- Timeline: Sept 2015 – March 2016
- Context: Client Project at Mentor Creative Group
Koverse is a technology product that allows users to prepare, index, and search big data, all at high speeds and with customizable security features.
The creators of Koverse approached Mentor Creative Group with the goal of making their application easier to learn and use. The application had a powerful back-end platform, but an unintuitive front-end—especially for non-technical users. Koverse often sent employees to their customers’ workplaces, training them on the software for months, which was extremely costly.
I worked with a small team at Mentor to redesign the user experience and user interface of Koverse. The result is a product that maps closer to a data scientist’s actual workflow and is much easier to use out-of-the-box.
The Koverse redesign project kicked off with a six-week discovery phase. I worked with a team of four to conduct user interviews, contextual inquiries, and usability studies on the existing Koverse product. We also performed a content audit on the product and competitive analysis on other data analytics applications.
We developed our research data into personas, user journeys, and research findings.
A member of our research team working on clustering our user interview findings.
Based on our research, we developed detailed information and user journeys about these five personas, and used them in our design work throughout the project.
UX Ideation, Wireframing, and Iteration
Once research wrapped up and design production began, two areas of the Koverse app for which I was the primary UX designer were navigation and search. I explored full sets of solutions for these features, ideating at the wire stage. I presented multiple directions to the rest of the design team for critique.
As the team was very agile, usually working sketch-to-code, further ideation continued as our developer began building out functionality from the sketches.
One of our major tasks for Koverse was designing a new navigation system. The existing product was made up of disconnected mini-apps, and the navigation was not mapped to a data scientist’s actual workflow. One of my first steps in exploring the navigation problem was to sketch out an exhaustive list of navigation pattern possibilities.
This wireframe shows a navigation solution close to the final iteration we landed on. I combined six+ navigation paradigms from my previous exploration into one cohesive navigation system.
Searching data in the Koverse system is one of the primary tasks for which data scientists use the application. Therefore we needed a robust search feature that allowed users to search by different operators (e.g. date ranges, columns, authors). Relatedly, we needed to decide whether data could be sorted into folders or have tags or both. These are some early search paradigm options I presented to the team for critique.
This wireframe shows a search option that is close to our final iteration. It’s predictive—it gives users suggestions as they type. When clicked, suggestions navigate the user to their expected destination and also populate the search field with plain-language shortcuts. Power users can learn these shortcuts and use them in search queries in the future.
I created a new visual brand for the Koverse product. At the early stages of the project I used moodboards and style tiles to work with the client to determine a visual direction. I then worked with our front-end developer to create type, color, and UI styles for the product. The final step was creating mockups and working with the developer in-browser to create a final polished product.
Users need to add data sources into the Koverse system in order to start viewing and playing around with their data. This screen allows a user to upload files.
Since users of Koverse are usually uploading or connecting enormous data sets (big data), Koverse allows them to check out a sample of that data before committing to the lengthy process of pulling the entire data set into the Koverse system. At this stage they can also apply data normalizations, and view the effect of those normalizations on the sample.
Once data sets are added, users can view overviews of their sets that include automatically-generated data visualizations.
The final iteration of the search feature surfaces recent searches and provides suggestions once a user starts typing.