Case Study
2019 – 2020
Data streams and mapping cover image
Project Setup
Business Objectives

Salesforce CDP (Customer Data Platform) is a declarative tool for marketers and analysts to deliver a single, integrated view of the customer. This feature allows users to ingest customer and engagement data into Salesforce CDP and harmonize their schemas into a unified data model.

Role

Lead UX designer responsible for end-to-end user experience from ideation through development for 6 months.

Primary Stakeholders

Product Managers, Developers, Architects, UX Researchers, Content Writers, Product Marketing.

Project Status

Launched in November 2020 as a part of the Salesforce CDP general availability launch.

Project Background

It’s amazing how much data one person can generate. And how many different sources that data can come from. Even a simple shopping trip can generate customer data related to sales messages, web traffic, purchases, preferences, location, and a multitude of other sources. Businesses need to keep all that information organized and accessible so they can gain a more complete understanding of their customers.

Defining the data model can be complex. Customers need to understand what data is collected (and how), the existing data structure, and how that data relates to other sources. There is a huge unfulfilled business need to bring all of that data together into a single, actionable view of the customer.

Challenge
🎯

The primary objective of this project was to create a foundational user experience where data from any source system can be ingested (by creating data streams) and unified (by mapping into a standardized data model) in Salesforce CDP.

Data Streams

Data is retrieved from the source by way of a Connector. Connectors establish trusted connections to the location where your data sources reside, building communication between servers so that your data can be continually accessed. Data Streams assist the connectors in that they dictate how often and when the connections should be established. And they also assist with actually populating the data into the data source object once the connector gains access.

Mapping

Now imagine that all your data is ingested and each data set is speaking its own language. How do we get them all to understand one another? These data sets must all conform to the same universal language to begin interacting with one another. That’s where the second phase, data modeling, comes in. Salesforce CDP utilizes an open-source, collaboratively defined data model known as the Cloud Information Model (CIM) to help with this unification.

Discovery

We kicked off the design process by running a series of design workshops to understand the concepts and technology we planned to build. The goals of these workshops and meetings were to:

1.

Break down product requirements.

2.

Understand the technology and architecture.

3.

Identify design goals.

4.

Document assumptions around personas and user goals.

5.

Create a research plan for requirement and design validation.

Outcomes
User Flows
  • Based on product requirements and technical architecture, we brainstormed, white-boarded, and identified the optimal user flows to accomplish the user tasks.
  • Salesforce has a number of tools in its ecosystem which provide the same functionality for data connections, ingestion, and mapping. We made an innovatory of these experiences and the existing research to plan and incorporate in our first version designs.
Design Goals
  • Make it Easy: The goal was to unlock time-to-value and facilitate ease of use for business persona which was often missing in existing internal toolsets.
  • Built on the Lightning Platform: Salesforce has a rich set of platform tools and a well-documented UI design system. The platform and the design system would give us standardized interactions and components which are well-researched and used everywhere throughout Salesforce, ultimately helping with faster development of the feature.
V1 Designs

After gathering the requirements and design goals, we started the design work by white-boarding and creating low fidelity concept mockups. These quick design wireframes helped me work with the Product Management and Engineering partners to represent the user flows and further question the technical requirements and scope from a user experience point of view.

As mentioned before, the designs were based on standard Salesforce applications heavily lifting standard design system patterns. While doing so, we documented my design assumptions which would be addressed in the design and persona validation phase.

Assumptions
  • This tool will be equally used by business users and analysts / data systems architect personas with the same level of confidence and ease of use.
  • Users are aware of the underlying data model and schema structures from their sources. They can use the knowledge with little-to-no guidance for bringing in and mapping the data into the Cloud Information Model.
View Prototype β†—
Concept Validation

We interviewed 9 participants from qualifying Salesforce companies. The participants were a mix of personas with Admin, Analyst, and Marketer roles. The study included a prototype clickthrough with a set of prompts, questions, and discussion on the feature functionality.

πŸ’¬

β€œ[I’m] fan-boying over this whole process. I’m totally stoked about it. Thanks for developing these awesome solutions that make life better for people like me.” – Crystal Springs Resort

Pros:

Product serves real needs.

  • Customers are experiencing pain points today understanding all of their data, and pulling it together in a way that is accessible for their stakeholders.
  • Overall, the experience was viewed really positively. Customers reported that β€œit’s all very exciting,” and described the product as simple, intuitive and helpful.
πŸ’¬

β€œSo Marketing Cloud is one of those products where… when you're setting things up, admins are very nervous because it can't be wrong, your data has to be right. Even the simplest thing, have to be very careful. Don't want to email someone who opted out. I'm always very very careful with it, very conservative about what I do. So going into it for the very first time, I didn't know the technology (referring to the data modeling) and what I was expecting. But if I'd done it once or I understood tech better, [it] would not has been as overwhelming.” – PayPal

Cons:

Points of confusion arose, which led to some anxiety around the experience.

  • Without full confidence in their understanding of the underlying technical concepts and the big picture of how the steps fit together, customers reported some hesitancy and lack of confidence in using the tool.
  • Mapping task were seen as tedious and time-consuming especially when working on multiple objects. Some customers asked for ability to save progress in a β€œdraft” state.
Iterations
πŸ’‘

How might we build trust with our uses and improve time-to-value for ingesting and unifying data from different source systems.

I.

Mapping is a tedious task and time-consuming.

From the user tests, the most time-consuming task identified was the field mapping step. The reason observed was the lack of familiarity with the CDP data model. The data model has a slight learning curve, and prerequisite knowledge is required even for any user persona using this feature.

To address this issue, we broke the user flow into 2 steps: data ingestion and mapping. With mapping as its own step, we could introduce draft and in progress modes to built user confidence and trust.

II.

Improving time-to-value for Salesforce Sources.

We wanted to further reduce the task completion time by introducing the concept of mapping bundles. A bundle would include standard source objects, attributes, and mappings to the CDP data model. These mappings would be pre-packaged and deployed without user intervention. This functionality would enable users to quickly get started with Salesforce CDP and serve as a learning mechanism for future custom configurations.

Explorations

In addition to the fore-mentioned enhancements, we also uncovered new technical requirements around the shape of the source data described below.

III.

Provide data transformation capabilities during mapping.

When going through real customer datasets, we observed that the attributes sometimes required lightweight transformations, e.g., changing date formats, concatenating fields, trimming values, etc. Such data transformation capabilities would help customers preserve data quality and improve the accuracy of their customer data.

This new requirement made the mapping experience more complex and demanded more UX explorations beyond our first version designs.

1.
‍
Transformation in a table column.
Pros:
  • Incremental patterns to v1 designs saw some positive feedback in the first study.
  • Low development cost.
Cons:
  • Potentially more confusing (especially for the business users).
  • Transformations could also be related to more than 1 source field, leading to a more complex interface and interactions.
2.
‍
Visual mapping experience. – Low-Fi
Pros:
  • Visual interface could be cleaner to read and navigate.
  • Covers complex cases of mapping multiple objects, relationships, and transformations.
Cons:
  • Custom UI components would mean more development time.
  • Requires more UI iterations and alignment with the Salesforce Lightning Design System.
3.
‍
Visual mapping experience. – Hi-Fi

As we continued to iterate and define these visual patterns, we conducted quick internal research with stakeholders and customer reps. The low-fi explorations gained a lot of support and traction. With backing from executives, we decided to spend more time exploring and bringing this concept to life. The explorations included visual and experiential considerations, including iconography, styling, color, and information architecture; while also validating customer use cases and datasets.

Final Designs

We consolidated all of our primary research findings and feedback from the internal studies and reviews to put together the end-to-end prototype for this feature.

I.

Data Streams
Research Feedback

We interviewed 19 participants with a mix of personas with data systems architect and business user roles. The study included small goal-based tasks. The primary evaluation metrics were task completion rate and time on task.

πŸ’¬

β€œThat was... probably one of the best mapping exercises I've done in my life. And I have to do this quite a lot with different packages and products. Very intuitive and user-friendly, very awesome.” – Cadillac Fairview

πŸ’¬

β€œThere's a fine line between technical and simple; the genius is making super-technical things simple. The simpler you can make things while letting you have control. Good balance.” – ThomasArts

Pros:
  • Customers reported that the starter bundle concept was valuable in increasing their speed to value and easing their understanding of the functionality.
  • Data Ingestion saw a 100% completion rate.
  • Data Mapping saw an 85% completion rate. Some users were confused due to unfamiliarity with the mapping concepts.
  • Some participants reported that the visual mapping experience gave them a sense of satisfaction.
  • All users preferred the visual mapper over the tabular experience as it was seen as: Easy to use, Scalable, and Intuitive
Cons:
  • Customers asked for more system intelligence to speed up the custom mapping process.
  • Provide a way for users to see the canonical data model in order to strategize their data mappings.
  • Prioritize the ability for users to preview and select which fields they want to map within data mapping.
  • Almost half of the users had concerns on how the mapping process might work inside their organization β†’ requiring a team to collaborate and be in charge on the data model.
  • Some participants expressed concerns with scale of the UI to support their schemas.
Development & Release

This project was a multi-release development effort with two engineering scrum teams leading the Data Ingestion and Mapper functionality. We had to make several tradeoffs and stack features for supporting incremental development. One of such tradeoffs was full accessibility compliance for the mapper on pilot release. Still, the pilot customer feedback was overwhelmingly positive and helped us plan for future improvements and new functionality.

After the GA (General availability) release in Aug 2020, the design team decided to bring 2 new UX designers to cover each feature individually. I helped with the transition and onboarding, and serve as a consultant on the project.

Since GA, we've implemented mechanisms to measure impact and success of the feature which include:
‍
Β Β Β Β β€’ Metrics - Daily & monthly active users. CSAT from sales teams, implementation partners, and customers.

Other Case Studies:
Calculated Insights β†—