Overview

When I joined the ONE design team in the Fall of 2018, they had already spent the majority of the year concepting a centralized data platform t0 include streaming data and users beyond Engineers. Using Apache Kafka, this ecosystem would leverage AI to deliver information to internal Operations teams to assess risk and protect customers in many ways. I joined the Streaming Data program to update Capital One’s first data transformation platform that would eventually serve as the foundation for the future experience. Our goal was to create a unified workflow for both producers and consumers while serving the nuanced differences in each user’s needs.

SDP Project Synthesis

While similar to other datasets in general structure, streaming data had a few complex differences from batch data that the team had not yet considered. The consumer and producer experiences were siloed and didn’t consider the way users toggled between the two modes of working. The pain points of the publishing process were a lack of clear guidance and documentation around definitions and naming conventions, and how to make the process more flexible, iterative, and collaborative. For consumers, we needed to better understand what are the markers of trustworthiness. Additionally we wanted to find areas to automate the process and shorten the time to production at each phase.

Methodology

We began with exploratory research in the form of interviews, contextual analysis, and collaborative journey mapping using existing personas. We wanted to evaluate our users’ mental models and hear them describe their process in their own words. Consistency in language and terminology became a theme. We also performed hybrid card sorts to further understand mental models, language, and priorities. We then created service blueprints for each user flow, mapping out both current and ideal states, including all stakeholders and touch points. After aggregating our insights, we generated design strategy and road maps, sharing frequently with our product and tech partners, and closely tracked our UX backlog as we moved into higher fidelity.

Key Themes

1. Data transformation is a team effort: Tech teams often engage with Product Owners, business partners, and / or consumers early on to gather requirements, and decide on a design approach.

2. Accuracy is key: Changes to the data are usually prompted by one of the following: mismatching datasets, consumers who need better / additional values, or because the first pass was not 100% correct after seeing it in QA / Production.

3. Provide guidance early and often: The majority of challenges and opportunities arise during the Design phase. In order to improve the speed and accuracy, give users a solid understanding of the requirements, as well as comprehensive feedback throughout.

4. Criteria for trustworthiness: Users looked to the freshness of the data, location, source, required and applicable fields, popularity, ownership, and timestamps as indicators of trust.