UX – putting the user first

Matt Balara explains how an established publisher, De Gruyter, completed an extraordinary transformation

How did your UX journey take off?

At OpenAthens’ 2022 AccessLab event, De Gruyter was announced as the winner of the third annual Best Publisher User Experience (UX) Award 2022. 

When 2020 began, De Gruyter had a site running on a third-party system, only a few internal developers, and no UX team or strategy. In February 2021 we had a new director of UX design, a well staffed internal development team collaborating closely with our partner 67 Bricks, and we launched a completely new degruyter.com, largely based on user feedback from a two-month beta period.

Since launch we’ve been iteratively learning about our users, evaluating UX performance, defining our goals and updating the site in an agile scrum process. We looked at our site and users through the lenses of data analysis and human behaviour.

User interviews

To understand our users, we conducted one-to-one, hour-long interviews with 10 librarians and 10 researchers in five different countries and a variety of disciplines, focused on the basic question: ‘How do you do what you do?’

The zoom interviews included a great deal of context – where they work, what their jobs entail, what tools they use, how the pandemic has changed their work – as well as screen sharing to observe how they perform their daily tasks, navigate our site and others, how they organise their files, how they work with email, Excel and other software, and the day-to-day problems they have.

Analysing the recorded interviews gave us a good list of commonalities and differences between user groups, primary tasks we could help them with, and frustrations we could alleviate.

Data analysis

For a quantitative perspective, we took the Google Analytics data and the feedback we’d received during the beta, as well as the database of customer service requests from the last year, and combed through them to identify user behaviours and complaints, and the most common problems we could solve to provide the most user impact possible. Customer service tickets and specific analytics metrics also form our ongoing UX KPIs for the future.

The interviews, data analysis, and beta feedback defined the majority of our scrum backlog. 

What does the development process look like?

Our process is quite streamlined: quick and dirty wireframes define an idea and basic interaction, more detailed screen design defines the look, and interactive prototypes nail down detailed interaction, and demonstrate how it feels in use.

With a UX team of only two people, we realised it was impossible to keep up a time-consuming usability testing process, and decided on a data-driven UX strategy for 2021, until we’re able to hire a UX researcher. We conduct reviews of our Hotjar data – heatmaps and recordings of mouse movement and clicks – and Google Analytics every two-week sprint, to understand what users are doing on the site, and how our changes have affected them.

One good example of tangible results from this approach was noticing a large number of users ‘rage clicking’ on DOIs (repeated quick clicks are a clear indicator of frustrated expectations). DOIs are required to be links by the DOI guidelines, but link to the page the user is already looking at, therefore clicks do nothing. After identifying the problem and creating a ticket in our backlog, DOIs were styled to not look like links in the next sprint, and rage DOI clicks all but disappeared. A simple, quick solution that removed a great deal of user frustration.

How about user feedback?

As mentioned above, our small team is unable to manage constant user recruiting, scheduling and testing, so we’ve made do: while they’re not users per se, a large number of our colleagues were librarians, researchers, authors and professors before joining De Gruyter, so integrating them and our developers into concept discussions and prototype testing sessions is an invaluable part of our process. 

Applying a lean, ‘Don’t Make Me Think’ approach, running short tests with three to five users, has quickly generated changes to either our design in progress, or the already deployed interface, with simple improvement tickets usually implemented in the next sprint.

Where have you improved?

Our partnerships with LibLynx and PSI have greatly improved user access to content, as shown by a dramatic drop in access-related customer service tickets since launch, and our statistics on authenticated vs unauthenticated users on the site. However our user interviews demonstrated what a confusing mess authentication is to most of them, and many users still ask, ‘Am I authenticated? How am I authenticated? Do I have access to this?’ 

Until more fundamental interface solutions were possible, we came up with three ideas:

• The ‘How access works’ page, which explains in simple language how to get authenticated, and dynamically displays the user’s current authentication status. This page is currently the second most-viewed page on our site, with an average time on page of over one minute. We’ve also received unsolicited user feedback thanking us for the clear explanation.

• A user’s authentication status is always displayed at the top of every page in a slim banner: green for authenticated, with details on how they’re authenticated, and yellow for unauthenticated. Both link to the page mentioned above.

• Dynamic download buttons, which make it clear whether a product can be immediately downloaded, or only after authentication, depending on the user’s state. We’ve also begun talks with the Seamless Access group, and are working to implement a solution in line with RA21 recommendations by the end of the year.

Last summer, an external consultant conducted a thorough accessibility review of degruyter.com. She gave us high marks in comparison to our legacy site, which she also reviewed, and delivered a short list of remaining issues for our backlog. We’re now compliant with Web Content Accessibility Guidelines (WCAG 2.1 AA).

What are your main goals and how do you keep track of progress?

Coming from a site, and process, with little to no consideration for UX, our primary goal last year was to achieve what we think of as ‘baseline UX’ – that is, no bells and whistles, but everything on the site is clear, readable and usable, driven by user needs, and we have an effective process in place for ongoing improvements.

From our ‘quant and qual’ analysis we understood a few things very clearly:

1. By far the biggest hurdle for every user is authentication.

2. Over 80 per cent of users come to our site from search engines, get what they want, and leave.

3. Load times make a huge difference in how long users stay on the site, and if they find what they’re looking for at all.

4. Users are always time-poor, so simple, clear and readable pages support them finding what they need quickly.

5. Less goal-oriented visitors need a browsing experience.

Authentication

Customer service tickets, one of our primary KPIs, related to content access have almost disappeared since we deployed the improvements already mentioned above. This year, we plan to move away from the default WAYF page LibLynx provides, and implement individual login, register, institutional authentication and passcode pages, clearly linked in context, to reduce user confusion. We also look forward to integrating institutional authentication with Seamless Access, and appreciate the Seamless Access group’s efforts to establish this standard.

Search engines

Since almost all users find us through a search engine, almost all users don’t see our site. They see one product page, making product pages the top priority for improvements. Interviews showed the most common use loop is search engine > publisher product page > download > back to search engine > another product page > and so on, so making our pages as discoverable as possible – and making the download call-to-action as obvious as possible – were the goal.

From the time of our previous platform, our search engine traffic has doubled, and downloads have increased by almost 50 per cent, so far more users are finding and using our content than before. An in-house SEO expert was hired in March 2021, and has been actively involved in our scrum process, so we expect even more improvements in the future.

The new technical infrastructure, built by our platform team of De Gruyter and 67 Bricks developers, delivers each of our roughly 10 million product pages in 2.8 seconds on average, in comparison to 14 seconds on our old platform, so this problem was already greatly improved at launch. We continue to monitor our Core Web Vitals and adjust to improve this wherever we can.

Simple, clear and readable

Clearly structured typographic hierarchy, nothing that distracts from a user’s goal, and calls-to-action that pull a user’s attention aren’t the sexiest UX topics, but experience and numerous studies (for example, Nielsen Norman Group) show that getting these things right helps users quickly find what they’re looking for.

Reviewing the state of the site in the beta, we have iteratively refined the typography, removed little used, distracting buttons, emphasised important calls-to-action and clarified the visual hierarchy of pages continually, with changes deployed every month since we launched, and will continue to do so.

Although most visitors arrive from a search engine and land on a single product page, we’ve identified some users – for example, authors interested in publishing with us, librarians curious about what we offer, and researchers keeping up with new publications in their discipline – who need a more general way to browse our huge product catalogue.

Our first step for this was to relaunch the homepage of our site, providing a two-level subject navigation, and lists of the newest and most popular products, dynamically generated based on publication and usage data.

We also worked on more comprehensive dynamic pages of new and popular products, focusing more on specific product type and discipline combinations, such as new architecture books or popular open access products on linguistics. These pages were deployed at the end of last year. The placeholder, search-focused homepage at launch in February 2021. The newer, browsing-oriented homepage deployed a few months after launch.

Evaluating impact

Our primary business model is B2B: we sell our products to institutional libraries, and provide access on our site to their professors, students and researchers. If these users are happy, our customers – librarians – are also happy. So, again, one of our main ways to judge the effectiveness of our UX improvements is customer service tickets, and what features they relate to. Since launch, overall ticket numbers have dropped almost 20 per cent, access-related tickets far further, and our new webshop has generated almost no tickets at all, despite very active sales.

Our secondary business model is B2C: selling our products directly to users on our own site, and through third-parties like Amazon. Our new webshop, launched this summer, has already generated double the revenue of the shop on our old site. Next year we’ll begin a comprehensive usability review of the webshop, with the primary goals of increasing time on site, number of pages viewed and online sales.

How is the UX approach being adopted across the company?

This is a long road, and we’ve admittedly only begun walking it. We’ve primarily approached this through education, and establishment of cross-disciplinary channels of communication.

User experience is a foreign concept for most people, so we’ve been holding regular company-wide ‘UX 101’ presentations to explain how we work, and why it’s good for De Gruyter and our users and customers. Longer term, once UX is a familiar idea, we aim to begin coaching colleagues in other departments on how they could integrate UX thinking and methods into their own work.

Beyond that, we’ve striven for open communication with colleagues in our customer service, sales, production, marketing and editorial departments. By integrating them into our concept development, using them for testing of our ideas and prototypes, and gathering their unique perspectives on their contact with, and feedback from, our users and customers, we’ve learned a great deal, and they have begun to see UX methods as a normal part of day-to-day work.

What are your future plans for user research?

Our most urgent plan for user research is to hire a user researcher, hopefully early next year. As already mentioned, our small team doesn’t have the time for the thorough, constant user research we’d like to do, and we currently depend on data from Hotjar, Google Analytics and customer service to help us define goals and evaluate our development.

Starting in the first half of this year, with the help of our new user researcher, we plan to conduct bi-monthly usability reviews with external users and focus groups with librarians, implement the infrastructure for A/B testing to evaluate how new ideas perform, conduct further, more focused interviews to identify user needs and opportunities to better serve them, and a complete usability review and refinement of our webshop.

No website is ever ‘done’, but we’re pleased with where it’s going and we are honoured our commitment has been recognised by OpenAthens.   

Matt Balara is senior manager for UX design at De Gruyter