Alectio - Redesign
Redesign onboarding for Alectio - a Data Prep Platform for Machine Learning
Understand Goal
Alectio encourages sustainable ML by valuing data quality over quantity, reducing resource consumption and ultimately leading to reduced technological costs. The existing platform included complex flows and provided users with a wide variety of options to get started with other ML projects journeys. The primary objective for the redesign of the platform was to simplify the onboarding process and provide users with a more intuitive flow on the platform.
Define Audience
At the time of the redesign, Alectio's platform was primarily intended to be used by machine learning engineers and scientists. We conducted a series of user interviews, the initial ones with the original platform to understand the goals and needs of the users, what motivates them, and what frustrates them. Based on the user interviews, I created personas that helped us make design decisions focused on the end user. The users' years of experience and knowledge can vary, but they all have similar motivations.
User context and needs
We discovered that Alectio’s platform was generally being used by ML scientists and engineers in the following context:
-
When - ML scientists would come to use Alectio’s platform when they would want to train a model, run an experiment or check the status of an experiment.
-
Where - ML scientists would usually be comfortable using the platform on their desktops, laptops from home, office, coffee shops, libraries. We envisioned the possibility of using a mobile app in the future if they want to check the status of their experiments.
Ideate and Prioritize
Our team formulated the feedback and ideas from user interviews and prioritized a few key areas to redesign which would have the highest impact to improve the overall efficiency of the platform. These were the areas we focused on:
-
Redesigning the platform layout
-
Project Onboarding experience
We started the ideation process by redesigning the user flow and creating paper sketches and low-fidelity mockups to initiate discussion within the engineering and the ML team. During these sessions, we gathered feedback and continued to incorporate the changes in our flow and the interface design.
Redesigning the Platform Layout
The original platform design had some content organization and navigation structure issues. It consisted of a side navigation menu for all available actions with no clear hierarchy and a top panel underutilized.
The proposed redesign focused on using the top navigation space and dividing the platform functionalities into sections for better accessibility and hierarchy. The primary menu in the redesign will allow users to navigate between two views efficiently - the Dashboard view which lists the user's usage statistics and other insights about projects and the Project view which serves as the starting point to utilize any of the functionalities offered by the platform. The primary menu also isolates and places the user menu and notifications for quick access and improved visibility.
Original Layout Design
Proposed Layout Redesign
Project Onboarding Experience
In Alectio ecosystem, a project is a {data, model} pair. Creating a new project is one of the most critical and essential parts of the Alectio platform. Many features of the platform require users to select a project, thereby making the project a prerequisite for additional activities. Originally, the platform allowed creating new projects either by clicking on the “New Project” button in the sidebar menu or cloning an existing project. There were a couple of issues with the existing approach which I have highlighted in the images below.
Besides the observations we made during the use of the existing platform, we also conducted user interviews with a few of the existing customers and the in-house ML scientists and developers to observe how they use the platform and identify additional issues. Based on our findings, we prioritized a few areas where we would focus to improve the experience of the platform. The proposed design changes for project onboarding functionality covered the following issues:
-
Master-detail layout to show an overview of completed and remaining steps to the user to keep them motivated and informed.
-
Create a wizard-like project creation template and pre-fill information where applicable to reduce the time for onboarding.
-
Add tooltips where necessary and simplify language to provide more information to users.
As a starting point, we worked on establishing a user flow for the project onboarding process which would feel intuitive to the user. We followed up with paper sketches and low-fidelity wireframes to brainstorm ideas about the redesign, gather feedback and improve the designs with the feedback received. This iterative process led us to refine our designs and decisions before we were ready to implement the changes on the platform.
Alectio is a B2B AI startup helping ML scientists increase their productivity and the quality of their models by curating and diagnosing issues in their training data. In my previous role as a UX Developer at Alectio, I worked to discover the areas of the platform where user experience can be improved and then prioritized the elements that need to be redesigned to improve the overall experience of the platform.
_edited.jpg)
Updated User Flow
Measure
To measure the effectiveness of the design solutions we implemented and to ensure that we were on the right track, we used a combination of behavioral and attitudinal metrics to measure, compare and track the quality of the user experience over time.
-
Task Success - This KPI showed us the percentage of customers who have successfully completed a specific task, in this case completing the creation of a new project. The task success for new project creation was calculated as the number of completed tasks divided by the total number of tasks attempted and we achieved a rate of 100% success. Additionally, the time per task was reduced significantly with the redesigned project creation module.
-
We also used attitudinal metrics to measure what people say and how they feel about our product. We ran a survey on the dashboard before and after the launch of design changes and asked users to respond to these statements on a scale from strongly agree to strongly disagree. The response was mostly positive with feedback to provide more value statements to the user to keep them posted about the impact of their actions.
In conclusion, my role as a UX developer in this Data Prep platform redesign project was to bridge the gap between the technical and design aspects of the project. I worked closely with the machine learning engineers to understand the capabilities and limitations of the model, and used that knowledge to design a user interface that was both intuitive and functional. I also conducted user research and testing to gather feedback and iterate on the design. Overall, the project was a success and demonstrated the importance of collaboration between the technical and design teams in the development of a successful machine learning platform.