I'm currently working as a visual designer for a small startup and because we are really small we don't have money to hire UX expert, so I'm trying to do UX as well.

We are measuring many kinds of data about customers with Google Analytics and with cca 1000 UIP/day and 2 years on the way, we have quite a huge data database.

My question is: Are there any general use cases or examples how to use these data to improve customers flow? Or at least which data I should look at? Which data could help me to find week points?

  • 2
    The problem with visitor analytics is that it's just numbers. It's quite difficult to decipher intent. Unless you are doing actual A/B testing, I'd always be wary of reading too much into the numbers.
    – DA01
    Commented Jul 28, 2014 at 18:39
  • If you talk to the marketing people, you'll find examples of how they set up goals and conversion funnels that mimic some of the user tasks and behaviour analysis that will be of interest and useful to look at.
    – Michael Lai
    Commented Jul 28, 2014 at 23:52
  • I have a real world case study on my blog colmcqux.wordpress.com/single-vs-multi-step-checkout
    – colmcq
    Commented Oct 22, 2015 at 16:15
  • 2
    You should approach UX the other way around. I.e: Don't start out with "Oh, I have a lot of metrics piling up, let's see if I can use them somehow..." Start out by identifying what you what to know, then pick some metrics that will give you the answers. Check my answer to a similar question here: ux.stackexchange.com/a/17029/95 (The question itself is quite interesting, though. The precondition, however, is that you know what you want to analyze.) Commented Oct 24, 2015 at 12:15

11 Answers 11


I would suggest the following approach:

  • Go talk with some real users, observe them using your product (i.e. users are confused about price of product);
  • Build some hypotheses about what might be improved (i.e. price is not available on the front page);
  • Then look for data that supports that hypothesis (i.e. users bounce once they reach the sub-page where the price is displayed).

If you start out by looking at the data, you might find stuff that the users were not able to tell you about (or maybe they where embarassed to tell you the truth), but it is also way harder to find a valid hypothesis as there are so many possible interpretations of the data and it would take a lot of effort to find the most important problems that way.

  • Yes! I've seen analytics data uses to prove opposing points. It's mostly just numbers and often becomes a game of theories rather than useful hard data.
    – DA01
    Commented Jul 28, 2014 at 22:46

Google analytics might not be enough to spot all possibilities to improve ux. but you use the following pages to get some insight.

in-page analytics - this will show you where users are clicking relatively. if you see they are clicking more on something you think might not be having a high priority you can switch placements with something that needs more priority


user scroll depth- if you can see users are scrolling far down and leaving the page, it might mean they are looking for something and they couldn't find it


Visitor-User Flow - This view gives an overview on the paths your visitors are taking and the bounce rate of each page (red curved line going down). you should focus pages with high bounce. maybe the users are not getting the information they are expecting, they cant spot the CTA.


besides google analytics you could use a heat map tool or even better, mouse flow tool to watch recording of your users mouse moves.

http://www.crazyegg.com/ http://mouseflow.com/

besides this i'd recommend to do cafe testing on a few people. no more than 5 will be able to spot major issues.


I agree that your starting point is where funkylaundry told you to start with. At the point you get into looking at you analytics data, below are most crucial data points you should look at (I did not list these in the order of importance):

  • Page Views (what are the most popular pages) - depending on your site and content you may only see a small subset of pages which deserves attention and other page view are below 1 - 5%
  • Entry pages (first page the user visits) - don’t assume the user will always land on the homepage
  • Exit pages - understand what is the last page people visited and why they left. was it because they found the information they were looking for or they didn’t and gave up
  • Average time people spend on a page - if you combine this statistic with the Exit pages, you will be able to say if the person has given up and left
  • Devices and screen resolution - more and more people use their mobile devices to find information they are looking for, are your pages responsive and over necessary information on any device?
  • Conversion rate - on pages such as Lead generation or checkout flows, etc.. - will give you visibility into how complex or simple your process is and how many people complete the flow and end up converting.. the longer the form is the lower the conversion rate will be
  • Page load times - slow performance can drastically lower your conversion rate
  • Search terms - Michael mentioned it above, understand what people are looking for and if they make common typos, which will allow you to create a list to correct results for that query
  • Error Pages - did you miss any 301 and 302 redirects and you are loosing some of your traffic that is trying to his a specific URL you no longer have (Michael, kudos on bringing us up)
  • Scrolling Statistics - how far down the page people actually scrolled before leaving. it will help you identify content people don’t care about or never got to.
  • Video play - how far into the video one got before abandoning the player
  • Carousel Statistics - how may slides did people see and which one of the items people click the most. Studies suggest that very small number of people click on the second or any other slide. The first slide has most interaction.
  • Customer Locale - study where most of your traffic comes from and optimize for those markets

This is by no means a definitive list, however these are most common items majority of the websites have today.


Look at your top 50 searching terms and their results if you have searching on your website. You might find out that people are looking for any specific information (site, product, content,...) which is important but it's hard to reach it. Or you might find out that some results are wrong. Then you can improve your this content to be easily reachable.

Look at your "Error pages" where users came from and also look at the bounce efect of your "Error pages". It's natural, that '404' sometime opens but you need to make sure that user will not leave and will have options where to go to.


A couple ideas:

Find the average screen resolution (1280x800? 1024x768?) of your visitors. Open your browser to this size and make sure that key elements are not hidden, look for CSS and other display quirks.

Look at the average page load time in Site Speed. If your pages are loading slowly there is a good chance customers are abandoning the flow.

In Site Content - look at the second and third most visited pages on your site and make sure that those pages are strong and that they are not dead ends (i.e. make sure there is a way to get into the customer flow from these pages). Look at the bounce rate for your homepage or customer conversion page. Try to decrease the bounce rate with better a better layout or call-to-action.

Look at your traffic sources, you may see that everyone comes to your site via links in email that you send out, might be time to make sure your email templates are clean and tightened up or you may find a source of traffic that you would want to reach out to for business development.

  • 2
    While your site should work for 'average' visitor browser sizes, it should also work for 'all' visitor browser sizes. Also note that screen size isn't the key number here--browser size is.
    – DA01
    Commented Jul 28, 2014 at 22:45
  • I think it goes without saying that a site should work for all browser sizes but the OP is asking for some optimizations he can he make based on data from Analytics. I called it "Screen Resolution" because that is how it is labeled by Analytics.
    – Max
    Commented Jul 28, 2014 at 23:45
  • But here's where the numbers can be misleading. Maybe the site works just fine for the average screen size and the reason other screen sizes aren't shown are because the site doesn't work on those other sizes. For instance, a site may not get any mobile users because the site simply doesn't work for mobile users. Focusing on desktop users in that situation would be counterproductive.
    – DA01
    Commented Jul 29, 2014 at 0:30
  • 2
    I think this is totally valid - kind of a 'survivor bias' (totally love this story: macgetit.com/4943/solving-problems-of-wwii-bombers). However, Google Analytics records visits the moment users hit your site. i.e. I can look at all my mobile-sized traffic and see that yes they do indeed have a huge bounce rate or a very low time on site and I can work from there.
    – Max
    Commented Jul 29, 2014 at 16:41
  • ah yes, a bounce rate would be a good stat.
    – DA01
    Commented Jul 29, 2014 at 16:45

Yes, you can use GA data to understand customer behavior.

Contrary to what some replies tell, I tend to go the different way round: I try to NEVER (important to me!) first look at the datasets and THEN ask the question "What could I read here?". This approach basically plays into the hand of the big data problem: If I do not have actual questions, then it does not matter how much data I gathered - I will always find different or no answers for the same question.

What I try to do instead is to ask "What do I want to understand about our user?" and THEN see, what data is available.

To do this properly, I usually divide my question into two parts. Let's assume I want to check behavior on the checkout, the following two questions are important from UX perspective:

  • Is the user interested in what she sees?
  • Does the page perform? If yes/no - why?

I try to clearly separate the two parts: User Engagement & User Performance. Why? We need that to identify levers.

Coming back to the checkout, the questions would be:

  1. "Is the user going to the checkout at all?"
  2. "Is the user going to the next step(s)?
  3. "Does the user finish the process?"

The result for question 1 answers the "Is the user interested in...?" question. If users WANT to buy, they start the process. Results for the second question answer the question "did the step before work out properly?" - so the same question suddenly answers a performance question - and question 3 also aims for performance.

This being said, we might learn interesting things. Some time ago, while optimizing a checkout, we found that people tend to register, go through 2 of 3 steps, then left. Wow - What an insight: We immediately went to the confirmation step and included trust elements, cleared up the overview of the items, put better focus on USPs and "final" CTA - and bam, there we go: CVR & satisfaction increase.

Summed up, we first checked "What is the user interested in?", then asked "Are we able to satisfy user's needs?". No, we could not - THEN we looked into data deeper, found that they spend looooong time on the confirmation page, THEN found that we had to optimize this very step. I think you get the point: Looking at CVR of checkout never would have gotten us there, neither would have page visits on step 3. The order of questions is important.

So my shot would be:

  • Define micro conversions (Add to cart, Checkout start, Newsletter signup, ...)
  • Check user engagement on these Micro CVRs
  • Check performance of Micro CVRs on different segments (is mobile different from desktop? Are especially new users exiting the checkout? Understand the situation of the user and why it is harder for HER than for another user)
  • Identify levers by looking at "Users were INTERESTED but did not PERFORM. Why?"
  • Go as deep into data as you need to pin point the problem. If you can not find it, go for research. Check best practice NOW - not before. Think first, then go ask for help. It will increase quality of YOUR product.
  • Fix it.
  • This seems like the right approach. In general we all do better following Abby Covert's advice: "'What' before 'How:" Know what answers you are seeking to avoid getting hypnotized by the analytics data.
    – Jim Ryan
    Commented Sep 18, 2017 at 13:50

The other answers provide good info. And I agree that analytics alone is no replacement for real UX research. But you're asking specifically how to use analytics data to improve UX.

I recently worked on an e-commerce site and consulted often with the analytics guy. I wanted to know the following things:

  • How many pages did the average user click through before giving up? We would try to keep the paths to the product pages shorter than that.
  • What percentage of people used the search feature, compared to those who clicked through the product catalog? We'd promote the more popular one more highly. (In our case it was 50/50.)
  • What were the most- and least-popular product categories? This could help us decide what to promote more.
  • What pages were never visited? We could decide to remove those pages or, if they're useful, make their presence known.

Those are the main ways we used analytics in our UX Design. I'm sure you can come up with more based on your own needs.


I would agree with those who say, "talk to users." Analytics data are great; I use them whenever I can, but bear in mind: they can only tell you "what," not "why". To Illustrate: I once worked on the redesign of a US website which got approximately 5% of its traffic from Ontario, Canada. Its Canadian visitors generally spent 30-60 seconds on the site, looked at 3-4 pages and left. After relaunch, all of those numbers tanked: we got fewer Canadian visitors, and those who did show up looked at one page and left immediately.

If you were just looking at the numbers, you might conclude the new site was performing worse than the old one. However, If you asked around to discover the story behind the Canadian visitor traffic, you would find as I did, that every last one of the Canadian visitors had been looking for an Ontario-based Child Protective Agency but finding a New York City private charity by mistake. The site redesign made clear that this wasn't what they were looking for, so they left.


Please don't downvote me for not giving the answer to your exact question :)

From what I read about your company, I am sure you will be interested by inspectlet, whish is a free tool which records and playback sessions, that you can review. Effectively it gives you an insight of how you users use your site.

The free version enables you to get 4 or 5 recorded session per days, over a period of time, it can give you some good information.

As an example, we had a site which had a high bounce rate, Inspectlet showed us that that was right, visitors where landing on a page via organic search, but they would exit after spending over a minute, scrolling and actually reading the content of the page then leaving. Which in fact was appropriate for an informative site.

Anyway, that's another good tool in your UX toolkit




  • Just a heads-up: Yandex.Metrika does the same, it's free and it comes from a large company that's been around since the '90s. The feature is called WebVisor. Commented Aug 11, 2015 at 7:10
  • Usertesting.com is similar. You provide a URL and a set of tasks and you get video of people doing a think-aloud usability test on your site. (I've been a tester, but have never used their services.) Commented Oct 22, 2015 at 18:32

The answer to this question is surely

use A/B testing.

You're lucky to have plenty of data coming in it seems. So A/B test your UX ideas and find the UX ideas that make the most money.

Many analytics packages have an A/B concept built-in so it's easy to get measurements.


Well as being from a visual design background my self, i learned that the analytics are just a tool to support or validate our decisions made with the design.

And with UI/UX its mostly about information and action

  • You need to give right information to the user with required hierarchy.
  • And provide clear action for the user to take on that information.

Analytics will help you know if what you have thought while designing and what the user experiences during usage is working the way you intend or not. If it does its great, if it does not you keep experimenting with it till you get what you want.

Also heatmap analytics are better with visual analytics as per my experience.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.