My Roles -
For -
With -
Timeline -
1 Year
Jul. 2021 - Jun. 2022
overview.
Self-Analytics Tools
I worked as a UX designer at Ant Group, Alibaba, from July 2021 to July 2022. My primary focus was leading the end-to-end design of a core self-analytics tool and scaling its design framework to 5 additional tools within the product ecosystem. These tools empowered 5,000+ internal, multi-functional data consumers to independently perform data queries and visual analytics, enabling timely insights while significantly reducing their reliance on BI specialists.
Watch the video to gain a quick glance!
Achievements
Process..
Problem.
Data consumers struggled with limited data analysis flexibility
At Ant Group, non-technical data consumers, such as marketing specialists and business development specialists, face challenges in accessing their work-relevant data or conducting analyses due to limited self-analytics support within existing reports. Such heavy reliance not only delays decision-making for report consumers but also places a significant burden on BI specialists. Such repetitive analysis tasks consumes time and resources, preventing BI specialists from focusing on high-value work.
Problem statement.
How might we support internal data consumers to independently and efficiently access and analyze work-relevant data, thus reducing reliance on BI support?
user research.
Interviews with pioneer user groups
I collaborated with product managers to conduct focus group interviews with 3 BI Specialists to gain insights into their current data support workflow, and typical tasks. Additionally, we interviewed a pioneer data consumer user group, comprising 2 marketing specialists, 2 business development experts, and 2 product managers to understand their current experience, pain points, and expectations for self-analytics.
user profiles.
What's the profile of our primary users?
While BI specialists will also use with self-analytic tools for ad-hoc analysis, our primary user group was data consumers - over 80% of whom are non-technical users with limited data literacy. We decided to design for the lowest technical level to ensure accessibility for all user groups and maximize the product impact. To clarify user profiles and guide future design decisions, I developed the following personas to represent our target users:
Key Findings.
Through stakeholder interviews, we’ve learned the causes behind such reliance, and the gap between the current data supporting tool, (data e.g., reports/dashboards, the internal report building tool), and the requirements from data consumers.
By diving deeper into those repetitive data analysis tasks, we’ve learned that most of the tasks were around metric insights, which helped us prioritize the function scope of the foundational self-analytic tool.
job to be done.
I worked with the PMs to develop a new workflow that empowers data consumers to independently gain metric insights through metric access, filtering, breakdown, visualization, etc, without relying on repetitive BI support. We also incorporated key design opportunities identified from user research, competitor analysis, and project team discussions, into the workflow.
Design direction.
Reflecting on the original problem statement “How might we support internal data consumers to independently and efficiently access and analyze work-relevant data, thus reducing reliance on BI support?”, through alignment with product managers, I refined the direction into:
Design a no-code self-analytic tool that prioritizes ease of use for novices and empowers data consumers with limited data literacy to explore metric insights confidently and efficiently.
Design Strategy.
🎯 Develop design strategies & align to roadmaps
Through synthesizing insights from user research, competitor analysis, and industry best practices, I established four key design strategies to guide the development of a self-analytics tool that meets the needs of non-technical users. I also prioritized the design focus and ensured their alignment with the product roadmap.
the challenge.
Challenge1: Simple but scalable tool structure?
The primary challenge during the MVP phase was designing a tool structure that was simple and intuitive for non-technical, novice users to ensure task success from their first-time use, while also supporting scalability to flexible, advanced query configurations, and maintain efficiency.
Competitor analysis.
Learn from industry practices
I researched relevant products, whitepaper for self-analytics, from which I summarized the UX patterns and design practices.
design concepts.
Ideate diverse core tool structures
To quickly get stakeholder feedback, I created wireframes to showcase different tool structures:
Concept comparison.
🤝 Let's go back to stakeholders for feeedback
To evaluate the concepts from multiple perspectives, I facilitated 2 stakeholder feedback sessions, one with a pioneer group of data consumers, and another with BI specialists and PMs. To ensure structured and systematic comparison, I gathered feedback on 3 key criteria: 1) novice-friendliness, 2) interaction efficiency, and 3) scalability for handling complex and advanced configurations.
The synthesized results are outlined below, with key stakeholder feedback highlighted for context:
refined concept.
Enhance the concept 3 with a balk add feature
Based on stakeholder feedback, I chose the top panel configuration as the core structure due to its lightweight and guided design, offering a more novice-friendly experience than the left panel. Additionally, we incorporated the bulk add feature for 1) novice-friendliness: Enable novices to quickly glance at data and gain confidence in their selections, and 2) query efficiency: Allow users to efficiently add multiple options at a time.
Usability testing.
Created clickable prototypes for usability testing
I created clickable prototypes using a typical analysis case, and conducted task-based usability tests with 4 users to identify potential usability issues.
Key iteration.
Enhance the empty state to guide new users
Key iteration.
Provide a collapsed mode to enhance clarity
final design > Query
🌼 Final Design: Guided, linear query experience
The guided, linear layout breaks down query building into clear steps, effectively reducing the learning curve and confusion, enabling non-technical new users to start queries from first time use.
final design > Query
Clear and efficient data selection
Illustrated tooltips are provided while the user explores the configurations, helping non-technical users better understand configuration terminology.
final design > Query
Novice-friendly yet scalable configs
To ease the learning curve for novices, the config panel prioritizes essential features for users, while advanced options (e.g., custom metrics and multi-layer filters) appear while hovering.
final design > Query
Collapsed mode for enhanced display clarity
Collapsed mode can be activated to provide clear reference to query configurations while maximizing space for results. This reduces information clutter and enhances efficiency in interpreting outcomes.
the challenge.
Challenge2: The default chart display can be messy… How to ensure the accessibility of visual analytics for any data query?
The major challenge we faced in the second phase was how to automatically generate clear and meaningful visualizations that adapt to diverse types of data queries, minimizing user customization efforts to ensure accessibility for users with limited data literacy.
💡 Can we add a "filtering effect" to the charts?
ideate.
🫶 Tackle the technical constraints with developers
Collaborating with developers, we came up with a feasible tech framework that can incorporate best design recommendations into the code to optimize the default chart specifications based on the data input, and then adaptively generate more accessible charts for users to look at. Additionally, we also decided to introduce the Chart Advice feature to educate the user how to interactively analyze the graph and if they need to change to a more appropriate chart type.
💡 Rules library for automatic visualization optimization
implementation.
Chart Optimization and Chart Advisor
By implementing the optimization rules, we refined the default visualizations to be more accessible. Additionally, the Chart Advisor was introduced to provide context-relevant suggestions on visual analytics, such as changing to a more suitable chart, using interactive analytics, etc.
refine.
A more accessible and adaptive design of data label list
While the rules I established improved visualization quality, post-MVP feedback highlighted challenges with interpreting and interacting with data labels. To address these, I refined the design, as shown below:
final design > results
🌼 Final Design: Adaptive, accessible data visualizations
the challenge.
Challenge3: How might we provide just-in-time smart support that users really need?
As we progressed to developing smart analytic features in the third phase, a critical challenge emerged: exploratory data analysis flows are inherently non-linear and flexible. This raised two key questions:
1. How can we provide contextual support that users really need, without disrupting users’ workflows?
2. Given the maturity level of smart analysis technology, how can we ensure the relevance and effectiveness of such smart support?
Scoping.
💡 Let's first scope down and focus on a typical use case!
Given the broad and ambiguous nature of the design challenge, I discussed with the PM, to first focus on a common, impactful use case for smart analytics: identifying root causes of abnormal fluctuations in line charts, and utilize this as a starting point to explore the pattern for self-analytics. Together, we pinpointed pain points at each stage of the process and identified opportunities where smart analytics could deliver contextual, real-time support.
user flow.
Co-create a feasible and effective smart analysis flow
To empower the current time-consuming, ineffective workflow, I collaborated closely with data scientists, product managers, and front-end developers to design a feasible and seamless workflow that integrates automated attribution analysis to empower the user’s existing analysis process.
To ensure the relevance, effectiveness, and technical feasibility of the smart support, we collaboratively refined the workflow by incorporating user control mechanisms. This balanced approach empowers users with relevant insights while maintaining efficiency in the analysis process.
final design > smart analytics
Intuitively access smart analytic features within the tooltip
Easily customize settings step-by-step
Clearly track the generation status
Effortlessly consume insights of fluctuation causes
Efficiently export to a report for sharing
Once the analysis is completed, the user can export to as a single report for saving, further editing, and sharing.
Design Guide.
🥳 Create a comprehensive design guide to ensure future consistency and streamline the design-to-implementation process
Throughout the end-to-end design process for the benchmark tool Metric Insights, I also oversaw and aligned the design efforts for other self-analytic tools in parallel, such as User Distribution, User Retention, and Attribution. These tools, in their MVP stage, were primarily supported by design contractors and product managers.
After validating the design patterns in Metric Insights, I developed a comprehensive design framework and strategies, along with detailed specifications for updated components. This framework enabled design collaborators, PMs, and developers to deliver a consistent and cohesive user experience across all self-analytic tools, both existing and future.
Product success.
🙋🏻♀️ Successful non-technical user adoptions
The self-analytic tool successfully onboarded 8,000+ non-technical users out of 12,000+ total users, showing its effectiveness in empowering non-technical users to independently handle basic analysis tasks.
💪 Significant efficiency improvement
As highlighted in our department meeting, the BI team estimated that the introduction of self-analytic tools saved approximately 150 workdays of BI effort daily. Simultaneously, data consumers experienced a significant increase in efficiency as waiting times for data support were substantially reduced.
😄 User satisfaction improvement
By addressing user pain points and improving usability, the tool enhanced the overall satisfaction level, particularly in the interactive analytics functionality.
Note: This is a rough comparison since we didn’t have an accurate baseline.
