Improving information architecture for B2B products
At Gympass, we are recreating one of our products from scratch. It’s a big challenge to accommodate all the new features since this product is going to grow a lot. In order to solve this problem at once, we decided to adopt user research methods to guarantee we have the best information architecture from the client’s point of view and consider the future of the product.
We began running a card sorting
The first step was to create a list of all the content we plan to have on the product in the future, also considering long-term changes that we were already aware of. The list had tooltips with descriptions for each topic to help participants understand all the items. It was revised by members of the team to make sure it would be clear to the participants.
After the pilot, we noticed that some items like Profile, Login and Language switcher were getting in the way of the main features. They are utility navigation, as named by Nilsen Norman Group, and weren’t up for discussion. For the final version, we just took off these items, making the activity more straight to the point and easier for the participants to use. We also noticed some users haven’t used the tooltips to understand any card better when they had questions, so we improved the test instructions by adding a topic with the description on the tooltips.
We recruited real clients for the experiment since domain knowledge was essential to understand what each feature and piece of content meant. But that cost us time until we got enough answers because we’ve narrowed the public a lot. So if you are in a rush, consider being more flexible with your screener and incentivizing the participants with a bonus or a discount.
We chose to run an open card sorting which left room to understand how the users would organize the cards freely. On the other hand, standardizing the categories was more difficult. In the end, it was super useful and we could identify consensus forming.
Then tree testings
To validate and refine the card sorting results we turned it into a content tree and selected the items that we needed to validate considering main actions, pain points, and concerns. We then wrote the tasks to test each selected item locale on the tree.
The tree testing is easier and faster for users to complete compared to card sorting. But at this phase, we knew from card sorting that recruiting real clients would not be the most efficient use of our time. As the test now was about finding subjects in groups that were previously defined, recruiting non-clients was a safe (and time-saving) option.
On the first tree testing, we noticed we could change the place of some items to improve findability. Based on the first round results we made adjustments to our tree, considering the comments from the participants to change the items to a more suitable place, and tested only the tasks that have changed, solving all major problems.
And more tree testings every now and then
Running card sorting and tree testing as complementary methods resulted in a solid information architecture defined for the product. Since then, we needed to run other specific tree testings, because new things are always going to appear, definitions are going to change, and also concerns will come up about more detailed information than we looked at in the beginning. Given this ever-changing and evolving product, running tree testings became part of our routine.
Thanks for reading! I'm happy to share some knowledge with the UX community. Enjoyed the content? Follow Gympass ♡
Thanks to all Gympass Design Team. Specially to Bruna Maia, Fernanda Uekita, Gerardo Marques, Henrique Sugiyama, Mateus Pinheiro and Sarah Kessler.
