Back to blog
Website Analytics Data: What to Collect and How to Use It

Website Analytics Data: What to Collect and How to Use It

Marketer reviews website analytics at a monitor

Website analytics data is the record of what visitors do on your site: where they come from, which pages they view, what they click, and whether they convert. You collect it with a tracking script or tag, read it in reports or raw exports, and use it to decide what to fix, test and keep. This guide covers what the data contains, how to read it without fooling yourself, what privacy rules apply, and where it stops being enough.

TL;DR

  • Website analytics data is made of events (page views, clicks, purchases) plus the dimensions that describe them (source, device, page). Metrics are counts and rates built from those events.
  • Your numbers are only as good as your implementation. Ecommerce revenue, for example, is accurate only when item data is correctly sent from your site code.
  • Interface reports and raw exports will not match one to one, because reports include modeling. Pick one source per decision and stay consistent.
  • Analytics cookies generally need consent under UK rules, and privacy has to be designed in from the start, not added at the end.
  • Analytics tells you what happened. A/B testing tells you what to change and whether the change worked.

Table of Contents

What website analytics data actually is

Web analytics is the measurement, collection, analysis and reporting of web data to understand and optimize how a site is used. It is more than a traffic counter: it also feeds market research and business decisions.

The data comes from a few places. The same reference lists HTTP request headers, metadata generated by the server, and application-level data sent along with requests. In practice, for a marketer, that means three layers:

  1. Automatic data. Page URL, referrer, browser, device type, approximate location, timestamp.
  2. Behavior data. Scrolls, clicks, form submissions, video plays, time engaged.
  3. Business data. Purchases, sign-ups, trial starts, revenue, plan type. This layer is the one you have to define yourself.

The first layer is free. The third is where most teams lose accuracy, and it is also the layer that pays the bills.

On-site and off-site analytics

On-site analytics tracks visitor behavior on your own pages. Off-site analytics looks at your wider internet presence: market share, visibility, buzz. Most of this article is about on-site data, because it is the data you control and can act on.

The building blocks: events, dimensions and metrics

Modern platforms such as GA4 organize collection around events and their parameters. An event is something that happened (a page_view, a purchase). Parameters describe it (which item, what value, which currency). Dimensions are attributes you slice by, such as channel or landing page. Metrics are the numbers you count.

How events, parameters, dimensions and metrics relate

Google's Events report documentation describes the basic view: how many times each event fires and how many users trigger it. Start there when you audit a setup. If an event you care about is missing or fires twice as often as it should, every report built on it is wrong.

When the automatic data does not cover what you need, you define your own. Custom dimensions and metrics let you analyze data you collected yourself from your site or app, beyond what is gathered by default. Think plan tier, logged-in status, product category, or the name of the experiment variant a visitor saw.

The metrics most teams should start with

MetricWhat it tells youWatch out for
Sessions and usersVolume of visits and of distinct visitorsUsers are counted by identifiers, so cross-device visitors can be counted twice
Engagement rateShare of sessions that were engagedIts mirror image is bounce rate
Conversion rateShare of visitors completing a goalOnly meaningful if the goal event is defined cleanly
Revenue per visitorMoney earned per visitDepends on correct ecommerce implementation
Traffic by channelWhich sources bring visitorsUntagged links fall into catch-all buckets

Google defines engagement rate as the percentage of engaged sessions, and bounce rate as the percentage that were not engaged. Bounce rate is not the old single-page-visit number many people still assume. If you compare against a benchmark from an older tool, check the definition first.

For a longer menu of numbers, see our walkthrough of common marketing KPIs, and for the ones that actually move revenue, digital marketing metrics that predict conversion.

Getting the data: collection in four steps

If you are starting from nothing, the path is short.

  1. Create an account and property. In GA4 that means an account, a property, and a web data stream, as laid out in Google's developer guide for websites.
  2. Install the tag. Add the measurement snippet to every page, directly or through a tag manager. Site builders such as WordPress, Shopify and Webflow usually have a field for the ID.
  3. Wait, then verify. Google's own help notes that data can take up to 24 hours to show up in some contexts, so do not panic after ten minutes. Our checklist of five website tests to run before reports populate helps you confirm that tags fire correctly in the meantime.
  4. Define your conversions. Mark the events that matter to the business: purchase, lead form, trial start. Everything else is context.

The site-builder part matters for small teams. If you run a store or a marketing site without a developer, the cleanest setup is the one the platform supports natively, with a short list of custom events on top.

Ecommerce data: where accuracy breaks

Revenue reporting is the most common place where analytics data goes quietly wrong. In GA4, standard ecommerce reports depend on specific event names such as purchase, add_to_cart and begin_checkout. Google's dimensions and metrics reference documents how total purchase revenue is built: purchase revenue plus in-app purchase revenue plus subscription revenue, minus refunds.

That formula is only as good as the item data behind it. If the item array is missing fields, or the purchase event fires on a page reload, your revenue is off and nothing in the interface will warn you.

Another trap is Event Value. Google explains that it mixes realized revenue with assigned importance values and is not currency-converted, so keep it out of financial dashboards and ROI math. Use purchase revenue for money.

Quick audit for a store:

  • Place a test order and compare the order total in your shop admin with the revenue GA4 recorded.
  • Check that taxes and shipping are treated the same way in both.
  • Refresh the thank-you page and make sure the purchase does not count twice.
  • Confirm that refunds are sent, or your net revenue will drift upward.

For a fuller treatment, read our guide to ecommerce analytics and sales data.

Reports versus raw exports: why the numbers differ

Many teams hit this at the same moment: the interface says one number, the warehouse says another. This is expected. Google's developer team explains that the interface adds modeling, such as behavioral modeling for consent gaps, which the raw BigQuery export does not contain. The export is meant for custom, supplemental analysis on granular, unmodeled data, not to replicate standard reports.

The practical rules:

  • Do not try to reconcile the two to the last decimal. Expect a gap.
  • Choose the source that fits the question. Use the interface for quick answers and trends. Use raw data when you need event-level joins, custom attribution or your own definitions.
  • Be consistent. Compare this month to last month inside the same source, never across sources.
  • Document your definitions of a session, a conversion and a new customer so that the next analyst gets the same answer.

Understanding how analytics stores and displays data helps too: aggregated tables give fast, unsampled results, while other tables give flexibility. For a perspective on how raw data feeds ROI conversations, this piece on analytics and marketing ROI is a useful companion read.

If your team wants programmatic access without a warehouse, the Analytics Data API gives code-level access to report data. Note that it does not work with legacy Universal Analytics properties.

The four types of analysis you can run

A common question is what to do once the data is flowing. Analysts usually sort the work into four types: descriptive (what happened), diagnostic (why), predictive (what will happen) and prescriptive (what to do). Our article on the four types of data analytics goes deeper. For most small and mid-sized teams, descriptive and diagnostic analysis cover 90% of the value, and prescriptive is where experiments come in.

A simple analysis routine

Try this weekly rhythm:

  1. Check the headline: sessions, conversion rate, revenue, by channel.
  2. Find the biggest change versus the previous period.
  3. Segment it by device, landing page and new versus returning visitors until the change localizes.
  4. Write one hypothesis: "Mobile visitors from paid search drop off on the pricing page because the plan table is hard to scan."
  5. Turn the hypothesis into a test.

Step 5 is the one most teams skip.

Web, product and experience analytics: how they fit together

The boundaries between categories have blurred. Gartner's market definitions describe web, product and digital experience analytics as related layers: web analytics for acquisition and navigation paths, product analytics for detailed behavior inside a digital product, and experience analytics for diagnostics such as session replay and heatmaps that help explain why people behave as they do. Many organizations stack the three.

For a small team, this does not mean buying three tools on day one. Begin with web analytics, add heatmaps or recordings when a page underperforms and you cannot see why, and add product analytics when you have a logged-in product to improve. Business intelligence layers on top once several sources need to be combined; see our overview of BI analytics.

Privacy and consent: collect responsibly

This is the part that gets teams in trouble, and rules differ by country. The UK regulator's cookie guidance, the ICO guidance on cookies and similar technologies, treats analytics cookies as non-essential, so under UK GDPR and PECR they require consent even if they feel harmless. That applies to first-party cookies too. The ICO also stresses shared responsibility: those who build analytics systems and provide website services must design them so clients can comply, and privacy by design has to be built in from the start of a project.

This is UK law. If your visitors are in the EU, the US or elsewhere, the rules differ, and US state laws vary. Check requirements for each market you serve, and ask your lawyer or data protection officer when you are unsure.

What to do in practice:

  • Map every script and cookie on your site, including those added by plugins and ad pixels.
  • Load analytics tags only after the consent choice where the law requires it.
  • Collect only what you will use. Fewer fields means less risk and cleaner reports.
  • Expect consent gaps in your data and account for them when comparing periods.

Our deeper guides cover this: privacy in web analytics for marketers and, for Dutch-speaking readers, Privacy bij analytics.

A public example worth studying

If you want to see analytics data presented clearly, look at analytics.usa.gov, which publishes official web traffic data for thousands of US federal government websites through the Digital Analytics Program. It is a good model of what to show: a small number of metrics, plain labels, and a clear time window.

Our perspective: data is for deciding

We build an A/B testing tool, so we see what happens after the dashboard. The pattern is the same across small and mid-sized teams: they have plenty of analytics data and very few decisions made from it. Reports pile up, and the page stays as it was.

Three opinions we hold firmly:

  • Fewer metrics, tied to one decision each. If nobody can say what they would change when a number moves, stop reporting it.
  • Analytics shows where, testing shows why. A funnel can tell you that 60% of visitors leave the pricing page. It cannot tell you whether a shorter headline or a clearer plan table would keep them. Only a controlled test can.
  • Speed of the tracking script is part of the data. Heavy scripts slow pages, and slow pages change the behavior you are trying to measure. That is why we keep our script at 5.4KB.

If you already use GA4, you can send experiment data to it; see how to launch GA4 A/B tests with a 5.4KB script. Our wider practical web analytics guide shows how we think about the full workflow.

— Juan

From analytics to experiments with Stellar

Once your analytics shows a weak page, Stellar lets you test a fix without waiting for a developer. You edit the page in a no-code visual editor, set goals, and read results in real-time analytics. Stellar works with WordPress, Shopify, Webflow, Wix, Squarespace, Framer and Bubble.

Pricing is based on monthly tracked users. The Sandbox plan is $0 per month, Plus is $29 per month, Pro is $99 per month and Scale is $399 per month. Enterprise has no published price. You can compare the plans on the Stellar pricing page and start with the free tier if you track under 25,000 monthly users.

FAQ

What is website analytics data?

It is the information collected about how people reach and use your site: traffic sources, pages viewed, clicks, conversions and revenue. Analysts use it to understand behavior and improve results, as described in the overview of web analytics.

How do I get web analytics data?

Create a property in an analytics platform, add its tag to every page, and define the conversion events you care about. Data usually starts appearing within hours, though some reports can take up to a day. Test your setup with a few real actions before you trust it.

Why do my analytics reports and exported data not match?

Reports include modeling and estimates that raw exports do not. Google says the export is meant for custom, supplemental analysis, so expect a gap and compare within one source only.

Do I need cookie consent for website analytics?

Under UK rules, yes: the ICO treats analytics cookies as non-essential, so consent is required. Other countries and US states have different rules, so check each market you serve.

Sources

Recommended