Our client's product was a data analytics software that specializes in processing natural language inputs and is used by federal law enforcement in several states. Our client wanted design recommendations that would support its current users and drive adoption by new types of analysts.
TEAM | Pierre Amelot, Ethel Chou, Kate McManus, Anuraag Jain
ROLE | UX Team Lead
TOOLS | InDesign, Sketch, Illustrator
METHODS | grounded theory, contextual inquiry, affinity diagramming, cultural modeling, interview interpretation, speed dating, user interviews, visioning and concept generation
Our client, uReveal, is a search support and analytics company that wanted to explore new features that could help them expand to new users and markets. uReveal's specialty is the conceptual processing of natural-language materials like police reports, and their software was currently used by police and certain federal agencies. Our job was to identify current and potential users who could give us insights into new applications for UReveal's technology and from our research with these users to design feature sets that would increase adoption and improve user experience with the software. Over seven weeks, our team conducted a full contextual inquiry, including stakeholder interviews, data consolidation, ideation and a final design proposal for the client.
We conducted contextual interviews with analysts as they went through their analysis processes.
We diagrammed the analysis process, focusing on painpoints and breakdowns in the existing process.
We used a cultural model to better understand the issues between agencies that affected the data analysis process.
We "walked" the wall with the noted generated in our contextual interviews.
We create six visions after ideating many, many more.
We storyboarded these visions and used them to get feedback from our orginal participants.
We began by identifying 1. current users and 2. analysts from analogous domains who were potential users of uReveal. In the current users group, we spoke to anti-narcotics and gang analysts. In the potential users group, we spoke to analysts from other federal agencies, such as the Federal Energy Regulatory Commission.
We then conducted contextual interviews with each user as they went through their analysis activities. Contextual interviews allow you to encounter user's thoughts about their work as it is happening, providing insights that are more accurate and in-depth than traditional retrospective interviews.
After conducting the interviews, we each presented our findings to the team, who collectively determined the notes that would make up the raw material for our affinity diagram.
We synthesized our findings with current users by creating a consolidated sequence-flow of their analysis workflows. This allowed us to pinpoint tasks that are completed daily, weekly, and monthly, and to identify painpoints and breakdowns in the current users' experience.
Our target users work in complex environments in which information is passed between individuals and agencies. We created a cultural model for the vaious agencies to better understand the interactions between them and the needs of each stakeholder in the analysis process.
We then synthesized our findings from new and potential users by creating an affinity diagram of all insights from our contextual interviews. As anyone who has created an affinity diagram knows, it is a slow and iterative process. You arrive at real insights only after pushing past the obvious connections between ideas—thematic clustering—and get to the more surprising connections. In our case, we worked a dataset of over 600 individual notes over a dense two days before moving on to visioning solutions.
After generating insights by "walking the wall", we identified the most pressing paint points, breakdowns, and opportunities presented by the research. We then began ideating possible design solutions for each of these areas. After coming up with lots and lots of ideas, we narrowed our visions down to the six most promising ideas. These were chosen for the magnitude of impact that they could have, coupled with how feasible they would be to implement.
We then sketched storyboards of these ideas to use in speed dating with our original interview participants. In order to get quick feedback about each concept, we showed them a sketch of how the concept would work and explained a short story of how the concept might fit into their workflow. Each analyst then told us their initial impressions of the idea, what they liked about it, and what they didn't like. This allowed us to see how readily each concept would be adopted by the target users, and whether any concepts sparked resistance or skepticism.
We presented the leadership team of uReveal with 6 recommendations for features and capabilities. Within two months, the management team notified us that they were moving forward with 4 of our recommendations for immediate development, and that the other 2 would be fast followers.
Data visualization tools that are built-in + dynamic
Many analysts create visual presentations that they share with their team. The presentations determine what investigations are funded and they need a way to make high-quality visualizations easily. These tools could replace the current motley collection of software that analysts use to visualize the insights that they get from uReveal.
Data cleaning (i.e. deleting duplicate reports and removing irrelevant results) currently takes up much of a skilled analyst's time. Automated data cleaning would allow the program to learn rules from the analyst's manual cleaning and apply those roles to future results.
Automatic concept generator that leverages machine-learning
Right now, constructing a concept is a manual process that requires the user to input every relevant word for a concept. With the Automatic Concept Generator, users will enter a keyword or phrase and the program will construct a concept based on 1. common shorthand or misspellings, 2. conceptual synonyms, and 3. associated words.
Online sharing platform for analysts
Aka GitHub for uReveal, it's an enterprise knowledge base that leans heavily on user-supplied libraries, issues, and documentation. This will allow users to self-educate on the specialized applications of uReveal that are most relevant for them, and will give users the libraries that will allow for more sophisticated concept-generation.
Realtime video input
Videos recorded by police bodycams and other devices are automatically uploaded to the database, analyzed, and prioritized, and relevant data is returned to the officer immediately. This allows the officer to make informed decisions when facilitating incidents in the field.
Identifying emerging patterns
Right now, it is up to analysts to notice new patterns in field reports and use that information in their analysis. By their own admission, analysts only notice some of the relevant patterns. In this vision, the analysis tool would notice emerging patterns around the concepts already being run (a new name is popping up, for example) and notify the analyst of them as they occur.
My team's work on this project had one large constraint: because our participants were analyzing classified and sensitive data, we were not able to conduct interviews in-person. Instead, we asked participants to go through their normal analysis activities while we were on the phone with them. While this gave us access to their in-the-moment comments and insights, we were not able to see what they were actually doing. Because a key component of contextual inquiry is being able to observe and ask questions (and catch discrepancies between how a participant describes an activity and what they actually do), we doubtlessly missed out on useful insights.
That said, I was surprised by how much insight we were able to get from these interviews, and I believe that the resulting concepts have the potential to greatly improve analysts' experience with uReveal. We continued to have follow-up conversations with our client for several months after submitting our final report, and I'm excited to see how these suggestions are encorperated into uReveal's next release.