Project Objective
My Role
The Outcome
Project Background
I was tasked with improving the taxonomy and design of the website navigation for AJ Bell, a UK based investment platform. The aim was to make it easier for users to find key content by addressing a number of UX/Usability issues that had been identified in a prior Usability Testing project we completed.
The AJ Bell target market was a mix of new vs experienced investors, with wide ranging levels of subject matter expertise and familiarity in the investing space. This posed an interesting challenge throughout the project, as these personas have very different user needs, and reacted very differently to the categorisation and labelling of navigation content.
This project was focused on the lead generation website only. Our primary metric was to therefore gain an increase in account openings, by helping users to find key pages and information that addressed their needs.
Our Approach
To meet our objectives, we separated this project into two parts which allowed us to isolate feedback between the taxonomy and design. All research and testing was conducted across the four AJ Bell personas with varying levels of investment experience, as well as both non-customers and existing customers. Stakeholder Interviews
Part 1: Creating and testing a new taxonomy:
Open Card Sorting (with 10 users)
Creating a new Information Architecture
Tree Testing (with 246 users)
Part 2: Creating and testing a new UI:
Designing the new UX/UI
Unmoderated First Click, Reaction and Preference Testing (with 456 users)
Although we had a solid understanding of the navigations UX/Usability issues thanks to some prior Usability Testing, we wanted to understand how users structured and labelled website information, and establish if different user groups think in different ways about the information.
Therefore I used Optimal Workshop to create a de-categorised list of website pages that 10 users could then categorise and label in a Moderated Open Card Sorting study.
Through this, we identified clear themes in how all personas categorised content, which drove our ideation when it came to creating a new Information Architecture.
Results from the Open Card Sorting were compiled together to get an aggregate view
Initial ideas for the new IA were created utilising both this insight and psychology principles like Millers Law to help decide the structure and number of categories.
I then ran a stakeholder workshop, where we discussed the categorisation, labelling and number of pages. Among others, stakeholders from the SEO and Compliance teams were invited to identify any pitfalls from the proposed changes.
We finished this workshop with two versions of the proposed Information Architecture that we wanted to test in an Unmoderated Tree Testing study.
Tree Testing
A total of 246 participants completed the unmoderated Tree Testing study. With these responses split across Control and the two variations.
In order to segment results by experience level in line with AJ Bell personas, we recruited a mix of participants who had:
Under £5,000 in invested assets and less than 5 years investing experience.
Over £5,000 in invested assets and more than 5 years investing experience.
A total of 10 tasks were given to participants in each Tree Testing study, the order of which was randomised. After each task, users were asked two follow up questions to understand the reasons behind the paths they took.
Tree Testing study and results
With UI feedback anticipated to largely be subjective design based comments, Unmoderated tests were run to measure changes with a larger audience.
We used our Figma prototypes to test control plus two variations on both Desktop and Mobile with a total of 456 users. The studies we ran included:
Preference Testing: Users were shown two versions (e.g. V1 Mobile vs V2 Mobile) asked to choose their favourite, then asked to explain why.
Navigation Testing & Design Survey: Users were tasked with finding pages across 4 key categories, then were asked a series of follow up questions.
Results were analysed, using LLM AI to help analyse the large number of open ended answers. Results proved particularly positive for one variation, with further opportunities for improvement identified.
The Outcome
The validated taxonomy and design were placed into a site-wide A/B test, which led to a statistically significant increase in account openings, our primary metric.
Further opportunities for improvement were also placed in the A/B testing backlog for future optimisation.