Einstein Prediction Builder
Competitive Analysis, Heuristic Evaluation, Usability Testing, Micro Guidance, Information Architecture, High Fidelity Mockups, Prototyping
4 weeks
4 UX Designers + Developer
Einstein Prediction Builder is a low-code tool to create custom predictions and deploy models.
Builders are tools that allow users from all backgrounds to create, customize applications and business process. It brings the power of AI to all Salesforce customers. Aimed to serve customers with low AI literacy. Increases team productivity, enables you to make informed marketing predictions, and helps you discover insights.
Adoption to EPB was low and users often abandon the builder before completing a task.
The idea was to improve the overall user experience of the builder. Prior research from the team at salesforce showed that their was a 6% drop-off rate between enabling the builder and deploying the prediction.
To guide non-technical and novice users to complete workflows successfully and build predictions.
Solution overview
Creating a better experience
We worked on a two week long design sprint to understand the reason behind the drop of rate and try to improve the overall experience. After narrowing down the scope of the problem we worked on providing guidance by focusing on three subareas:
  • Promoting user guidance and giving context: creating an introductory interface, showing users potential outcomes of the model and how it could benefit the business. Utilizing the Salesforce mascot (Einstein) to provide guidance with data and terminology.
  • Creating Focus & Progressive disclosure: Segmenting the canvas into 5 parts namely, navigating, progress indicators, working area, assistance and CTA's.
  • Feedback and validation: Providing progress indicators and phase status' through icons and catching errors in data through the data checker.
20% decrease in user task abandonment rate.
30% reduction in errors users committed per task.
Understanding existing user workflows through heuristic evaluation
Upon conducting an initial analysis, we realized that EPB consisted of long and complicated workflows, which led us to conduct a heuristic evaluation and create user flows to map the existing workflows. The key observations were:
1. No clear visibility of system status: The sequence of steps, phases was often unclear which confused novice users. There was also a lack of explanation
2. Easily overlooked information: Users often overlooked the options listed below the fold.
3. Limited cues about next steps: There was no clear indication of progress between each step and sub step.
4. Insufficient knowledge bases: Users were unable to identify the characteristics of a dataset.
Understanding EPB's users
With a focus on understanding AI maturity levels of our users and increasing the intuitiveness of the platform for non technical users, we conducted primary research interviews. We interviewed 5 users with a non technical background to understand their pain points and how they used EPB. The key takeaway was that users often left tasks due to confusion.
Defining opportunity areas
We distilled our findings and narrowed our problem scope based on impact vs effort matrix into 3 main problems.
  1. Increasing user knowledge bases: Helping users identify characteristics of a dataset through visual guidance.
  2. Educating non technical users to accurately segmenting data thus increasing the number of predictions deployed.
  3. Indicating progress:  There are numerous subsets inside each phase, making it easier for beginners to become confused.
We came up with two design principles
  1. Visual Learning: Making the experience of EPB more scannable visuals that are more engaging.
  2. Clarity: Salesforce advocates for clarity in their Lightning Design system. Reduce confusion for low AI maturity users.
Target Group: Focusing on the underserved.  
Out of the salesforce persona repository we decided to focus on our underserved users, i.e. Biz Ops interpreters. They interacted the most with our software and were the least data aware of our user group.
Feature #1
Creating context
Lillian, a first time user, explores what EPB is about through the Introductory Page. She is able to see potential outcomes EPB can bring to her work and further, benefit the business. It allows her to visually learn about the canvas.
Walkthroughs take her through the EPB’s canvas which she can access and dismiss at her own convenience.
Feature #2
Progressive disclosure & creating focus on the canvas
Lillian, comes to a 3-part canvas with clear segmentation of navigation, working area and assistance. Having clarity around the working area helps her focus on one task at hand.
Feature #3
Staying in loop through progress indicators
Lillian, is able to navigate easily through each phase in EPB through the Navigation Tree. Phase statuses (icons) make her aware of prediction progress and also calls attention to any errors along the way.
Feature #4
Providing user guidance through assistant
Lillian can always see the guidance panel on the right part of the page to better understand the process and terms. If she has more questions, she can leverage the resource from the trailblazer community.
Feature #5
Real time data checker to catch errors early
Lillian can always see the Data checker on the right part of the page to better understand the system progress, data status and how to recover from errors.
Identifying outcomes and impact
We evaluated our designs through usability testing and surveys. The results were:
Decrease in users Task Abandonment Rate
Of users were satisfied with the new design
30 %
Reduction in errors users committed per task
If I could take another go at this project
  • Improve the visual language.
  • A/B Test different elements .
  • Collaborate with developers to further test our solution based on other external conditions.
Reflection and learning outcome
  • Tackling functional complexity and depth within enterprise software.
  • Working on a new domain and gaining expertise.
  • Embracing flexibility and thinking holistically about the impact of our design on the company's vision.
  • Working within existing design system constraints.