APR 2025

Reimagining Data science workflow with Fabric Notebooks

A notebook is core feature in Fabric. It's an interactive coding tool where data scientists write code, experiment, and collaborate all in one place, making it easy to explore data and share insights.

My Role

As the design lead for Microsoft Fabric’s developer experience, I, along with a talented global team of designers, researchers, product managers, and engineers, set out to revolutionize how data professionals interact with their analytics tools. My mission: make data science as frictionless, collaborative, and powerful as possible-all while driving tangible business value for Microsoft Fabric.
(Note that all features introduced below has been released in product)

The Challenge: Coding UX Were Broken

USER RESEARCH

Method: 60 minute 1-on-1 sessions of usability testing coding products.
Participants: 4 data engineers and 4 data scientists.
Theme: Semi-structured interviews and allows free explorations.

We conducted user research that included interviews, testing, and direct observation of data scientists and engineers in their daily workflows. By listening to their experiences and watching how they interacted with coding tools, we uncovered common pain points such as slow performance, complicated setup processes, difficulties with collaboration, and concerns about data security.

Data scientists and engineers faced a minefield of obstacles:
•   Setup Nightmares: Spend hours install languages configuring environments in their local machines.
•   Lag & Frustration: Slow performance for handling large scale data.
•   Collaboration Chaos: Sharing work meant endless email threads and version confusion.
‍•   Need Help with Coding: Spend hours searching for coding solutions on GitHub and developer forums.

Our research was clear: Users spent majority of their time fighting tools instead of finding insights. We knew we could do better.

Our Vision

Collaborating with the product leaders, we aligned our goals: make Fabric notebook a go-to coding tool for every data project.

Data scientists and engineers faced a minefield of obstacles:
•   Zero-friction onboarding: get to learn the core workflow in minutes, not hours.
•   Stellar performance: no more waiting, just coding and running with built-in features to power innovation.
•   Real-time collaboration: coding together seamlessly and securely.
‍•   Get help from AI: Get help for coding and debugging from AI directly in notebook, answers within seconds.

Frictionless Onboarding

USER Feedback

We started collecting feedback from analytics tools and conducting empathy-interviews. One of the most frequently received feedback was that developers have a hard time familiarizing with the notebook interface and understanding what new features have been released.  

First-time onboarding tour

The initial experience a customer has with a product is vital. The aim of the first-run experience (FRE) is to ensure users quickly understand the product's value. Direct them simply and clearly to their starting point, minimizing obstacles and unnecessary friction for a smooth and rewarding beginning.

New feature tour

Properly introducing new features helps keep users aware of updates and boost feature adoptions. Use non-intrusive methods like in-context prompts or tooltips that fit seamlessly into the user's workflow. Present features when they are most relevant to boost engagement and adoption.

Easy Setup and Lightening performance

USER RESEARCH

Method: Benchmark think-aloud interviews, 90-minute 1:1 sessions.
Participants: 12 data engineers and 12 data scientists.
Theme: End to end data engineering benchmark study.

We conducted user interviews that focuses on first-time users for Fabric who had experience with other online coding tools. In the research, each participant was assigned a list of routine tasks for data engineering workflow, the tasks includes getting data, transforming data, and visualizing data. The participants were asked to performed the same tasks on Fabric and other tools with a random order. The task success were recorded in both behavioral measures (success rate and completion time) and attitudinal measures (how difficult the users feel when working on Fabric vs. other tools. High-level paint points are shared below:

Difficult Setup

To start coding, data engineers and scientists must set up a proper environment. It takes hours for every data projects to install all their preferred languages, libraries, and properties in the right version.

Slow Performance

To run codes, the market standards is to take around 15 minutes to start a session. It is perceived as "too slow", and one of the reason data scientists do not prefer web solutions for programming.

High Dependency

To transform or visualize data, data engineers and scientists need to rely on public libraries, which are sometimes laggy and unreliable. Advanced tooling usually adds up the costs of operations.

Lightening-fast start up

Our engineers had optimized backend performance, and my team focused on enabling the power of "defaults" - creating intuitive default settings that cater to the majority of our users' needs. This approach eliminated unnecessary setup steps for new notebooks.
Now, users can seamlessly choose their preferred programming languages and select a pre-built, optimized environment from our curated list, all with just a few clicks. They don't have to install anything on their computer or do any configurations. Notebook takes care of it at all.
This enhancement reduced the initial setup-to-code time from 15 minutes to under 5 seconds, dramatically boosting user productivity and workflow efficiency.

Transforming and analyzing data with built-in visualizations

In leading the design of the Fabric Notebook, my team prioritized reducing repetitive work by incorporating UI gestures. This approach eliminates the need for complex code configuration, as the notebook automates code generation. Users simply drag a table into the code editor, run it, and our built-in visualization tools—enhanced with AI recommendations—create charts and reports effortlessly.
Our visualization tools and AI recommendations have garnered praise for being available out-of-the-box, requiring no additional library installations. This achievement highlights my ability to blend innovative design with technology, significantly enhancing user experience and efficiency.

USER RESEARCH

Method: Competition benchmark, 90-minute 1:1 sessions.
Participants: 12 data engineers and 12 data scientists.
Theme: End to end data engineering benchmark study.

As a result, the Fabric notebook has the highest outcomes compared to major market competitors.

  • Fabric notebook: Task Success 84%, Attitudinal 77%
  • Competitor 1: Task Success 80%, Attitudinal 76%
  • Competitor 2: Task Success 71%, Attitudinal 68%

secure real-time collaboration by design

USER RESEARCH

Method: Interview and wireframe testing, 90-min 1:1 sessions.
Participants: 9 participants across data engineers, scientists, and analysts with various coding experience.
Theme: Notebook collaboration testing.

Users love collaborating in coding, especially when they can see what others are working on in real-time. However, the lack of supporting features concern users. When coding on the same project, users expressed concerns about changing other's work accidentally, going back and forth with questions, exposing sensitive data to collaborators who should not have permissions.

Real-time collaboration supercharged

Real-time collaboration and comments in Fabric notebooks supercharge teamwork by letting multiple people code in different languages, comment, and brainstorm together instantly. With live updates and in-line feedback, teams can solve problems faster, share insights on the fly, and keep everyone in sync-turning data projects into a truly dynamic, collective effort.
With comments, notebook lets users leave conversations right on specific code cells, making teamwork fast and focused. Tag teammates, spark discussions, and solve problems-all without ever leaving your notebook. It’s real-time collaboration, supercharged!

USER FEEDBACK

Method: In-product feedback tool

Through our feedback tool, users expressed one concern for real-time collaboration feature - afraid of irreversible changes to others' work. This stops them of freely collaborating.

Version history safeguarded

Understanding the pain points, my team collaborated on robust built-in version control, automatically saving checkpoints and allowing manual checkpoints to mark key milestones. This not only safeguarded users' progress but also enabled them to track the evolution of my work, compare different versions through a diff view, and restore previous iterations when needed. Furthermore, for more advanced collaboration, users can leverage Fabric’s Git integration , which allows them to back up and version notebooks, revert to previous stages, and manage content lifecycle directly within the Fabric environment.

AI assisted Coding

USER RESEARCH

Method: Survey study taking around 12 minutes
Participants: 183 data professionals

Our research team sent out a survey to attendees for Fabric Conference in Las Vegas. The survey attempted to capture insights on changes in the landscape of AI tools utilization, and how data professionals are using AI in their daily work.

As a result, 87% of all respondents have low to intermediate levels of familiarity with A tools, but 90% of respondents are willing to adopt AI tools into their data workflows. Across all surveyed AI tools, ChatGPT is rated the most common and most helpful, followed by Microsoft Copilot. Two major concerns of frequently integrating AI into data workflow: 1) data security - not trusted to share data with AI tools, and 2) inefficiency to provide context to AI tools.

It gives us an advantage to build the AI-assistant tool inside of notebook, because the data are not shared outside of Fabric, and notebook automatically understands the context of all content.

AI assisted coding with Notebook Copilot

Copilot is your best partner in working with data. Copilot in Fabric notebooks offers an interactive chat panel and in-cell AI assistance, enabling users to generate code, fix errors, add comments, and optimize performance directly within their workflow. By simply typing natural language questions-such as asking for data visualizations, code snippets for machine learning models, or explanations of complex code-Copilot responds with relevant code or insights, tailored to the context of the notebook’s data and structure.

results

The features introduced above has been released to public users and helped Fabric notebook growth.  A few key results showing early success:

Steady growth

Since the start of 2025, Fabric notebook's month active users has grown by 67%.

Increased customers

By the end of March 2025, Fabric announced that the customer base has grown 75% in the last year, from 11000 to over 19000.

Healthy retention

By the end of April, 82% of Fabric notebook users are continuously active.

Better execution

Starting a session in a default environment is reported very fast - often start within 3 to 5 seconds, because of optimization.

Growing usage

Since the start of 2025, the computing resource consumed by notebook users increased by 68%.

Trusted AI

Since the start of 2025, Copilot usage in notebook has increased by 148%, remaining from the developer community.

✨ Reflection

Leading the redesign of Fabric Notebooks was a defining experience in my journey as a design leader. This project reinforced my belief that transformative user experiences are built at the intersection of empathy, creativity, and strategic vision. By immersing myself and my team in the daily realities of data professionals, we moved beyond surface-level fixes and tackled the root causes of their frustrations.

Most importantly, this project taught me that great design is not just about solving existing problems-it’s about unlocking potential and pushing the boundaries. By removing barriers and empowering users, we didn’t just improve a product; we helped shape how organizations turn data into impact. I’m proud of what we achieved and even more excited for the innovations ahead

Back to top