Plants of the World Online – Royal Botanic Gardens, Kew

I chose a card sorting exercise as a way to understand how different types of users grouped the information. I conducted the activity with more than 150 external participants and over 30 internal participants, including Kew scientists who regularly used the application.

Participants grouped individual pieces of plant information according to how they felt they belonged together, then assigned names to their groups. I also asked them to rank the information in order of importance.

This gave me two complementary perspectives: how users mentally organised the content, and which information they prioritised.

To interpret the results, I created a 3D cluster visualisation and a similarity matrix. These helped identify which information types participants consistently grouped together, revealing four clear categories:

• Taxonomy / nomenclature: Scientific names, synonyms, publication details and taxonomic relationships.
• Images / gallery: Plant photographs, illustrations and herbarium imagery.
• General description: Morphology, geographical distribution, habitat, common names and plant uses.
• References / other resources: Research papers, source references and links to additional Kew data.

The analysis provided an evidence-based foundation for the information architecture, rather than relying on assumptions about how a scientific resource should be structured.

By conducting the research with both scientific and more general users, and keeping their results separate, it allowed me to compare how each type of user views the information.

The ranking exercise revealed an important distinction between the two audiences.

Scientific users placed greater importance on taxonomy, nomenclature and publication-related information. General users prioritised imagery and descriptive content more highly.

This reinforced the central challenge: presenting a huge amount of information in a way that allowed both audiences to find what they needed without forcing them through an experience designed primarily for someone else.

When considering how to present so much information on the page, I explored a sticky side navigation as one possible approach. However, keeping everything on a single page would have created a lengthy experience in which users needed to navigate up and down to find relevant sections.

Instead, I introduced three primary tabs: Taxonomy, Images and General info. This approach was a direct result from my research, as it grouped related information into focused views, allowing each type of user to quickly access the content they considered most valuable. It also reduced the amount of content presented at once, preventing less scientific users being overwhelmed by the amount of scientific information available. The Taxonomy tab was visible by default, reflecting the slight majority of scientific users and ensuring that specialist information remained at the forefront of the service.

This also meant that I could use the space at the left for a sticky section header. As some of the sections can be quite long and data heavy, separating the header from the content makes each section much clearer, and having the header always visible prevents the user from losing track of where they are.

The research also influenced how information was prioritised within the page itself. I used the ranking results to determine which details should appear in the header, including the scientific name, taxonomy, publication information and accepted status. This allowed users to access important identifying information immediately, while keeping the more extensive supporting content organised below.

I also introduced a secondary level of tabs to help users jump directly to relevant sections within the main view. This divide of information made it easy for the user to get directly to the information they needed, whilst preventing a long list of tabs as before. The visual distinction between the two levels made it clear that the secondary navigation belonged to the selected primary tab.

Together, these decisions created a layered information architecture: important information was easy to access, related content stayed together, and users could navigate directly to the level of detail they needed.

Questionnaire responses indicated that users did not always find the information they were looking for easily. I therefore improved the search experience as part of the wider redesign, addressing discoverability alongside the information architecture.

My solution included an update to the visual design of the page, as well as improving the relevancy of the auto-suggestions, to ensure it was suggesting the type of content people were searching for. I used the data from the search logs, as well as questionnaires focussed specifically around search, to analyse what people were searching for. In the previous application, the autocomplete suggested characteristics that a user could search for (e.g. yellow flowers) however this created a cluttered experience, and research showed that users were not conducting that type of search.

For that reason we decided to remove the characteristic suggestion, and focus on suggesting the taxa in an order that made it much easier for the user to find what they were looking for.