What to measure in website analytics and how to read the metrics
What to measure in website analytics and how to read the metrics
1. Why even look at website analytics
Website analytics are not needed just for pretty graphs. They answer very practical questions: who came, from where, what they did, where they left, and why they didn't complete the application. If you only look at traffic, you might be happy with the number of 10,000 visits and not notice that 9,700 of them came from a random query and left no trace on the site.
The site almost always operates in conjunction with a task: sales, leads, calls, subscriptions, applications, reading materials, returns to the catalog. When the site analytics is set up correctly, it helps to see where money is lost, where the content does not meet the request, and why one landing page generates 12 inquiries while another generates 0.
Traffic alone proves nothing. A person can visit a page, spend 40 seconds, and leave because they couldn't find the price, didn't see the button, or opened the page on a phone where the form shifted down. This is where site analytics starts to work as a diagnostic tool, rather than just a report for the sake of it.
2. What basic metrics should be tracked
You should start with five things: users, sessions, page views, traffic sources, and conversions. Immediately add bounce rate and engagement, because without them it's easy to confuse genuine interest with a short random visit.
Users — how many different people visited during the period.
Sessions — how many visits they made.
Page views — how many pages were opened.
Traffic sources — where they came from: search, advertising, social media, direct access.
Conversions — how many times the goal was achieved: application, purchase, call, subscription.
Bounces — visits without further interaction according to the counter's logic.
Engagement — how much the user actually interacted with the site.
One number without context means almost nothing. For example, 1,000 users and 1,000 sessions is one thing, but 1,000 users and 3,500 sessions is a different story: people are returning, comparing, reading to the end, clarifying.
Another practical example. If traffic sources show growth from search, but conversions are not increasing, it means that the website analytics indicates a problem not with volume, but with the quality of incoming visitors or the page itself.
3. How to read user behavior on the site
Behavior is easiest to read as a route. The user came to the landing page, opened 2 more sections, lingered on the service card, then left or clicked on the form — and in this sequence, there is already a lot of meaning.
Page depth shows how far a person has gone within the site, but it should be interpreted carefully. 4 pages per session can sometimes be better than 9 if the four pages have contact with the price, cases, and form, while nine views are just wandering through empty sections.
Time on page also requires caution. 6 minutes on an article is a good sign if the text is long and the user read to the end; 6 minutes on a catalog page may indicate confusion, especially if the purchase button is hidden below the first screen.
Exit points help identify places where users leave most often. If 35% of visitors leave from the cart, and even more from the delivery page, website analytics shows a specific area where the price, delivery times, and payment method need to be checked.
Sometimes it's useful to look at the path step by step, rather than the total number of views. One person might enter the blog, go to the service, and reach the form in 3 clicks; another might get to the cart but stop at delivery. Two identical traffic numbers, but the behavior is completely different.
4. What metrics are important for conversion
For a business, conversion is not an abstract term. It is 1 application, 1 call, 1 purchase, 1 download of the price list, 1 appointment for a consultation. Therefore, in website analytics, goals should be set not only for the final action but also for intermediate steps.
Micro-conversions are actions that indicate movement towards a purchase: viewing contacts, clicking on the phone, adding a product to the cart, opening a form, selecting a tariff, downloading a catalog. Without them, it's hard to understand where users are getting lost even before the main goal.
The cost of a lead and the effectiveness of landing pages are read together. If the ad brings 200 visits but the page generates 2 applications, then the problem may lie in the traffic, the offer, the form, or a mismatch between the promise and the content. Here, website analytics is more useful than any intuition.
The path to the goal is examined separately. If a user goes through 4 screens and 2 forms before making an application, it means they are facing unnecessary steps. In practice, reducing the path by one step sometimes has a greater effect than changing the headline.
A good habit is to compare the goal and micro-conversions in one table. This way, it becomes clear where there is a lot of interest but few applications, and where interest almost immediately turns into action.
5. How to assess the quality of traffic
The quality of traffic is evaluated by sources, channels, and campaigns. If the report shows many visits from ads but few conversions, it is necessary to look not only at the cost per click but also at what kind of people came, from what query, and which page they landed on.
Search traffic often looks 'clean', but there are traps here too. A query may be informational, while the page is commercial; a person wanted a comparison but received a request instead. As a result, there is a visit, but no useful action.
Non-targeted traffic usually shows several signs: very short sessions, high bounce rates, identical behavior patterns, strange sources, and spikes without a clear reason. If there are many such visits, the website analytics shows not audience growth, but noise.
You also need to be careful with campaigns. The same ad can yield different results on different devices, in different regions, and on different landing pages. Without these breakdowns, it's easy to praise a campaign that actually brings empty clicks.
If you doubt the honesty of the source, it is useful to first check if a suspicious click is disguised as a normal one. In such cases, it helps how to check a website for fraud, because junk traffic sometimes comes from completely unexpected places.
6. How not to misinterpret numbers
Numbers love context. A 20% increase in a week may look great, but if it's the pre-holiday period, comparing it to a regular week explains nothing. Seasonality, promotions, weekends, and advertising spikes often change the picture more than a redesign.
You should compare not only with last week but also with the same period from the previous month or year if the business is seasonal. For an online store, the same traffic in November and February can behave differently, and conclusions without comparison will be too bold.
Segments also change the meaning of metrics. Mobile users may read longer but purchase less; new visitors may leave quickly, while returning ones convert better. If you mix all groups into one report, website analytics will become flat.
Another trap is changes on the website. After changing a form, button, or card template, the numbers may 'drop' not because of advertising, but due to the interface. In that case, you need to look at the change date and compare periods before and after, rather than arguing with the report emotionally.
7. Which reports to check regularly
The useful minimum is 5 reports. Site summary, entry and exit pages, conversions, traffic sources, devices, and errors. If time is short, this set is already enough to see half of the problems.
Site summary — a daily or weekly picture.
Entry pages — which pages users encounter first.
Exit pages — where visits end.
Conversions - what happens with goals.
Sources - where the audience comes from.
Devices - how mobile, desktop, and tablets behave.
Errors - forms, 404s, loading issues.
Ideally, these reports should be reviewed on a schedule, not once every six months before a meeting. It's convenient to check the summary on Monday morning, conversions in the middle of the week, and landing and exit pages once a month.
If the report shows a spike on mobile, you need to open the site on your phone and manually go through the path to the goal. Yes, it's an old method. And it often finds a problem in 2 minutes that isn't immediately visible in the numbers.
For a light pause between analyzing numbers, sometimes it helps to completely step away from analytics. For example, jokes about students. Jokes for free. Short they are read faster than a long report, and after such a pause, it's easier to return to the numbers without overheating.
8. How to turn analytics into action
The purpose of website analytics is not to accumulate reports. The purpose is to solve problems: where users are lost, which page doesn't convince, which channel brings empty traffic, which form is too long.
The working scheme is simple: identify the problem, formulate a hypothesis, change one thing, and look at the numbers again. If the number of submissions increased after reducing the form from 5 fields to 3, then the hypothesis worked; if not, check the text, trust, price, or the placement of the block.
It's good when website analytics are tied to a specific task for the week. For example, on Monday they look for the page with the highest exit rate, on Tuesday they check the path to the application, on Wednesday they compare traffic sources, and on Thursday they see if mobile devices have dropped off.
Sometimes it's useful to look at the website as a series of small tests. One headline, one button, one form, one screen. This makes it easier to understand which change had an effect and which just coincided with an increase in traffic.
When the numbers start to suggest the next step, website analytics stops being a report and becomes a working tool. And then there's no need to guess what to measure in website analytics and how to read the metrics — the answer is visible in 3-4 points of the report and in one specific action on the page.
