As the sole designer, I redesigned the process for assigning and distributing trainings, assessments, and policies in a course and learner management system.
Our LMS platform distributes required content (assessments, policies, trainings) to employees who must complete or attest to it. In the existing product, our users would have to use a dense, settings-heavy interface built for power users created over 10 years ago.
This design best fit the use case of a power user who creates and manages content assignments on a frequent basis. But the majority of our actual users were admins who typically assigned required content around once a year. For them, every extra page and unused setting is a place to stall, second-guess, or make a mistake. That unneeded friction and complexity was showing up directly in customer churn numbers.
At the outset of the project, our monthly customer churn rate was 7.8%, much higher than the company target of less than 4%.
To gauge the usability of the current design, I conducted moderated usability tests with 30 participants who were current users of our product. To gauge usability, I measured time-on-task for the following tasks:
Task 1 (Role- based assignment)
Create an assignment that trains Engineers on Data Privacy with a due date set one week out and an end date two weeks out.
Average time on task: 2.6 minutes (156 seconds)
Task 2 (Role + individual user)
Create an assignment covering all code of conduct content, assigned to all Vice Presidents in the company plus one named person (Jennie Chen), with a due date one month out and an end date two months out.
Average time on task: 4.42 minutes (265 seconds)
For us, these times were far too high for simple tasks that were part of our key workflow. These initial tests revealed some areas where our users got tripped up. In general, users were having a difficult time sifting through the maze of settings, some buried behind modals and tabs. For Task 2, I found that the additional step of including an individual user almost doubled time-on-task. In discussions with stakeholders, using these test results as a benchmark, we set the definition of success as the following:
I interviewed internal subject matter experts and product stakeholders (across product management, customer success, and content/training) to inquire further about our users’ issues. These interviews revealed pain points which fell into a few themes:
For an overview of the original flow, I mapped the existing assignment journey step by step and annotated each point with the friction it created.
The original flow:
Each step had its own sub-decisions and unexplained concepts, things that a once-a-year user most likely wouldn't recall.
Main Takeaways
As a result of the usability tests, my interviews with SMEs, and mapping out the old design, I created these goals to guide my redesign:
My new design is a questionnaire style flow that simplifies the steps and guides users through each of them, while providing clear labelling and instructions, instead of dropping them into a complicated and dense settings page.
What I'm calling step 0 allowed users to choose how they would like to begin creating their new assignment, either Select Content or Select Assignees. For this first iteration, we wanted to gauge whether this flexible starting point would be useful or whether our users had a definite preferred path.
If users choose Select Content as their starting point, they can then select which content they would like to assign. Users can search and filter by content name and types. Available Content and Selected Content are separated so users can see a running selection of what they’ve chosen.
Users can select who they would like to assign the selected content to. Users use tabs to select by either Individual Users and User Groups. The assignees can be filtered by their location, job titles, departments, or custom groupings. I included the ability to “skip” individual assignees, in the case that admins would like to select a group but exclude a couple of users in that group. This was a highly requested feature from customers.
All settings related to the assignment are configurable from this page. Settings have been grouped (basics, schedule, advanced) for findability. The most important, most used settings are surfaced.
Default values are selected based on the most likely selections. This allows users who are unsure or don’t typically adjust certain settings to not have to consciously make a selection.
All previously available settings are now in a collapsed “Advanced settings” section. The focus here was visual hierarchy and progressive disclosure, so that we have feature parity with the legacy design without the complexity getting in the way of our typical users.
I made the key decision to reduce the use of modals. The old design made excessive use of them and there were many instances of modals on top of modals. By using modals more sparingly, information and controls are surfaced and hidden more thoughtfully.
The last step is a confirmation screen organized into Content, Audience, Basic Settings and Schedule, so all of the assignment details can be verified before committing. The old design didn’t give an overview of all the user’s choices before completion. A mistake in this task could mean sending a mistake to thousands of assignees. This step shows the full who, what, and when before completion.
I used Claude Design for ideation and rapid prototyping, moving quickly from high fidelity mockups to a fully clickable prototype. I used this prototype to quickly validate and iterate concepts with stakeholders. I also used the prototype for the next phase, moderated usability testing for the new design. Overall, using AI tools accelerated the pace at which I could iterate, create complex prototypes, and make adjustments in response to research insights.
To test our new design, I ran a similar test to what I conducted to gather our benchmark metrics. I recruited another 30 participants for moderated usability tests with almost the exact same tasks as before. However, I did make these minor changes:
Task 1 (Role- based assignment with an exclusion)
Create an assignment that trains Engineers on Data Privacy, excludes one named person (Jennifer Smith), sets a due date one week out and an end date two weeks out, and repeats every year.
Task 2 (Group plus a named individual user)
Create a non-repeating assignment covering all code of conduct content, assigned to all Vice Presidents in the company plus one named person (Jennie Chen), with a due date one month out and an end date two months out.
These changes were made to test the usability of the added recurrence and exclusion features.
Across the 30 sessions, every participant completed both tasks without difficulty, and almost all of them described the experience as modern, intuitive, and faster than the original design. Notably, there was one participant who had never created an assignment before. He completed both tasks confidently and quickly, a direct signal that the guided flow is learnable for the new/once-a-year user it was designed for.
The findings translated directly into a design changes for the next iteration:
Removing Step 0, the flexible starting point, simplified the flow even further.
Another 30 usability testing sessions were conducted after making changes based on the first round of tesing. Feedback and metrics improved further. These were the final results:
Task 1:
Task 2:
While we missed of the 50% improvement goal for Task 1, we were still very happy with the result. Overall, these tests showed our new design was much more learnable and quicker to use.
The new design was rolled out to users after 3 months of research, design, and development. Four months after release, the customer churn rate for our product had reduced from 7.8% at the outset of the project to 3.9%.
Overall, I am very happy with how this project went. Although we fell short of the 50% improvement goal for the first task, the reduction in churn felt like the more important metric of success.
I thought that the research and design phases went well. The original usability tests gave us a great benchmark to test our new designs against. The SME interviews gave me a broader understanding of our users' issues which proved to be very helpful.
This project highlighted how AI tools could accelerate my workflow. I was able to iterate and create prototypes much quicker than I could in the past. My design and research process benefited greatly from using Claude Design.
Ideally, I would have loved conducting usability tests with more complex tasks. I was given a tight timeline for participant recruitment and testing which limited how complex the tasks could be and how long each session could be. The tasks I wrote were also quite simple assignments. I think additional tasks diving into advanced settings could have given us even more useful insights to iterate on.
Overall, this project has been a lesson on the importance of designing for your audience. The original design had a valid use case, but it did not fit with the majority of our users. That mismatch was such a big frustration for customers, it led to churn. Being able to create a new flow from the ground up, while keeping our audience in mind, allowed me to design a flow that was intuitive and easy to use for our audience, while retaining feature parity with the original design.
