In Fall 2022, we collaborated with Salesforce's Einstein AI Builders team on a university sponsored project to enhance the user experience of Einstein Prediction Builder.
Within 13 weeks, I remodeled the configuration experience by making crucial information that was previously missing or overlooked visible and easily accessible to users.
1 Data Analyst, 1 Product Manager, 3 Designers
UX/UI Design, Heuristic Walkthrough, Wireframe, Prototyping
Einstein Prediction Builder (EPB) is a low-code data analytics tool in Salesforce. Its configuration process, using non-code UIs, allows users such as data analysts and business analysts to quickly create machine learning models within Salesforce. These models can then be seamlessly integrated into various business processes.
Previous interface of Einstein Prediction Builder (EPB)
EPB users often feel confused when they do not receive adequate feedback after performing operations within the system. They also express frustration when error notifications are not easily noticeable, and helpful instructions in general or for addressing these errors are not provided efficiently.
We redesigned the EPB configuration process user interface with a focus on enhancing user experience by improving the visibility of critical information. Additionally, I developed a powerful Data Checker that effectively visualizes operational feedback and error-related information for users.
Redesigned EPB interface with new Data Checker
At the outset of the project, we had no prior experience with EPB and lacked a clear understanding of the problem that needed to be solved. Therefore, we collaborated with our data scientist, Churchil, and conducted a heuristic walkthrough to gain insight into how the product works. This helped us sketched out the current user flow map and identified two usability issues in EPB that needed to be tested later.
User Flow Map that we drew
A challenge with this project was the limited contact with end users. To research and develop concepts, we consulted with internal stakeholders such as product managers and UX researchers and reviewed their research documents. Through this process, we gained insights about user groups and usability issues.
Interview with stakeholders at Salesforce
Review research document from Data Analysists
EPB has diverse user groups with varying levels of expertise in data analysis and business, each with unique goals, occupations, and knowledge about EPB and data analysis. To cater to both novice and expert users, it's important to create a flexible and efficient platform and provide clear explanations of technical jargon.
“I want in-detailed instructions that can guide me through my configuration process on EPB.”
“I want to be able to focus on my configuration process without bothering with useless information.”
Based on a report by a data analyst from Salesforce, there is a significant drop-off rate between users who enable EPB and those who complete the configuration process to build the predictor. Therefore, I plan to prioritize improving the UX of EPB's configuration process to maximize business and design value within a limited timeframe.
I conducted usability testing with two data science students for the configuration process, during which participants verbalized their thoughts and actions and answered interview questions afterward. The testing provided valuable insights into the problems hindering user experience in EPB's configuration process.
Research Goal: Validate assumptions and uncover other potential problems
Participants: Validate assumptions and uncover other potential problems
Primary Tasks:
During testing, users often seek feedback on the success of their actions to build confidence and proceed with tasks. The lack of feedback raises doubts about the product's capability.
“...I want to know if my operation was done. What was the results of it? Did I successfully change the settings of the object that would influence the end result?”
Users seem to abandon using EPB not because of lack of functionality or power, but because they feel overwhelmed by the information presented in the tool. They are unsure about what information to use and how to use it, the impact of their actions, and how to troubleshoot errors.
“...how might we make information in the configuration process of EPB more significant and visible to all types of users in an efficient and flexible way?”
Based on design insights, we decided to redesign EPB's information display to make it more accessible and user-friendly. The end result would enable users to easily find hidden and missing information while being flexible and efficient enough to accommodate both novice and expert users.
To start our design process, we analyzed the existing information architecture of the interface and created new wireframes to address the issue. The previous interface suffered from a cluttered layout, which contributed to the problem. In the new wireframes, I emphasized a clear visual hierarchy to improve information organization and made room for displaying additional information.
I developed high-fidelity screens for the interface, incorporating both “quick examples” and detailed instructions based on feedback from Salesforce design team. This approach caters to both novice and expert users, providing bullet-point instructions for the former and allowing the latter to focus on their tasks by closing instructions and navigation menus.
- Clear navigation bar indicating the actions users will perform
The navigation bar is renamed into the actions that users will perform on each page, making it clear to understand.
- Visualized Data Checker provides operational feedback
The new Data Checker can intuitively and directly show users the status of the dataset they are operating on. Users can also be notified at the first time when some error happens.
- Closable side bar containing instructions and explanations
All “Help” information for current page are put inside a side bar so that users can quickly and easily find the information they want. Expert users can close the side bar to focus on their configuration.
- “Need help?” window to ask questions to consultants
The configuration in EPB is a complex process, and there are always problems that are out of the range that can be covered in the FAQs. So I implement a "Need Help?" window where users can ask questions they encounter to a consultant from Salesforce.
The original Data Checker used by EPB to notify users about errors was not effective, as it was too small and easily overlooked. I redesigned the Data Checker to make it more prominent and informative, effectively serving its intended purpose.
❌ Data Checker provides users with information on each of their operations, but is too small for users to notice.
❌ Some images are used to implicitly tell users about where they are and what they can do, but most users can not really understand it.
✅ The new Data Checker represents the dataset in a visually consistent way to clearly show users where they are.
✅ It also sticks on the screen so that users can easily find it, and displays details on the configuration so far to help users understand the current status of the process.
❌ Data Checker tells users when some error happens, but is too small for users to notice.
❌ Although the Data Checker shows users on what part of the configuration is wrong, it does not provide any suggestions or instructions to help users solve the error.
✅ The new Data Checker notifies users about error with significant visual cues and icons.
✅ It also provides details on why there is an error, and tells users what they can do to solve this error.
Due to project restrictions as a sponsored project by Salesforce and time limitations, I was unable to recruit real EPB users as participants for usability test in a Figma prototype. However, I still wrote a usability test plan which contains the tasks participants would perform, follow-up questions after the test, and key metrics that measures impact and success of the project, and recruit some data science students into the test and see the impact.
Collaborating with product managers and data scientists posed a challenge due to “language” and expertise differences. To address this, I communicated and check-in with them frequently, sharing my research results and updated concepts. This helped me overcome barriers and receive valuable feedback for a successful project outcome.
Due to project restrictions as a sponsored project by Salesforce, I was unable to interview real users of EPB, which was a departure from previous school projects. Despite this, I utilized stakeholder interviews and user testing with data science students to empathize with users. However, I would welcome the opportunity to interview with real users to understand the obstacles they face directly.