UX reviewing that you don’t need to do anymore


The number of usable websites in the world is low. This has a big impact on businesses. For example, eCommerce is losing hundreds of billions of dollars every year because of bad user experience.

Let’s take a look at a few reasons behind the lack of usable websites [1].
  • Usability evaluation requires experts to do it, and there’s a shortage of such experts
  • Doing a usability evaluation is a manual expert work making the cost of it beyond the budgets of many companies
  • Websites are developed in high-speed with frequent updates, and the sites are more and more complex, making usability evaluation and improvement of websites increasingly challenging
  • Evaluation gets overlooked because of market pressure to release rapidly and often, due the absence of distribution barriers
  • Usability evaluation can only be carried out up to a limited depth
  • Different usability experts give conflicting reporting
The good news is that bots can perform a significant number of usability evaluations currently done manually by experts. For example, in a study [1] where Jakob Nielsen’s and Marie Tahir’s book Homepage Usability [2] was transformed to automated rules, the team was able to trivially implement 51.48% of the usability best practices with high accuracy. Obviously, with the help of more advanced techniques, this number can be pushed much higher.

Let’s go through a few concrete types of observations that Attractive.ai catches and reports.

Layout problems and responsive design

A bot can detect if elements get incorrectly covered by other elements, essentially breaking the UI. Here’s one real-world example from tesla.com. Noticeably, an important order button is broken on their front page for the Model S.

Inconsistent in design

Visual consistency is one of the most important user experience factors. It increases usability, helps to locate the most important content, and evokes a strong emotional response. It strengthens your brand. Here’s an example where Huawei’s mobile menu layout changes when you change a page. This is confusing to users because something as common as navigation elements change places unexpectedly, forcing the user to scan the UI for the desired elements instead of just clicking in the previously learned location.

Accessibility

Of all web users, 15% have some sort of disability that makes it hard to read all the contents of a regular website. The EU has recently introduced the accessibility directive to make the web more inclusive. Here’s an example of a very common problem where the text is not readable because of a busy background image.

Cultural problems

Because your database has columns for the first and last name, it doesn’t mean your users have to suffer from it. Here’s a very common problem of asking for first and last names separately instead of asking just for a full name, which is consistent and works without problems in all cultures. Even extremely international companies like Uber make this mistake.

The basics of the web

Obviously, any tool should check for links. What’s more annoying than getting a 404 not found page when you try to book a flight or buy an interesting book. Here’s Wendys.com privacy policy page where every link is broken.


Another introductory level mistake 90% of websites do is having contact information non-clickable. To save gray hairs from your customers, just make phone numbers and email addresses clickable. Here’s Starbucks demonstrating bad practices.

What's left for us?

Of course, humans are still needed for many parts of usability evaluation. For example, we need humans to say whether high-level processes are logical and achieve the desired business goals. However, we can already cover a significant part of the work by bots. This releases our time for more creative work, for example, designing better UX solutions. 

References

[1] Justin Mifsud, Alexiei Dingli: USEFul: A Framework to Mainstream Web Site Usability through Automated Evaluation
[2] Jakob Nielsen, Marie Tahir: Homepage Usability: 50 Websites Deconstructed