- Improving upon an already existing flagship database product, increasing subscriptions from customers by enhancing the existing offering and improving the user experience.
- Work type
- Databases, UX Research and Design, International
- Continuous, over a period of two years
- Senior UX Designer (me, London)
- Junior UX Designer (Pune, India)
- Technical Product Owner and Business Product Owner (Heidelberg, Germany)
- Business Analyst, Project Manager and 6-8 Developers (Pune, India)
Springer Materials is a subscription-based online database, providing information for chemists, engineers and other scientists on the physical properties of elements and compounds. As Senior UX Designer on the project, it was my responsibility to develop the site to improve engagement, reduce attrition, and develop value for customers. In order to achieve this, I engaged a programme of work that would help address these issues.
Research and Analysis
To start, I conducted analysis of the current information architecture of the site. I started by conducting user research and testing sessions, which included the following explorations:
- How do you use Springer Materials in your everyday work? – this helps me understand how Springer Materials fits into their working life, as well as giving me wider context around the requirements and decisions that cause them to need to use the site.
- What kind of information do you look for? – the breadth of data on the site is quite wide, and used by scientists across a range of disciplines. This helped me to define what information was more popular on the site.
- Can you show me how you get to this information? – switching over to a little more of a user testing scenario, I would get users to show me how they would find information, and observe the number of steps and ease with which they found what they were looking for.
Following these research sessions, I identified some of the more popular data sources on the site, and drew diagrams to explain how they were currently located on the site:
As the example above shows, users have to take a large number of steps to locate a simple piece of data. This was an example of classic “experience rot” – continued added functionality without overall scope degrading the overall user experience. One key reason for this was due to the fact that Springer Materials was originally created from the Landolt-Börnstein series of data books, which had been scanned to PDF, and broken up into relevant sections, resulting in users having to download the relevant section, and then go through the PDF to find the datapoints they required. Along with the series of unnecessary steps to locate the PDF in the first place, this was a distinct problem for users, as they came to Springer Materials to find information quickly in the first place.
From the research, I had understood that users had trouble locating information on the site, which caused frustration, abandoned user journeys, and had an impact upon sales of subscription renewals to the product. It was clear that we needed to reduce the number of steps taken to find information. The way that we did this was through improving the search on the site, which would allow people to get to the data they needed more quickly.
Improving the search – bringing context to results
In order for the site to answer user queries more quickly, we needed to provide better context for the search terms. Our previous search method used a text-based search, similar to that found in many library systems, where it took the inputted terms, and searched for incidences of those words in the text it held. A user looking for “tin” would get results for “tin” (the simple metallic element), but then also get results for “TiN” (Titanium nitride, a ceramic compound). We clearly needed to provide better context for these searches, in order to reduce unwanted search results.
We collected our ideas together, and elected to improve the search in two ways: first by introducing a suggestions drop down below the search input, allowing the user to specify what they mean by what they are typing in. The second part involved bringing in a graph-based search system. This meant that if the user selected something from the suggestion drop-down, it also incorporated a knowledge base that provided more context about that subject. For example, instead of seeing “tin” as merely a combination of the letters t, i and n, the system sees “tin” as the metallic element, with the chemical symbol “Sn”, the atomic number 50, and other details, which helps provide more relevant details. You can see evidence of this on the live site – a search for “tin” on the ‘focused’ search (which is what we called it to avoid technical confusion) gets 263 results, whereas searching using the old style text search provides 3653 results.
“This is great. It really will save me a lot of time searching in the future.”
Corporate researcher at MRS Boston 2016
Improving user pathways – reducing steps
Not only did changing the search help reduce the number of reduce the number of steps the user took to get to information, but we also used the studies I had previously done to analyse where we could improve some of the paths taken after searching to reduce those number of steps there, too. I provided a series of proposals of which paths could be reduced, and provided them to the team to implement:
While some of these proposals were acted upon, some of the paths in Springer Materials have yet to be amended, as they became quite a long and involved task. However, we managed to improve the key “red routes”, or most popular user journeys, which reflected well when tested with users.
One of the key parts of the plan was to try and bring the data from out of the PDFs, into a digitised format that could be easily manipulated and experimented on by the user. This proved a difficult task, as it was not merely a case of copying the data from the PDFs onto the site, but also having to curate the data, in order to ensure that it was organised in such a way that allowed it to be found easily. Getting stakeholders to buy into this idea took time and persuasion, but I ended up getting them to hire the services of allied post-doctorate teams based in Germany and the USA, who assisted us with the curation, and made the data ready for us to use.
We combined the new data with producing a new set of pages to display the data on, which we referred to as the Landolt-Börnstein Digital, or Springer Materials Interactive pages. These pages would not only display the desired datasets, but also provide the user with interactive visual elements, which showed the spread of the data, and allowed the user to select a slice of the information, before having it represented in tabular format.
I also encouraged the team to look at the whole user journey, providing suggestions on how the search snippets could help to show basic information before getting to the whole dataset. This, sadly, wasn’t acted upon, due to resource restrictions, and we focussed just upon producing the actual interactive page. However, going through several different iterations, which were tested at each stage, and met with some very positive responses.
“This is awesome. I don’t think anyone else is doing this. Where do I sign up to get this for my university?”
Student researcher at ACS 2016
Issue: finding users to research with
One of the major problems I found with trying to conduct user research on this project was the was in which I wasn’t in easy contact with the users. On other projects I had worked on, Project Owners and stakeholders would often be in contact with the users, as those owners were often both the people who used the product, and the people who purchased access to it. However, in this case, there was a more complicated set-up involved.
Due to this rather protracted relationship between the production team and the customer institution and its users, I was not provided with users with which to conduct research and testing sessions with. However, I used the following methods to locate the much-needed users to work with:
- Online recruitment – researching customer institutions online, on sites such as LinkedIn, Science Direct and others, I would collect lists of possible candidates. I then would cold-approach them, which had a relatively low success rate
- Recruitment drives – I would send notices round to customer institutions asking for volunteers. This also didn’t yield high numbers of recruits, but helped to complement the ones I had already attained
- Website advertising – using banners on the site, I would appeal to users to take part in research and testing. This had some degree of success
- Guerilla testing at conferences – I attended conferences where we had a sales presences, such as the ACS and MRS conferences. I would then approach people who came to the stall. This was a fairly successful way of getting quick 15 minute sessions of user feedback and testing, but the people I talked to weren’t always active users of the product.
For more information, you can read the blog post I made about these methods, and how others can use them for their research. This did provide sufficient numbers of users for getting feedback for the work we were doing, but I did have to continually justify the low numbers of results to Project Management, who would have, on other projects, been responsible for providing those very users. My persistence paid off, as the results showed that my work was having an effect.
As mentioned before, we received some good feedback from our improvements, when testing them individually with users. We also managed the following gains:
- The introduction of the Interactive pages brought it to the level of a “best in class” in its field, winning accolades from professional bodies.
- In 2017, the Indian Government bought licences to provide Springer Materials in all libraries across the country, which greatly widened the subscriptions and exposure of the product.
- A site survey, run towards the end of my tenure on the product, showed that 69.2% of users classified the product as “great” (in a selection between “great”, “okay” and “bad”), an improvement of 22% from a site survey run towards the start of my time on the project
I felt I definitely contributed well to the original remit of improving engagement, reducing subscriber attrition numbers and developing value for the users. The examples above show only the more significant parts of the two years I spent on the project, and don’t necessarily account for the issues faced with pushback from business interests, problems with user recruitment, and other hurdles I faced during the project. However, despite this, I can happily say that my work has had a lasting effect on the product, and has taught me a lot about getting results in less than perfect circumstances.