Thomson Reuters: Machine Learning News App
Thomson Reuters — Deep Credit Model UI
What's the problem?
Thomson Reuters holds a patent for its Text Mining Model, which uses machine learning to scan news articles and assign a “default score” — a measure of risk that a company might default, used by investment professionals in financial analysis. The original model, based on keywords, produced unreliable results. A new Deep Learning model, capable of interpreting phrases and sentences, significantly improved score accuracy. The challenge was to design a new UI within Eikon (TR’s financial platform) that clearly demonstrated the value and robustness of this upgraded model.
What do we want to achieve?
• Build an intuitive UI and workflow to display the new Deep Credit Model within Eikon.
• Emphasize the improved accuracy of the default score.
• Support financial analysts in integrating the model into their decision-making process.
What are challenges we faced?
• Limited budget for extended user research.
• Ensuring the new UI felt consistent with Eikon standards while showcasing a cutting-edge model.
What was part of the team and what was my role?
As a UX Designer, I collaborated closely with the Head of Data Science and a Design Strategist to develop user-centered solutions that effectively integrate data insights and strategic design principles. This interdisciplinary teamwork ensured the creation of intuitive and impactful user experiences that align with both technical feasibility and business goals.
How was the process I followed?
The design process began with research and foundations, where I collaborated with internal subject matter experts to define initial personas and journey maps. This helped in understanding how investment professionals currently integrate the Text Mining Model into their workflows. Next, during the design exploration phase, I created sketches for two UI concepts that aligned with the existing Eikon design standards. These concepts were further developed into low- and high-fidelity prototypes using Sketch and InVision. Finally, in the testing and iteration phase, I conducted internal testing and refined the designs based on the feedback received. I also prepared and carried out a user research study involving five active users of the Text Mining Model to ensure the designs effectively met their needs.
What was the final solution?
Delivered a new UI within Eikon that highlights the improved Deep Credit Model and communicates the robustness of the new default score. The experience allowed financial analysts to trust and adopt the updated model with greater confidence in their risk assessments.
Sketches
Final Designs
