Hi-fi prototype + interactive prototypes + user flows + usability study + design led A/B testing
@Microsoft
Figma, Copilot, userzoom
Product designer
Title
Time
My main contribution
Company
Tool
2023 Feb - 2025 Feb
Design a new Linux workload creation experience
What did I do?
I worked as product designer in Azure core
I built on Azure portal: a cloud provider interface used by 200 million active users
I designed latest experience on scenario-based workload creation called Executable doc
I designed
Project background
Why building Linux workload creation
Azure architecture comprises two principal systems: Linux and Windows
70%
User running cloud using Linux
15%
Service are related to Linux on Azure
70% of potential customers can only use 15% Azure service
To achieve growth
To convert more new users: we need a service support Linux, providing an easy on boarding experience and building Linux ecosystem.
Project initiative
How to better serve our Linux users
User painpoint
Multi-cloud users and newcomers to Azure often spend hours navigating extensive documentation and independently learning and testing workloads. As a result, they often end up feeling confused and frustrated throughout the process.
Business incentive
“Azure is widely known for Windows-based deployments. To demonstrate its versatility as a comprehensive development platform, we aim to showcase available options and offer tailored workloads for Linux users, driving product growth.”
Personas
Who are we designing for?
Previous experience
This is what currently on boarding looks like
⚠️
Reading through countless documents can be overwhelming.
Learning without hands-on experience in deploying workloads makes it even more challenging.
Competitive study
How well did our competitor do?
Most competitors lack workload-based deployment during onboarding
However, they outperform Azure in educating users by providing full context and actionable deployment guidance
Generative study
We research on what's the determining factor for user retention rate
When a new user deployed a task within their first 20 minutes in Azure portal
+12%
User conversion rate
+25%
More satisfied with the experience
+30%
User learn about Azure better
Design goals
What to improve with design
Learn as they go
Users can deploy workloads when they are learning about Azure
Users can find the right packaged workload based on their needs
Fast deploy
Break a long deployment into steps -- easy to learn and process the information
Provide summarized documentation on each step as they go
From MVP to North Star
Design iteration in phases
Product development was optimized by structuring design in phases.
For successful rollout, our approach needs to establish an educational journey from start to finish.
The users begin by understanding Azure's potential, and then transition to ideating based on their requirements.
We've constructed a web page:
Selecting workloads
Initiating deployments
Confirming successful deployment
From MVP to North Star
Selecting workload page
This page allows users to explore what Azure offers and find the best match for their needs.
Users can quickly preview and understand all available options presented in a grid format.
Deployment selection page
From MVP to North Star
Crucial milestones on deployment page
Due to restrictions on the engineering team, the deployment page was designed and rolled out gradually, based on technical feasibility. We shipped it in phases to gather quick feedback from users.
In this stage, we aim to achieve one-click deployment, aligning with the ‘fast deploy’ design requirement.
Deployment page first version (shipped 2023)
One click deployment
Hover over to view details
Usability study
Quick feedbacks from users
We conducted another round of usability study with first version of deployment page, using user feedbacks to lead our design decisions
User feedback
❤️ They responded positively to seeing the terminal and quickly deploying workloads.
⚠️ They want a clearer understanding of the process, greater transparency, and more control over the deployment
In this stage, after implementing one-click deployment, we focus on guiding users through the process by breaking down the workload into clear steps.
Deployment page second version (shipped 2024)
Step by step deployment
Hover over to view details
From MVP to North Star
Deployment page North Star version
North Star version combined previous two versions and feedbacks from the research study
From MVP to North Star
Features break down
Here is design reasoning on each component in North Star design. It was inspired from the previous study and tested with our potential customers.
Left ToC
Search
Basics
Create a resource group
Create a virtual machine
Enable Azure AD Login for a Linux virtual machine in Azure
Store IP address of VM in order to SSH
SSH into VM
Users can use ToC to quickly navigate to the section they would like to see
Users can learn an overview of steps how to create this workload, a high level view or index
Usability testing I conducted indicating
users with previous different cloud provider experience are only interest in seeing that Azure is doing differently or some step with complex deployment to learn how to our expert code certain sections
Code block
Usability testing I conducted indicating
users want to practice when they’re learning the new workflow or product. This keep them engaging and showing full transparency of how our expert deploy the workload
Create a resource group
A resource group is a container for related resources. All resources must be placed in a resource group. The azure group command creates a resource group with the previously defined $MY_RESOURCE_GROUP_NAME and $REGION parameters.
Resource group name
myVMResourceGroup$RANDOM_ID
export RANDOM_ID="$(openssl rand -hex 3)"
export MY_RESOURCE_GROUP_NAME="myVMResourceGroup$RANDOM_ID"
export REGION=EastUS
az group create --name $MY_RESOURCE_GROUP_NAME --location $REGION
Run this step
Users can run the code in the code block live on the cloud shell, see the result in terminal and configure costumed parameters as they need to.
Users can learn how each step are done with descriptions in natural language with the code block (the best way to learn how to code than just reading the document)
Users can use run all to deploy the whole workload within one click, which will save them time without learning for hours to get started. They will be able to use all preconfigure data that will support them having a working workloads.
Our generative study shows
users are 15% more likely to stay and use the platform longterm if they can deploy something on the platform within the first 15 minutes they’re on the platform
Action footer
Run all
Get the source files
Give feedback
Terminal
Users can use terminal to trouble shoot real time
Users can learn what has been done/deploy on the steps and learn how long each step will take and what’s the terminal output
Usability testing I conducted indicating
Users like learning model as Jupiter note book. A real life feedbacks on what they deploy and that will help the learning and better sense of control
Requesting a Cloud Shell.Succeeded.
Connecting terminal...
Welcome to Azure Cloud Shell
Type "az" to use Azure CLI
Type "help" to learn about Cloud Shell
Your Cloud Shell session will be ephemeral so no files or system changes will persist beyond your current session.
Connie [ ~ ]$
Results and metrics
How did me measure success
In our earlier experimentation shown good conversions 5.2% bump which could translate in 18M LTV(lifetime value) for Azure
“This was really easy to use, and I even picked up some new tricks to make my deployments more efficient.”
— A happy customer at study
Next project
Stay in touch. I will follow you back, promise
