Google Analytics is a set of reports that count visits to a website, and nearly all of the confusion about it comes from five or six words that sound like they mean one thing and are defined to mean another. I have used it, and I did not need a statistics background to read it. I needed to know what each number was counting. That is a smaller job than it looks, and Google’s own help pages do most of it.

Ultimate Google Analytics Tutorial 2025 (Small Business Essentials), Stewart Gauld on YouTube

What the code collects, and what it does not

Analytics works by a small piece of JavaScript on each page of your site. Google’s help page on how Analytics works describes it in three stages: the code collects pseudonymous information about how each user interacts with a page, along with details such as browser language, browser type, device, operating system and traffic source; Analytics then aggregates and organizes that data by criteria such as device type or browser; and the processed data is stored in a database where it cannot be changed.

That last point has a consequence for a small owner. Settings you choose affect data going forward. A filter that excludes your own visits, for instance, helps only for the visits that arrive after you set it. Your own clicks from the counter, checking whether the site loads, are in the numbers until you tell the system otherwise, and a site with a few dozen visits a day can have a noticeable share of its traffic be you.

The traffic source detail is worth holding on to, because it answers the first question a business owner asks, which is where people came from. The categories include search engines, ads and email campaigns, which line up with the places a small business actually spends effort.

Users, new users and sessions

A dining room seen from the host stand with several tables occupied.

The first report most people see is the acquisition overview. Google’s help page for that report calls it a pre-made overview report that summarizes acquisition data, and says it shows whether marketing is bringing in new visitors and whether efforts to bring people back are working. By default its cards cover total and new users, real-time users over the last 30 minutes, new users and sessions grouped by channel, sessions from Google Ads campaigns, and average revenue from new users over their first 120 days.

Two definitions on that page carry most of the weight. The page says users shows the total number of users who visited your website or app, and that new users shows the number of users who visited for the first time. A user here is whatever the system counts as a visitor. It is not the same thing as a customer, and the number should be read as an approximate head count, the way you would eyeball a dining room from the host stand.

A session is the other unit, and the channel cards on that page are built from it: the page describes a card of new sessions grouped by channel, with the groups that produced the most listed first. A session is a visit and a user is the one who made it, so the two counts are different questions. If sessions run well above users, some of those users came back more than once, and if the two numbers sit close together, most visitors came once.

A card worth looking at is new users by channel. It is the cheapest answer to the question of which of your efforts brought a stranger to the site. The report does not tell you that the stranger bought anything.

Engagement rate and bounce rate

Older versions of Analytics had a bounce rate that confused a generation of site owners. The current version defines it differently, and the new definition is a good deal more useful. Google’s page on engagement rate and bounce rate defines an engaged session as one that meets any of three criteria: it lasts longer than 10 seconds, it has a key event, or it has 2 or more screen or page views.

From that, the page defines the other two numbers. It states that the engagement rate is the percentage of engaged sessions on your website or mobile app. And it says the bounce rate is the percentage of sessions that were not engaged, the opposite of the engagement rate. The two add up to a hundred percent, so you only need one of them.

Read that definition against the way a small site gets used. A person who lands on your hours page, finds the closing time and leaves in eight seconds is counted as a bounce. That visit did exactly what the page was built for. A high bounce rate on a page whose whole job is a phone number or an address is not a failure. A high bounce rate on a page meant to start a purchase deserves a closer look.

The honest limit is that the number tells you what happened and nothing about why. A ten-second threshold is a rule somebody picked. It will mislabel some good visits and some bad ones, and the way to use it is to compare one page against another page on your own site, not against a number you read somewhere.

The page also notes one exclusion: events such as first_visit, first_open and session_start are left out of the engaged-session calculation even when they are marked as key events. A key event is an interaction you have told Analytics matters to your business, such as a purchase or a form submission, and it is the third path to an engaged session.

Events, and what a metric is

An open hand-ruled ledger with a steel ruler and pencil on a workbench.

Everything above is built on events. Google’s glossary of dimensions and metrics defines a metric as a quantitative measurement, such as an average, ratio or percentage, and it says event count shows the total number of times an event is triggered. A page view is an event. So is a click you have set up to be tracked. The reports are different ways of grouping and counting the same stream of events.

This is the idea that makes the interface stop feeling like a wall of unrelated charts. A dimension is a label, such as the channel, the device or the page. A metric is a number, such as users, sessions or events. A report is a table of metrics split by a dimension. If you can say, for any report, which dimension is on the rows and which metrics are on the columns, you have read it.

I mean this literally. The next time a report looks strange, say it aloud: rows are channels, columns are users and sessions. The strangeness usually goes away within a sentence.

Channels, and a caution about where visitors came from

The traffic sources in Analytics are split by scope. Google’s page on scopes of traffic-source dimensions says the information on session-scoped dimensions appears in the traffic acquisition report, and the information on user-scoped dimensions appears in the user acquisition report. The practical meaning is that the same visitors can be sorted two ways. One report asks where each visit came from. The other asks where each person was first found.

A neighborhood restaurant understands this without any vocabulary. The regular who first came because a friend mentioned the place, and now walks in every Friday, was acquired once and has been a returning visit ever since. A user report credits the friend. A session report credits whatever got him through the door last week.

The two answers can disagree, and neither one is wrong. If the question is how a new stranger finds you, look at user acquisition. If the question is what is bringing traffic this month, look at traffic acquisition.

A weekly routine that fits on one page

A printed bar chart on a table beside reading glasses and a mug.

A report nobody opens is worth nothing, so the habit matters more than the tool. Here is a routine short enough to survive a busy week:

  • Open the acquisition overview and note total users, new users and the top channel by new users.
  • Open the engagement numbers and note the engagement rate for the whole site.
  • Pick one page you care about, such as the page where people order or the page that has your hours, and see how it did against the whole site.
  • Write the four numbers in a notebook or spreadsheet with the date.

That is all. The point of writing numbers down is that a single week tells you almost nothing, and ten weeks tell you something. A figure without a previous figure beside it is a trivia answer.

I taught a course on information systems at Queensborough Community College in 2008 and 2009, with a direction I chose, new media in small business. One thing a classroom teaches is how often people stop reading a tool because its vocabulary is hostile. Analytics has that problem, and the cure is the unglamorous one of learning five definitions.

What the numbers will not tell you

The reports do not know who your customer is. They do not know whether a visit led to a phone call, a walk-in or a sale, unless you set up key events and keep them honest. They do not know about the orders that arrive in person after a visit to the website, and for a small business with a counter, those can matter more than anything the screen shows.

I will also say this against my own point. A cleaner reading of the numbers can make a small owner feel in control without improving anything. Knowing the engagement rate does not sell a briefcase or a plate of eggs. The tool is useful when it changes a decision, such as which page to rewrite or which channel to leave alone, and it is wasted when it only produces a pleasant feeling of having checked.

For informational purposes only. This post summarizes Google’s help pages as they were read in preparing it; interface labels and definitions change over time, so confirm them against the current Analytics Help.

Sources