reports. Analyst can also share explorations. Marketer can also manage audiences, events and key events. Editor controls the property settings. Administrator can do everything, including adding and removing people.

When we ask for access, we ask for the lowest role that lets us do the work, and you stay in control because only an Administrator can remove us. Always make sure at least two people in your business hold Administrator, so access is never locked behind one person who has left.

For example, to check your reports we need Viewer, to mark an enquiry form as a key event we need Marketer, and to change the property's settings we need Editor.

Universal Analytics.

Universal Analytics was the previous version of Google Analytics. Google stopped processing new data in it in 2023 and has since removed it, so GA4 is the only version in use.

If anyone quotes you a figure from the old version, it cannot be compared directly with GA4. The two count visits differently, so a drop when you moved across was not a loss of customers.

Google Ads link.

Linking Google Ads to GA4 lets the two systems share data. GA4 can then show what your ads cost and which campaigns brought visitors, and key events can be imported into Google Ads so it learns which clicks led to enquiries.

The link only works properly if auto-tagging is switched on in Google Ads, because that is what tells GA4 a visit came from an ad. Data from a newly made link can also take a day or two to settle.

For example, a plumber spending R12,000 a month on Google Ads can see in GA4 that the 'emergency callout' campaign brought 70 sessions and 6 enquiries, rather than only seeing clicks in the ads account.

Search Console link.

Search Console is Google's free record of how your pages appear in search results. Linking it to GA4 brings those search figures into GA4's reports, next to what visitors did once they arrived.

It matters because the two tell different halves of the story. Search Console shows impressions and clicks from Google search only, while GA4 shows every source and what people did on the site, so the two click and session counts will never match exactly.

02. Users and sessions

Active users.

Active users is GA4's main count of people. A user counts as active if they had an engaged session, or if GA4 recorded a first visit to your site, so someone who lands and leaves immediately may not count.

Because GA4 identifies people by the browser or device, not by name, the same person using a phone and a laptop counts as two. Treat the figure as a good guide to the number of people, not a headcount.

For example, a landscaper might see 1,200 active users in a month. If one customer researched on a phone at lunchtime and again on a laptop that evening, they count as two of those 1,200.

New and returning users.

A new user is someone GA4 has not seen on your site before, on that browser. A returning user has visited before.

It helps to read the two side by side. Many new users suggests your marketing is reaching fresh people, while a healthy number of returning users suggests people trust you enough to come back before they enquire. Clearing cookies makes a returning customer look new.

Sessions in GA4.

In GA4 a session is one visit: it starts when someone views a page and, by default, ends after 30 minutes of inactivity. If they come back after the gap, that counts as a new session, and the session timeout can be changed in the settings.

It matters because most figures in your reports, including engagement rate and channel counts, are calculated per session. One long browsing evening can be one session or three, depending on the gaps.

For example, a visitor reads your pricing page, answers the door for 40 minutes, then returns to fill in your enquiry form. GA4 records two sessions, and the form is credited to the second.

Engaged sessions.

An engaged session is a session that lasted longer than 10 seconds, or included a key event, or had at least two page views. A session that meets none of those is counted as not engaged.

It is a rough measure of whether a visit was real. It cannot tell you whether the visit was good, only that the person did something other than glance and leave.

Engagement rate and bounce rate.

Engagement rate is the share of sessions that were engaged. Bounce rate is simply the opposite: the share that were not. GA4 shows both, and they always add up to 100 per cent.

This is not the same bounce rate as in the old Universal Analytics, so do not compare the two. A high bounce rate is not always bad: someone who lands on your contact page, reads your phone number and rings you may leave quickly and still be your best visitor.

For example, if your website had 400 sessions in a month and 260 were engaged, the engagement rate is 65 per cent and the bounce rate is 35 per cent.

03. Events and key events

Events.

In GA4 everything is an event. A page view is an event, so is a scroll, a click on a link and a form submission. Each event has a name, such as page_view or generate_lead, and can carry extra details called parameters, such as the page address or a value in pounds.

This is why tracking can be configured in countless ways and why the same action can be recorded under different names by different people. It is also why events need to be checked, not assumed.

Recommended and custom events.

Some events are collected automatically, such as page_view. Google also publishes a list of recommended events with fixed names, such as generate_lead for an enquiry or purchase for a sale. Anything else you invent yourself is a custom event.

Using the recommended names where they exist means reports, integrations and benchmarks understand the event without help. Custom events are fine for anything unusual but are easy to duplicate or misspell.

For example, a quote form should send generate_lead rather than a custom name like 'form_done_2', so anyone reading the account later knows at once what it means.

Key events in GA4.

In GA4 a key event is any event you have marked as important to your business, such as a form submission, a call tap or a purchase. Google renamed these from 'conversions' to 'key events' in 2024, so the old term now belongs to Google Ads and the two are no longer the same thing.

Only real business outcomes should be marked. Marking scrolls or page views as key events makes the headline number look larger and tell you less.

For example, a dentist marks booking_confirmed as a key event. If they also mark 'scroll', GA4 might report 900 key events in a month when only 31 people actually booked.

Key events in Google Ads.

A GA4 key event can be imported into Google Ads, where it becomes a conversion that Google Ads can count and optimise for. The same enquiry is then counted in both places, but by different rules.

GA4 credits an event on the day it happens and Google Ads credits it to the day of the ad click, and they use different attribution rules. So the two systems will often disagree, usually by a small amount, and that is normal rather than a fault.

For example, GA4 might show 48 key events for the month while Google Ads shows 41 conversions, because some enquiries came from people who clicked an ad the month before.

DebugView.

DebugView is a GA4 screen that shows events arriving in real time from a device in test mode, which is how we check that a form, a call tap or a purchase is genuinely recorded.

It is the difference between 'the setup says it should work' and 'we saw it work'. We test in DebugView before calling any tracking finished.

04. Traffic sources

Channel groups.

A channel group is the way GA4 sorts every visit into a handful of labelled buckets, so you can see at a glance where your visitors came from. The standard buckets include Direct, Organic Search, Paid Search, Paid Social, Organic Social, Email, Referral, Display, Affiliates, Cross-network and AI Assistant.

The rules are set by Google and cannot be edited in the default grouping, although a custom grouping can be made. A visit that fits no rule falls into Unassigned. The buckets come from rules about the source and medium, which is why how links are tagged matters so much.

For example, a visitor who clicks your result in Google goes to Organic Search, one who clicks your Google ad goes to Paid Search, and one who follows a link in your newsletter goes to Email, provided the newsletter link is tagged properly.

Source and medium.

Source is where a visit came from, such as google, facebook or a newsletter's name. Medium is how it came, such as organic, cpc (paid clicks), email or referral. GA4 shows them as a pair, for example google / organic.

This pair is the raw evidence behind the channel groups. If a channel looks wrong, the source and medium pair almost always shows why.

For example, google / organic means a free Google search result, google / cpc means a Google ad, and mail.yourbusiness.com / referral could mean a webmail link that was never tagged.

UTM parameters.

UTM parameters are short labels added to the end of a link, such as utm_source=newsletter and utm_medium=email, so GA4 knows exactly where a click came from. They are written by whoever builds the link, not by Google.

Spelling and capital letters matter, because 'Email' and 'email' are counted as different things. Using one agreed set of labels is the simplest way to keep reports clean, and without them many visits land in Direct.

For example, a link in your June newsletter could end with ?utm_source=newsletter&utm_medium=email&utm_campaign=june_offer, and your report would then show which campaign brought the 22 sessions.

Auto-tagging and click IDs.

When auto-tagging is on in Google Ads, Google adds a click ID called gclid to the address of your landing page. GA4 reads it to know the visit came from an ad and to bring in campaign and cost details. Meta adds a similar one called fbclid.

Some redirects and forms drop that ID before the page loads, and the visit is then recorded as Direct or Unassigned. We check that these IDs survive the journey, because they are also what lets offline sales be tied back to the ad that started them.

Direct traffic.

Direct means GA4 has no record of where the visit came from. It includes people typing your address, using a bookmark, or clicking a link in an app that passes no information, and also any visit whose source was lost along the way.

It is the most misread number in the report. A rising Direct figure does not mean more people know your name. It often means tracking is losing the source of visits that really came from ads, emails or social posts.

For example, if 500 sessions in a month are split with 140 Direct, that is 28 per cent. If you also send a newsletter with untagged links, some of those 140 are really Email.

Unassigned.

Unassigned is the label GA4 gives a visit when its source and medium fit none of the channel rules. It is a fallback, not a source.

A large Unassigned share usually points to untidy link tagging, such as a medium named 'promo' that Google's rules do not recognise. It is fixed by tidying the labels, and it does not mean the visits were poor quality.

For example, a link tagged utm_medium=ppc lands in Paid Search, while utm_medium=paid_adverts may not match any rule and ends up in Unassigned.

(not set).

(not set) appears in a report when GA4 has no value for that column. It is a gap, not a page, source or customer, and it can show up under almost any heading.

A little is normal. A lot in one column, such as landing page, usually means a setup problem that someone needs to look at, so we never read much into a row labelled (not set).

Referral.

Referral traffic is visits that arrived by clicking a link on another website, such as a directory, a supplier's page or a news article. The other site's name appears as the source and the medium is referral.

It is a useful sign of which sites send you business. It can also include payment or booking pages that sent your own customers back to you, which then look like a new visit and may need to be excluded.

For example, a wedding venue may see 30 sessions a month from a local directory listing, which tells them the listing is worth keeping up to date.

AI Assistant traffic.

AI Assistant is a newer default channel in GA4 for visits that come from AI chat tools such as ChatGPT and Gemini. It is recognised from the link's source and is kept separate from Organic Search.

Google's AI Overviews and AI Mode in search are not counted in this channel, so they remain inside Organic Search. This means the channel is useful, but it will understate how much AI is influencing enquiries, because many people read an answer and never click.

05. Reports and explorations

Reports and explorations.

Reports are the ready-made pages in GA4, such as Acquisition and Engagement, which show the common figures without any building. Explorations are a separate area where you build your own table, funnel or path from the raw events.

Explorations are more flexible but are limited by data retention and sampling. For most business owners, the monthly report we send is built from both.

Date ranges and comparisons.

Every GA4 figure belongs to a date range, and the range at the top of the screen changes every number below it. You can also switch on a comparison, which puts a second range, such as the previous period or the same time last year, next to it.

If two people quote different numbers for 'last month', check the date range first. The most common reasons for a mismatch are different dates, different properties and different time zones.

For example, 310 sessions in September compared with 355 in August is a fall of 45 sessions, or about 12.7 per cent, but September also has one day fewer.

Segments.

A segment is a slice of your visitors that you choose, such as only people who came from Google Ads, or only people on a mobile. Applying one shows the same report for that group alone.

It is how you answer questions like whether mobile visitors enquire as often as desktop ones. A segment changes the numbers it shows, so always check which one is switched on before comparing two screens.

For example, a physiotherapist might see 6 per cent of desktop sessions end in a booking against 2 per cent on mobile, which suggests the mobile booking page needs attention.

Thresholding.

Thresholding is GA4 withholding data that could be used to identify an individual. Google says it can apply to reports that include demographic data or audiences defined by demographics, and to search query rows that do not have enough total users. When it applies, the data quality icon at the top of the report says so.

It matters most for small websites, where a row can easily have too few users. Some figures can look lower than the truth, or missing, because rows were hidden. A wider date range often brings them back.

For example, a website with 80 visitors a month filtered by age and city may show an empty table, because every row is too small to show without risk of identifying someone.

Sampling.

Sampling is when GA4 works from a portion of the data and scales it up, instead of counting everything. Google says it can happen when a report or exploration uses more events than the property's quota, which is 10 million events for a standard property, a level a small local business is very unlikely to reach.

A data quality icon on the report shows when sampling has been used and what percentage of the data the figures are based on. In that case the figures are an estimate and should be read as close rather than exact.

Data retention.

Data retention is how long GA4 keeps the detailed, person-level data that explorations rely on. A standard property can keep it for 2 months or 14 months, and Google's default is 2 months.

It only limits explorations and funnels, not the standard reports. Lengthening it applies to data that has not yet been deleted but cannot bring back anything already gone, and shortening it deletes the older data. It is one of the first things to check, because a 2 month limit quietly caps what any custom analysis can look back over.

For example, a business that wants to compare this spring with last spring in an exploration cannot do it on a 2 month setting, but could if it had been changed to 14 months a year earlier.

Data freshness and processing delay.

GA4 shows some figures within minutes in the Realtime report, but the standard reports can take up to a day or two to settle. Figures for the latest day can therefore change a little afterwards.

This is why we never quote a figure from a day that is still settling, and why Google Ads cost data in GA4 can take longer to appear after a new link is made.

06. Consent and data quality

Modelled data.

Modelled data is GA4's estimate of the visits it could not observe because people declined cookies. It uses what it learned from visitors who did consent to fill the gap in the totals.

It only works if consent mode is set up in its advanced form, and only once there is enough data. Google sets minimum volumes: at least 1,000 events a day from visitors who declined, for at least 7 days, plus at least 1,000 daily users who agreed on at least 7 of the previous 28 days. Meeting them does not guarantee modelling, so a small website may never receive it. The figures are then simply those who agreed.

For example, a site with 1,000 recorded users may really have had more, because some visitors declined. Modelling can estimate those, but a very small site would normally see only the 1,000 observed.

Cross domain measurement.

Cross domain measurement lets GA4 treat a visit that moves between two of your websites, such as your main site and a separate booking or payment page, as one journey. It works by adding a short code called _gl to the link.

Without it, the customer appears as a brand new visitor on the second site and the sale is credited to a referral from your own website. It is fixed by listing both domains in the data stream settings.

For example, a hotel's website sends guests to a booking engine on another domain. Without cross domain measurement, 100 bookings may all be credited to 'referral', hiding the Google search that actually brought those guests.

Internal traffic filters.

An internal traffic filter removes visits from your own team, such as staff browsing the site or us testing it. You tell GA4 which IP addresses are yours, create the filter, and switch it from testing to active.

Without it, your own visits are counted as customers, which inflates sessions and can look like interest that isn't there. The filter only works from the day it is switched on, and it cannot recover earlier data.

For example, if a team of five checks the website daily and creates 40 of 500 monthly sessions, that is 8 per cent of the traffic that was never a customer.

Unwanted referrals.

An unwanted referral is a website that GA4 treats as the source of a visit but which really just sent your own customer back, such as a payment provider after checkout. GA4 lets you list these so they are not counted as a new source.

If they are not listed, a sale can be credited to the payment page rather than to the ad, search or email that really brought the customer, which hides your best sources.

Reporting identity.

Reporting identity is the setting that decides how GA4 recognises the same person across devices and sessions. The Blended option uses signed-in IDs where it has them, then device IDs, and then modelling.

Changing it can change your user counts, and it also affects whether thresholding applies. It rarely needs changing, but it is a good first question when two reports show different user numbers.

07. Google Tag Manager basics

Container.

A Google Tag Manager container is the single bundle of code that sits on your website and holds all your tracking inside it. Once the container is installed, tags can be added or changed in Tag Manager without anyone editing the website itself.

Each container has an ID starting with GTM-. When we ask for access, it is to a specific container, and it matters that your site carries only one, because two can count every action twice.

Tag.

A tag is one job inside the container, such as sending a page view to GA4, recording an enquiry for Google Ads or firing the Meta pixel. Each tag does one thing when it is told to.

If an enquiry stops being recorded, the cause is almost always a tag that is paused, set up wrongly, or no longer being triggered by anything.

Trigger.

A trigger is the rule that tells a tag when to fire, for example 'when someone clicks a link starting with tel:' or 'when the form is submitted'. A tag without a trigger does nothing.

Triggers are the most common point of failure after a website is rebuilt. They match on things such as button names, page addresses and form IDs, and the rebuild changes those, so a tag can sit untouched and never fire again.

For example, a trigger that waits for a button called 'Send' will stop working the day the web designer renames it 'Submit enquiry'.

Variable.

A variable is a piece of information that tags and triggers can use, such as the page address, the text of a clicked button or the value of an order. Some are built in and some are made to suit your site.

They let one tag be reused in many places. A measurement ID stored as a variable can be changed once instead of in every tag.

Data layer.

The data layer is a list of facts that your website passes to Tag Manager, such as 'this was a purchase for R2,400', 'this form was submitted' or the product that was viewed. Tag Manager reads them and hands them to GA4 and your ad platforms.

It is the most reliable way to track events, because the website states what happened rather than Tag Manager guessing from what is on screen. It needs a developer to set up, and it is usually worth it for shops and for forms.

For example, a shop's website can push a 'purchase' message with the order value and ID, so GA4 records the sale exactly once with the right amount.

Workspaces and versions.

A workspace is a private draft of changes in Tag Manager, and a version is a saved snapshot of the whole container. Changes in a workspace are not live until a version is made and published.

This is the safety net. Earlier versions are kept, so a mistake can be rolled back to the last good version without anyone rebuilding anything.

Publishing a container.

Publishing is the step that makes a saved version live on your website. Before that, the change can be tried in Preview mode, which shows on a private copy of the site which tags fire and why.

Publishing affects every visitor at once, which is why it is a deliberate step, and why it should be done only after a test, by someone who understands what has changed. At red&white we always prepare and test changes first. Who presses publish depends on you: some clients give us Publish access and we publish once the test has passed, others prefer to publish themselves.

For example, a new tag for enquiry form tracking can be tested in Preview with a dummy submission, and only published once it has been seen firing.

Tag Manager access levels.

Tag Manager access works in four steps, plus no access. Read can only look. Edit can make changes in a workspace. Approve can also create versions. Publish can also make a version live and has full control of the container.

The access we ask for depends on how you like to work: Publish if you are happy for us to put tested changes live, or Edit or Approve if you would rather publish them yourself. Either way, make sure at least two trusted people in your business hold the highest role.

For example, an agency with Edit access can build a new enquiry tag, but it cannot go live until someone with Publish access presses the button.