Data Integration Framework

Designing a scalable, extensible data integration ecosystem to bring external and internal data.

Product Design Leadership

Data Integration Framework Overview

Context and Problem

Customers had data everywhere. Prism could only see some of it.

Workday Prism Analytics is a powerful platform for financial and HR reporting — but its value depends on the richness of the data it can access. Organizations were storing critical data in cloud warehouses, SaaS platforms, and legacy file systems that Prism had no native pathway to reach.

The result: customers were forced to build custom ETL pipelines or rely on brittle Studio integrations to get external data into Prism. This created maintenance burden, slowed time-to-insight, and limited adoption among data-forward customers.


Data Integration Journey

Data Integration Journey showing the flow from Load Data to Prepare Data, including steps like Create Schema, Set up data load configuration, and Data Load and Error corrections

End-to-end data integration journey: From identifying data sources to monitoring activities


Goals

Existing data integration framework lacked flexibility, transparency and had performance issues. Customers were bringing in high volume and high frequency data which demanded a revamp of the current architecture for data containers.

Our goal was to design the end-to-end experience for a self service data integration framework that will enhance data management capabilities by providing:

Complete Data Transparency

Full visibility into data flows and transformations

Data Validation

Robust validation to ensure data quality and integrity

Powerful Schema Management

Flexible schema handling for diverse data sources

Fine-Grained Access Control

Granular permissions and security at every level

Comprehensive Monitoring

Real-time visibility into integration health and status


Design Approach

A framework, not just features.

Rather than designing one-off connector flows, I advocated for a unified Data Integration Framework — a shared UX model, connection management system, and design pattern library that could scale across dozens of future connectors without bespoke design each time.

This meant working upstream with engineering to define the DCT (Data Change Task) framework as a reusable scaffold, and establishing UX conventions for credential management, connection testing, field mapping, and error states that would apply across every connector type.

Key Design Decisions

Principles that shaped the framework.

1 Framework-first thinking

Every pattern — credential forms, connection testing, error messaging — was designed as a reusable component, not a one-off. This cut design time for Phase 4's four connectors significantly.

2 Security integrated in UX

Credential handling, OAuth flows, and IAM configuration were designed in parallel with security review ensuring secure patterns were user-friendly — not bolted on.

3 Consistent experience across every connector

Design scales from 6 to 10+ connectors without visual degradation, groups related connectors by type, and uses familiar platform iconography to reduce cognitive load.

System Thinking: Object Modelling

By focusing on Object-Oriented UX (OOUX) principles, we shift the conversation away from individual screens and toward a shared understanding of core objects, their attributes, and their complex interdependencies. This approach is inherently superior because it establishes consistent patterns and logical relationships early in the process, preventing the "drift" that often occurs when designers focus on aesthetics while engineers focus on data structures. Ultimately, this systemic alignment reduces technical debt and ensures that the final product is as intuitive to navigate as it is robust to build.

Object Modeling Diagram showing system relationships between Connection, Configuration, Schedule, Base Datasets, Base Tables, Derived Datasets, and Activities
Click to zoom

Object Modeling Diagram: System view showing how the new connection object interacts with other catalog objects — click to enlarge


My Role

UX Strategy & Design Leadership

  • Led UX efforts in defining the strategy and roadmap for data integration experience
  • After Phase 1 launch, transitioned into management
  • Hired and trained designers to work on subsequent phases of the roadmap
  • Team co-led several customer workshops along with research team
  • Provided design deliverables and worked with the scrum team until final launch

Design Process

MVP - Phase 1

Competitive Analysis
Analyzed competitive landscape to understand industry best practices and identify opportunities for differentiation in the data integration space.
Competitive analysis comparing data integration solutions across the industry

Competitive landscape analysis of data integration platforms

Workshops
Conducted a Design Sprint with cross-functional team members with the goal of understanding and providing better problem framing. We collected different ideas from the group and converged to create high-level concept designs.

Cross-functional design sprint workshops with stakeholders

Concept Design
Created concept designs to outline the flow for bringing in data with flexibility and provide a self-service experience to our customers without relying on IT teams.

Concept walkthrough demonstrating the self-service data integration flow

Design Focus Areas
Our design efforts were broken down into creating experiences for 3 key areas:
1 Creating Tables from Various Sources

Tables are foundational data containers that serve as a central location for data from multiple sources with schema definition and data management capabilities.

2 Data Operations: Data Change Task

Provide the flexibility to ingest data incrementally in an automated fashion with a reusable framework for Table Data Change operations such as Append, Replace, Delete. Define and map fields that are needed.

3 Connections & System Integration

Create reusable connections and activity monitoring, integrating with the data catalog.

DCT Overall Flow: End-to-end data change task workflow
Click to zoom

Data Change Task: Complete stepper workflow overview — click to enlarge

User Feedback & Iterative Designs
Conducted multiple rounds of user research and usability testing to validate design decisions and iterate based on feedback from customers.
User Empathy Map for Data Integration showing Says, Thinks, Does, and Feels quadrants for Nina - HRIT Analyst

User Empathy Map: Understanding the HRIT Analyst persona and their pain points


Design Improvements

Key UX optimizations driven by user feedback, focusing on workflow efficiency, cognitive load reduction, and system coherence.

Improvement 01

Unified Mapping, Data Types & Smart Suggestions

Before
  • Mapping fields and converting data types were two separate steps
  • Users had to move back and forth between screens
  • Auto-mapping algorithm created duplicate mappings due to hierarchical naming
  • Data preview panel consumed valuable mapping space
  • No intelligent assistance for mapping decisions
After
  • Combined mapping + data type conversion into a single unified workflow
  • Switched from fuzzy hierarchical to exact mapping logic
  • Added ML-powered field mapping recommendations with confidence scores
  • Moved data preview to expandable bottom panel
  • Expanded mapping workspace for better visibility
Faster configuration time, fewer navigation errors, reduced duplicate mappings, improved user confidence.
Unified mapping with inline data type conversion
1 Inline data type conversion
2 Source-to-target mapping
3 Data preview at bottom
1
Inline Data Type Conversion Users can adjust data types directly in the mapping row without navigating to a separate screen.
2
Source-to-Target Field Mapping Unified view shows source fields mapped to target fields with type indicators.
3
Expandable Data Preview Sample data moved to bottom panel, freeing up mapping workspace.
ML-powered smart mapping suggestions
1 Smart suggestions panel
2 Confidence scores
1
ML-Powered Suggestions Panel Intelligent recommendations appear contextually based on field names and data patterns.
2
Exact Matching with Confidence Switched from fuzzy to exact matching logic, showing match confidence percentages.
Improvement 02

Schema Changes with Existing Data

Before
  • Manual, multi-step workaround required for schema changes
  • Deleting or modifying fields required recreating tables
  • High operational friction and risk of data inconsistencies
After
  • Introduced Alter Table capability
  • Users can modify table schema even when data exists
  • Eliminated manual workaround processes
  • Reduced engineering dependency
Greater flexibility, faster iteration cycles, lower operational friction.
Create Table using source schema
1 Create from source schema
2 Tables with metadata
1
Create Table Using Source Schema One-click action to generate table structure from source data.
2
Tables with Metadata View existing tables with last modified and data change timestamps.
Schema editing with existing data
1 Add Field
2 Editable field types
1
Add Field Capability Add new fields to schema even with existing data.
2
Editable Field Properties Modify field types, required status, and default values inline.
Improvement 03

Load & Transform Workflow

Before
  • Loading data and transforming data were completely separate stages
  • Transformations required moving to a different step
  • No lightweight manipulation during load stage
After
  • Enabled lightweight transformations during load (filters, custom calculations)
  • Users can perform quick data shaping inline
  • Reduced need for separate transformation passes
Streamlined pipeline setup, fewer steps, improved efficiency.
Inline filtering during data load
1 Filter step in workflow
2 AND/OR logic builder
1
Integrated Filter Step Filter data during load without switching to a separate transformation tool.
2
Visual Filter Builder Construct complex AND/OR filter conditions with a drag-and-drop interface.
Custom calculations during load
1 Inline calculation builder
1
Inline Calculation Builder Create calculated fields with CONCAT, math, and string functions without leaving the mapping workflow.
Improvement 04

Connector Selection & Configuration

Before
  • Fragmented connector setup across different interfaces
  • Inconsistent credential and authentication flows
  • No logical grouping of connector types
  • Each connector required bespoke design work
After
  • Unified tile grid with consistent selection experience
  • Logical grouping by type: Workday Delivered, Applications, Data Warehouses
  • Familiar platform iconography reduces cognitive load
  • Framework scales from 6 to 10+ connectors without redesign
Reduced design time for new connectors by 70%, consistent user experience across all data sources.
Connector selection interface with grouped tile grid
1 Workday Delivered
2 Applications
3 Data Warehouses
1
Workday Delivered Connectors Native integrations (File Upload, Data Catalog, Workday Reports) with zero external configuration.
2
Third-Party Applications SaaS connectors (Salesforce, ServiceNow, LoanIQ) with OAuth-based authentication.
3
Cloud Data Warehouses Enterprise data platforms (Snowflake, BigQuery, Redshift, Azure) with IAM/key-pair credentials.

Phased Delivery

Building the ecosystem incrementally.

The connector roadmap was structured in phases tied to release milestones, each expanding the surface area while validating the framework's extensibility. Each phase introduced new connector types — from internal Workday data, to file-based, to direct cloud warehouse connectivity.

MVP
FOUNDATION

Internal Connectors — DCT Framework, File-Based & Data Catalog

Established the core DCT architecture and connection management patterns. Introduced the first connectors: internal Workday data flows, file upload ingestion (CSV/Excel), and native Data Catalog integration. Proved the framework could handle diverse connector types from day one.

💎 DCT Framework 📄 File Upload 📚 Data Catalog
P1
PHASE 1

Workday Reports & SFTP

Extended the framework to pull from native Workday Reports (addressing a top enterprise request) and SFTP file servers — a widely used legacy data pathway for regulated industries. Validated credential management flows and secure connection testing UX.

📊 Workday Reports 🔒 SFTP
P2
PHASE 2

Salesforce & Amazon S3

Introduced the first external SaaS and cloud storage connectors. Salesforce addressed the most-requested CRM data source; S3 opened the door to data lake access. Security reviews completed for both new OAuth and IAM credential patterns.

☁️ Salesforce 📦 Amazon S3
P3
PHASE 3

Cloud Data Warehouse Suite

The most ambitious phase — four major cloud data warehouses in a single release cycle. Designs completed and partnered directly with engineering for implementation. Each connector used the mature framework with tailored authentication flows (key-pair for Snowflake, service account for BigQuery, IAM for Redshift, Azure AD for Synapse).

❄️ Snowflake 📊 Google BigQuery 🔴 Amazon Redshift 💎 Azure Synapse

Impact & Outcomes

A platform, not just a product.

The Data Integration Framework transformed Prism from a closed analytics environment into an open, extensible data platform. Customers gained self-serve access to their most critical data sources without engineering intervention.

60% Reduction in schema validation errors
75% Connection reuse rate
3 Products adopted the framework
10+ Connectors shipped across phases