Website Analytics: What to Actually Measure

Website analytics is the measurement of how people find and use your website, and in 2026 that means Google Analytics 4 for most UK and Irish businesses. A business owner needs five numbers each week: total sessions, sessions by channel, session key event rate, top landing pages by sessions, and engagement rate on those landing pages. Four of those five live in two GA4 reports, Traffic acquisition and Landing page, so a weekly review takes about ten minutes. GA4 calls a completed enquiry a key event, not a conversion, and it will not record one until somebody configures it. Traffic without enquiries is a diagnosis, not a success.

Most business owners have website analytics installed and almost never open it. The dashboard loads, the numbers mean nothing in particular, and the tab gets closed again. This guide fixes that by naming the five numbers worth checking every week, the ones safe to ignore, and how to get Google Analytics 4 reporting properly without hiring anyone.

Which Five Website Analytics Numbers Should You Check Every Week?

Five numbers, two reports, roughly ten minutes. That is the whole weekly routine, and it covers volume, source, and outcome without requiring you to understand a single thing about event parameters.

The five sit in Google Analytics 4 under Reports in the left menu. Two of them come from Acquisition > Traffic acquisition. Two more come from Engagement > Landing page. The fifth, key events, has its own report under Engagement. If your left menu shows headings like “Generate leads” and “Examine user behaviour” instead of Acquisition and Engagement, your property is using the Business objectives collection rather than the Life cycle one. Both contain the same reports under different names, and you can switch by going to Reports, then Library, and publishing the collection you prefer.

What follows assumes you already know roughly what you want the website to do. If that part is genuinely unsettled, measurement will not resolve it, and a session on digital marketing strategy for your business is the more useful starting point, because a metric only becomes meaningful once there is a decision waiting on it.

That last one has a catch worth knowing about, because it defeats a lot of people. Engagement rate is not shown in most GA4 reports by default. Google’s own documentation on engagement rate and bounce rate sets out the fix: open the Landing page report, click Customise report in the top right, choose Metrics, click Add metric, type “Engagement rate”, and save. You need Editor or Administrator access on the property, and if you cannot see the Customise report button, that is why.

Engagement rate itself is worth defining precisely, because the plain English reading of it is wrong. GA4 counts a session as engaged if it lasts longer than ten seconds, or includes a key event, or involves two or more page views. Any one of the three is enough. So a visitor who reads one page for twelve seconds and leaves counts as engaged. Engagement rate is not a measure of enthusiasm; it is a measure of whether the visit cleared a low bar. Bounce rate in GA4 is simply the inverse, and it is not comparable to the bounce rate in Universal Analytics, which counted single-page visits instead.

Why these five and not others

Each of the five answers a different question, and together they cover the chain from arrival to outcome: how many came, where from, what they did, which door they came through, and whether that door suited them. Drop any one and a blind spot opens. Add a sixth and the routine stops being weekly, because it starts feeling like work.

Set a recurring fifteen-minute slot, look at the five, and write one line about anything that moved. The writing is the part that matters. A number checked and forgotten has no more value than a number never checked.

Which Website Analytics Metrics Should You Ignore?

Ignore any number that goes up when nothing good has happened. That is the whole test, and it eliminates most of what sits on the default GA4 home screen.

These metrics are not wrong, and several of them have proper diagnostic uses further down the line. The problem is that they read as progress when they are not, and business owners reasonably act on what looks like progress.

One habit protects you from most of this. Before reading any number, decide what you would do differently if it went up by a fifth, and what you would do if it fell by a fifth. If both answers are “nothing”, the number is decoration.

A note on timing, since it causes a lot of unnecessary alarm. GA4 does not report in real time outside the Realtime report. Standard processing commonly takes 24 to 48 hours, so yesterday’s figures are usually incomplete and today’s are close to meaningless. Check last week, not last night.

How Do You Set Up GA4 Without a Developer?

You can get a working GA4 setup yourself in under an hour on most sites, and the parts that genuinely need a professional are narrow and identifiable. The honest split matters here, because plenty of businesses either pay for work they could do in twenty minutes or attempt work that quietly breaks their data.

The part you can do yourself

Start at analytics.google.com and create an account and a property for your business. GA4 will then ask you to create a data stream: choose Web, enter your site address, and give the stream a name. It produces a Measurement ID beginning with a G. That ID is what your website needs.

Getting the ID onto the site is where people expect to need help, and usually do not. On WordPress, the simplest route is Google’s own free plugin: install Site Kit by Google from Plugins, Add New, then follow its setup flow, which connects Search Console and Analytics without touching any code. One thing the plugin will not do is create the Analytics account for you, so make the property first. On Wix, Squarespace, Shopify, and most other hosted builders, there is a dedicated field in the settings where the Measurement ID goes; search their help centre for “Google Analytics” and paste it in.

Then three configuration steps that take about ten minutes between them and materially improve the data:

• Leave enhanced measurement switched on. It is on by default for new properties and captures scrolls, outbound clicks, site searches, file downloads, and video plays with no further work.
• Exclude your own traffic. Under Admin, Data streams, your web stream, Configure tag settings, Define internal traffic, add your office IP address. Staff visits otherwise inflate exactly the pages you check most.
• Link Google Search Console. Under Admin, Product links, Search Console links. This is what lets you see the search queries behind your organic sessions rather than guessing at them.

Finally, mark your enquiry as a key event, because until you do, GA4 records no outcomes at all. If your form redirects to a thank-you page, go to Admin, Events, Create event, and build one that fires on that page path. If the event already exists in the list, you simply click the star beside it to mark it as a key event. Allow 24 to 48 hours before expecting it in reports.

The part that needs a professional

Four jobs are worth handing over, and they are the same four on almost every site.

The first is tracking anything that does not produce a new page: a click-to-call link on mobile, a chat widget, a PDF download behind a gate, an embedded booking tool. These need Google Tag Manager, and Tag Manager is where self-taught setups most often go wrong in ways nobody notices for months.

The second is consent. Under UK GDPR and PECR, analytics cookies require active, informed consent before they are placed, and the Information Commissioner’s Office guidance on cookies is explicit that a banner with a pre-ticked box, or one offering only “accept”, does not meet the standard. Doing this properly means a consent management platform wired to Google’s Consent Mode, and doing it improperly means either a legal exposure or a data set with a hole in it, depending on which way you got it wrong.

The third is e-commerce. Purchase and basket events carry values and item detail, and a partial implementation produces revenue figures that look plausible and are wrong.

The fourth is anything spanning two domains, such as a booking system on a separate address. Without cross-domain configuration, GA4 treats the handover as a new visit from a referral, and your enquiries appear to come from your own website.

A word on the cookieless alternatives

GA4 is not the only option, and for some businesses it is not the best one. Privacy-first platforms such as Plausible, Fathom, and Matomo in cookieless mode collect aggregated traffic data without setting cookies or processing personal data, which under most readings of PECR removes the consent banner requirement entirely. The trade-off is depth: no individual journey tracking, limited segmentation, and no Google Ads integration. If you run Google Ads, GA4 is close to unavoidable. If you do not, and your questions are mainly about channels and enquiries, a simpler tool answers them with less setup and fewer compliance headaches.

Pricing on these platforms changes often enough that it is worth checking the vendor’s own page rather than any article, including this one.

Where Is Your Website Traffic Actually Coming From?

Open Traffic acquisition and read the channel column before anything else, because volume without source is uninterpretable. The same 500 sessions mean very different things depending on where they came from.

GA4 sorts visits into default channel groups. The ones most UK SMEs see are worth knowing by sight.

Organic Search is someone who found you through Google or Bing without clicking an advert. It is usually the channel with the highest intent and the slowest to build. Growth here is the clearest evidence that content and search work is paying off, particularly when the session key event rate holds steady as the volume rises.

Direct is the honest answer to a question GA4 could not resolve. Some of it is people typing your address or using a bookmark, which is genuine brand strength. Much of it is untagged email links, clicks from documents, links opened in apps, and privacy tools stripping the referrer. Treat a large Direct figure as a measurement gap first and a compliment second.

Referral is a click from another website. Small referral numbers from a directory or a supplier’s site are normal. A sudden spike from a domain you have never heard of is usually spam.

Organic Social is unpaid social posts. This is where a lot of businesses discover an uncomfortable gap between effort and result, since social frequently delivers respectable session counts alongside a session key event rate close to zero.

Paid Search and Paid Social should be checked against what the advertising platform itself reports. The two will never match exactly, because they attribute differently, and expecting them to match wastes a great deal of time.

Email only appears if your links carry campaign tags. If your newsletter clicks are landing in Direct, that is a tagging problem, not an email problem.

Unassigned means GA4 received data it could not place, often because consent was declined or a tag fired oddly. A small share is normal. A large one is a setup issue.

The useful comparison is not which channel sends the most sessions. It is which channel sends the most key events per hundred sessions. A channel delivering 5,000 visits at a 0.2% key event rate is doing less for the business than one delivering 400 at 4%, and the second is where additional budget belongs.

How Do You Connect Website Analytics to Actual Enquiries?

You connect them by deciding what counts as an enquiry, marking it as a key event in GA4, and then checking the number against what actually arrived in your inbox. That last step is the one almost everybody skips, and it is the one that makes the data trustworthy.

Key events replaced the term “conversions” inside GA4 in March 2024. The measurement logic did not change, only the label, so older guides describing GA4 conversions are describing the same thing. The word “conversion” still appears in GA4, but now refers specifically to a key event that has been imported into Google Ads, which is why the two columns will never agree.

What to mark, and how to count it

For a service business, three or four key events is plenty. The contact form submission is the obvious one. A click on the phone number matters more than most owners expect, particularly on mobile, though it needs Tag Manager. A quote request or booking, if you have one, belongs here too. Everything else is a micro conversion: useful for diagnosis, not for reporting.

GA4 lets you choose how a key event is counted, either once per event or once per session. For enquiry forms, once per session is almost always the sensible setting, because a visitor who submits twice by accident should not read as two leads.

The reconciliation that makes it real

Once a month, count the enquiries you actually received and compare that figure with the key event count for the same period. They will differ. What matters is whether the gap is stable and explainable.

Consistent undercounting in GA4 usually points to declined consent, which is expected and not a fault. Consistent overcounting points to test submissions, spam, or an event firing on page load rather than on submission. A gap that swings wildly month to month means something is broken. Doing this reconciliation twice is enough to tell you whether your numbers can be trusted, and it is worth more than any dashboard.

Where the picture is that the traffic is fine and the enquiries are not, the problem has moved out of measurement and into what the page asks of the visitor. That is a different discipline with its own methods, covered in ProfileTree’s guide to improving the conversion rate on pages you already have.

“The businesses that get the most from their website data are almost never the ones with the most sophisticated tools. They are the ones that decided in advance what question they were trying to answer, built the measurement to answer it, and made someone responsible for acting on what it shows,” says Ciaran Connolly, founder of ProfileTree.

When Do the Numbers Say You Need Help?

Three patterns are worth acting on, and each of them is visible from the weekly five without any further analysis.

Traffic rising, key events flat. Sessions climb across two or three months while the key event count sits still. This nearly always means the traffic arriving does not match what the pages offer, or the pages give visitors nothing clear to do. More content will make the chart look better and change nothing commercially.

One channel carrying everything. A single channel supplies most of your sessions and most of your key events. It is working, but the business is exposed to one algorithm, one platform decision, or one competitor outbidding you. The fix is not to reduce that channel; it is to build a second one before you need it.

Strong volume, weak engagement on the money page. Your main service page collects healthy sessions with an engagement rate persistently below 30%. Something is wrong between the search result and the page: the wrong queries, a slow mobile load, or a page that answers a different question from the one being asked.

Each of these points somewhere different, and it is rarely obvious which from the analytics alone. A structured review of the site across technical, content, experience and conversion layers is what separates the three, because the cause of a flat key event count sits in one of those four places and analytics only tells you that it exists. Where the question is broader than measurement, ProfileTree’s guide to running a full website analysis covers the evaluation side of the same work.

The rest of the time, the answer is simply to keep the fifteen-minute slot, write the one-line note, and let three months of it accumulate. Patterns you would never spot in a single session become obvious across twelve, and by then you have something better than a dashboard: a record of what changed and what you did about it.

FAQs

The questions below come up most often from business owners setting up or reviewing measurement for the first time.