How Data can be Used to Improve UX Design
In the current age of technology, if an entrepreneur comes up with a grand idea, chances are they’ll need a pretty sweet website to go along with it. And if they want their idea to really sell, they will also need a website that really catches the eyes of the audience. This is where the wonderful world of UI/UX design comes into play.
UI vs UX Design
First, we should probably establish a rough understanding of what UI design means compared to UX design. UI stands for “User Interface”, and UI designers are usually concerned with things like which fonts and colors to use, what kind of padding to put on widgets, etc. Basically, UI is for the most part all aesthetics. UX, on the other hand, stands for “User Experience”. UX designers are charged with deciding how users actually interact with a website, determining things like: where on the page should this widget go? Should we open a new window when “button” is clicked, or open up a modal pop? Is a scroll better for this list of products, or should we introduce pagination?
While data can certainly play a role in both, for this blog, we will be exploring the ways in which UX design can be drastically improved with the help of a data scientist. The below sections will outline a few of the many, many applications data science can play in the UX design process.
Figuring out which Webpages Should get the Most Attention
This question is very applicable to any e-commerce company. Imagine you are selling a plethora of goods, stored on a large amount of different webpages. As any savvy business owner would wonder, you might be asking yourself “Which products are selling the most?”, “Are people viewing a certain product, but falling off somewhere in the buying process?”, or “Which products are not even being viewed”. These are all questions that can be addressed by proper UX design.
A UX designer working for an e-commerce company, for one, needs to make sure the user experience for an end-to-end purchase is seamless. If the different webpages are Product > Cart > Shipping > Billing > Checkout, and we notice a loss of users, or high “bounce rate”, on the Shipping page, then the UX designer knows where to focus their work. Something about the Shipping page might be confusing, or intimidating, to a user.
Similarly, if the company wants to upsell a certain product, then it falls to the UX designer to focus their efforts on the webpage that showcases that specific product. They need to make sure that when a user interacts with the page that contains Product X, that Product X is displayed clearly and beautifully. Using data analytics, we can gather information regarding the percentage of times this specific product is clicked for each webpage view, and then work iteratively with a UX designer to increase this percentage.
A Website Help Center Utilizing a Recommendation System
Many websites out there are not easy to use, and thus, website owners decide to have a “help” section of sorts, often times with a means to chat with a representative or view an archive of helpful articles.
For a UX designer, one question might be “How should I most appropriately show the list of articles?” Maybe some kind of folder metaphor, or maybe simply just alphabetize them. While not terrible approaches, they are a little bit archaic.
With technology today, companies are able to keep track of user data. Using this data, we can use machine learning algorithms to tailor a list of appropriate articles to display to an end user, and present those within a “popular” or “most relevant” articles section. Instead of searching through hundreds of articles to find the one most relevant, wouldn’t it be sweet if the system guessed what you were looking for before you even knew it? And if the guesses are not helpful, there is always a fallback option of scanning the entire list!
A Powerful Homepage using Natural Language Processing
A website homepage should more or less accomplish the following: be an aesthetically pleasing webpage with some kinds of links to other places in the website where actual processing can occur. Now imagine an insurance company’s member portal. As an individual paying for insurance, there are a multitude of things you can do on such a website: file a claim, enroll in some coverage, cancel some coverage, add a dependent, add a beneficiary, pay some bills, etc etc etc. The list is huge.
A homepage for this website should, to some degree, have a way for the user to navigate quickly to any subpage of the website to accomplish any of these tasks. However, if a UX designer tried to cram all these links and descriptions along with whatever else belongs on a homepage, clutter would be an understatement. Additionally, a non-tech-savvy user may be intimidated by such a page, panic, and shut their laptop.
Instead of all these links, widgets, and descriptions, a data science powered Natural Language Processing widget with one heading titled “How can we help you today” would both remove clutter and put the user at ease. If a user came in due to a broken leg and wanted to see if he qualified for an accident insurance claim, he could simply type “Broke my leg” in the widget and be brought to the appropriate webpage to file a claim, thanks to natural language processing, an extremely powerful sector of data science.
