PatternAI (fictional name) democratizes the process of extracting insights from complex datasets. Starting with a set of criteria set by the user, PatternAI proceeds to a cluster analysis (machine learning technique) and exposes the best match to the user's request.
My work on this project is subject to confidential agreements, therefore all visuals will be stripped from actual data points.
Turn a very technical project nurtured by data scientists into a tangible product, and validate the interface usability. Then, ensure passation to the dev team for implementation.
In the next sections I will thoroughly explain my UX process. If you are a bit impatient, jump to the final design 🖥️
PatternAI can help them save time by automating repetitive data manipulations, and extracting insights faster
PatternAI is also meant to be usable for users with a less extensive knowledge in data analysis, but who are eager to get insights from their datasets
To understand the value proposition of the product, I had a few brainstorming sessions with the Data scientists who owned the project. After really immersing myself in their reality - which my past career as a data analyst really helped to do - I started crafting the experiences to make their vision become a reality.
The team was already enthusiastic about the first wireframes - but I convinced them of the importance of testing with potential users before going further in the design. Since PatternAI is a technical tool, it could easily become hard to use and understand if we designed it only based on our assumptions. We therefore recruited a dozen business professionals with a different range of data knowledge, and had them execute a few tasks on the prototype.
Assess the ability of users to upload their dataset, and define their analysis criteria
Determine if users can extract the right insights from the Clusters page
In the dataset upload screen, give the user the option to edit the data column type automatically detected by PatternAI. This provided the user more control, and helped avoid errors later in the data processing
I transformed the "criteria definition" step into a query format to reflect the way users asked themselves questions about their data. The previous format was too technical, and made it hard to visualize what PatternAI would extract
I added confirmation modals in the data upload step, to give users reassurance that their dataset was in the right format to proceed - or, invite them to make some modifications
In the results page, offer a clear reminder of the criteria set by the user - previously the information was not visible enough, and users thought they had to go back to the previous pages to see it
.png)
.png)
.png)
.png)