The Mitsue-Links UX Blog shares some of our insights and opinions about UX in Japan, experience design and cultural differences between user research in Japan and the world.
If you want to find out more about us, please contact us at firstname.lastname@example.org
Culture Shock: Japan's Winter Holiday Season
It's the holiday season here in Japan, and after a long hiatus, we'd like to get the Mitsue-Links UX blog up and rolling again with a festive start. We've talked before about the importance of cultural differences in Japan and their effects on the behaviors of users. So, I thought it'd be great to look at the Christmas and New Year's holiday traditions in Japan, some of which are observed nowhere else in the world.
Growing up as a Christian in the United States, Christmas was perhaps one of the most important times of the year. Presents were purchased months in advance, lights were hung on every street, trees were precociously decorated, and seasonal tunes were played on the speakers of every store in town. And every year, on Christmas day, we'd tear into the presents that Santa left for us, go to church, and then retreat to a relative's house where the family had gathered for a massive feast of turkey, ham, and all the fixings. It was a time of celebration, giving, and family.
Dig Deeper into Your Site's Bounce Rate Using Google Analytics
Recently, I had the chance use Google Analytics for a couple UX projects here at Mitsue-Links. Google Analytics allows anyone to view a vast amount of data concerning site usage and user behavior. The bounce rate is one of the most basic pieces of data available. Many of you will probably be familiar with what a bounce rate is, but just in case, a bounce occurs when a user visits a site and leaves without clicking through to another page within the site. Therefore the bounce rate is the proportion of users that have bounced off your site.
On paper, a high bounce rate seems bad. It could indicate that the majority of users are leaving the site without any engagement whatsoever. It could mean that users found the site too difficult to use, or perhaps they found the contents of the site to be useless. A quick search will bring up plenty of blog posts detailing the negative effects of high bounce rates, and what you can do about it. However, most of these sites fail to mention that a high bounce rate isn't necessarily an indicator that the site is performing poorly.
Your Bounce Rate (Maybe) Isn't That Bad at All
Just because a user bounced does not mean that the user had no positive engagement with the site. For example, bounces can also occur if the user finds the exact piece of information they were looking for on the first page that they landed on. If the user's intention were to find the phone number they could call for support, and they landed on the contacts page where the phone number was immediately available, they would have no reason to click through to other pages within the site. Having completed their objective, the user will leave the site satisfied, and with no need to click through to any other page within the site. This kind of bounce could occur at a variety of page types, such as product pages or blog posts.
Here's the thing. The bounce rate (be it high or low), on its own, is neither good nor bad. It's just a number. The bounce rate needs to be considered along with more contextual information, such as the page the bounces are occurring at or how the users are arriving at that page, before a conclusion should be reached on whether it is a problem or not.
So, how exactly would you differentiate between good bounces and bad bounces?
Figuring Out Where and Why the Bounces Are Occurring
The first step should be to check where the bounces are occurring, and to find any potential reasons that the bounces may be happening there. The overall site wide bounce rate can be useful monitoring bounce rates over long periods of time, but its scope is too large to pinpoint where exactly the bounces are occurring.
The All Pages report in Google Analytics is a good place to start looking for where bounces are occurring. However, you shouldn't simply sort the list of pages on the site by bounce rate. You'll likely find that the pages that come to the top of the list will have very few views, and bounce rates of 100% or 0%, depending on the sort direction. Instead, use the weighted sort feature. Google Analytics' weighted sort feature allows you to sort the bounce rate (and other metrics) depending on how impactful it is for your site's performance. This list will give you a better overview of where the bounces are occurring (or not occurring) within your site.
Once you have a rough idea of the pages that might have extreme bounce rates, the next step is to figure out how the users are coming to that page. Are they searching for a keyword, or being referred to from a certain site? The context under which the user was visiting the page may indicate the intention they had upon landing on your site, and this can help you decide whether the bounces (or lack thereof) are beneficial or harmful. For example, suppose you have a fashion e-commerce site that has a product details page for a jacket with a very high bounce rate. If many users are coming to the page via search engine results to see what colors are available for the jacket and this information is easily accessible to the user, they may be bouncing off the site satisfied after finding the information they were looking for. On the other hand, if the bounces are occurring because the users were coming to the site looking for shirts, this could mean that there are SEO issues with the page leading to the bounces.
Digging Even Deeper into the Bounce Rate
One metric that becomes available using custom events is scroll depth, which allows you to see if your users actually scrolled through your page or not. However, be careful not to take a high scroll rate to mean that users are reading and interacting with your page. It may simply mean that the user was just quickly scanning through the page before deciding that they were not interested. This can be mitigated by measuring the user's time on page along with the scroll depth.
The amount of time the user spent on a certain page can be a good indicator of whether users were interacting with the contents or not. For example, a long duration may indicate that many users were reading your blog post, and a short duration may indicate that users only looked at the page for a short amount of time, and determined that the contents were not worth looking into. Whether spending a long time is preferable (i.e. your blog post is being read) or not (i.e. users were getting lost on your FAQ page) will depend on the contents and purpose of the page.
The average time on page metric is available within Google Analytics by default. However, the way Google Analytics measures time on page is not ideal. Google Analytics measures this metric by calculating the time difference between when the user lands on the page, and when they clicked through to a different page within the site. Since bounced users have not actually clicked through to a different page within the site, it is impossible for Google Analytics to calculate how long they spent on that page. For these bounced users, time on page is recorded as 0 seconds, even if they had actually spent several minutes on the page. Since this 0 second value is taken into account when calculating the average time on page for all users, this means that the actual amount of time users spent on the page can be much different from what is shown to be the average time on page on Google Analytics. For example, suppose both users A and B, from the image at the beginning both spent 10 minutes on the homepage. For user A, the 10 minutes spent on the page is recorded when he clicks through to the next page. For user B however, the time on page is never recorded because the user bounced and never clicked through to a second page within the site. Instead, the user's time on page is considered to be 0 seconds. Therefore, the average time on page for all users is displayed as 5 minutes on Google Analytics, even though in reality, the average amount of time users had stayed on the page was 10 minutes.
Looking at scroll depth in conjunction with time spent on page, should give you a clearer view into how users were behaving on the page before they bounced.
As mentioned, the bounce rate is just a number which should neither be considered to be good or bad. Look at which pages the bounce rates were occurring, how users were landing on that page, and how they were behaving once they landed. I've come to realize that looking at data from Google Analytics on the surface level can give us an overview of how the site is doing, but there is a lot more information available when you dig deeper which can help diagnose where and what things are actually occurring on the site. The key is to keep digging deeper in order to figure out how and why your bounce rate is at that value. Don't be worried if you think your bounce rate may be too high. You might find a very good reason for it!
For more information, check out these articles
- End of Dumb Tables in Web Analytics Tools! Hello: Weighted Sort
- How to Use Timer Triggers in Google Tag Manager [V2]
- A Google Analytics plugin for measuring page scrolling
How Can the Peak-End Effect Help Users Get the Best out of Any Experience?
In April, I attended an interesting presentation at the CHI conference in Seoul, South Korea (you can read about my overall experience at CHI here) called, Examining the Peak-End Effects of Subjective Experience.
So first of all, what exactly is the peak-end effect? When we experience something, whether it's buying something on a website, or eating at a restaurant, the moment-to-moment experience is not constant throughout. Some parts tend to be better or worse than other parts, and when the user takes into account the entirety of these experiences when they think about re-visiting that experience. The peak effect states that intense positive or negative moments will be weighed more heavily when reviewing the overall experience. In addition, due to the end effect, the user will be heavily influenced by how the experience ended when rating it. These two effects combined, the peak-end effect, can have a large impact on how the user feels upon looking back at an experience.
I'm sure it's something a lot of us have experienced in our daily lives. Take the amusement park, for example. Wait times for rides can sometimes be over an hour. Even though the ride itself may be 10 minutes or so long, the experience it can give us can be so intense, that it may make up for the entire length of the wait, leaving you with a positive overall experience. The effect can work negatively as well. A single negative peak--whether it is a movie plot, eating at a restaurant, etc.--can leave a bad impression, no matter how the rest of the experience went.
From the presentation, we also heard of another example where researchers from Carnegie Mellon University had users compare progress bars that advanced at different rates. Some would load quickly at first, but slow down near the end; others advanced at a constant rate; and some loaded slowly at first but sped up near the end. Users preferred this final progress bar the most, even though the different types of progress bars that they saw all took the same 5 seconds to complete. Here, the end effect had left enough of a positive impression for users to rate that particular progress bar higher than the others.
In the presentation I attended, the lecturer spoke about his experiment that required users to perform tasks with identical objectives: setting 25 digital sliders on a computer screen to designated levels. The sliders were spread over a series of five pages with the number of sliders on each page being varied. Some pages had only 2 sliders and were easy to complete, while other pages felt more tedious to complete, having up to 7 sliders. These were distributed so that each user would experience an easy peak or difficult peak during their task, and could end on an easy page or difficult page. The researcher then asked users which of the sequences they preferred. The results showed that users slightly preferred tasks with just easy peaks or just easy endings, but a large difference in preference was only seen when the task had both an easy peak and an easy ending.
As UX practitioners, how can we leverage these results for improving products and services? Let's take websites for example. Unfortunately, not all online tasks are enjoyable; we need to fill in long forms, sites may be slow to load, and things may not work as expected. Hopefully, we can mitigate these pain points, but when it's inevitable, it's important to remember about the peak-end effect. A positive peak experience and a nice ending can leave a positive overall impression, and hopefully, keep the user coming back.