The tools were technically capable. The UI made experts feel like beginners.
Oracle Unity Data Platform unified customer data across CX products — Eloqua, Service Cloud, Commerce. The underlying ML and data pipeline were enterprise-grade. But System Administrators had to configure everything by hand-editing raw JSON files. Gavin, a senior admin managing 500+ job configurations, was spending more time debugging JSON syntax than actually managing data.
The competitive angle: every competing CDP had visual configuration. Oracle's advantage was the depth of its data model — Redwood design system, cross-product unification, ML-driven identity resolution. The design challenge was making that depth accessible without a computer science degree. The constraint: a small admin team managing critical data integration for Oracle's largest enterprise clients.
Three problems. Every System Administrator said the same things.
Deep analysis of the legacy MCPS-Manager. Interviews with System Administrators across Oracle's enterprise clients. Competitive review of Segment, Treasure Data, and Salesforce CDP. The findings converged around three problems that the existing tool made structurally worse.
JSON complexity turns routine tasks into error-prone marathons
Creating a single job configuration required hand-editing nested JSON structures, manually verifying syntax, and running test imports to check for errors. A task that should take 10 minutes regularly took 45. New admins needed weeks of training just to become productive.
Cognitive loadNo visual feedback — you only know it's wrong when it fails
Column mapping was entirely textual. Admins mapped source columns to Unity Data columns by typing field names, with no preview of how the mapping would resolve against live schema data. Errors surfaced only at import time — hours after configuration.
No validationDiscovery is impossible at scale
Gavin managed 500+ job configurations stored as individual JSON files. Finding, comparing, or auditing them required custom scripts or manual file browsing. There was no search, no filter, no way to see status at a glance.
Searchability"The old system forced me to become a JSON expert just to manage configurations. I should be focused on which sources to connect and which columns to map — not wrestling with syntax."— Gavin, System Administrator · Research interview, Oracle Unity
Gavin manages 500+ configurations. He thinks in data sources — not JSON.
Research surfaced a clear primary user — not a developer, but a trained System Administrator responsible for connecting Oracle's CX products to enterprise data sources. His mental model was built around sources, columns, and jobs — not the data structures underneath them.
The System Administrator
Manages data integration workflows and job configurations for Oracle Unity. Responsible for connecting sources, mapping columns, monitoring ingestion, and troubleshooting failed imports. Technical — but not a developer.
"I think in terms of 'what data source am I connecting' and 'which columns map to which attributes' — I don't think in terms of JSON keys and nested objects."
What Gavin actually needed
A visual-first interface that matched his mental model: browse sources, drag columns to map them, preview the configuration before saving, and search across all jobs without scripting. Speed and confidence — not raw control over data structures.
The design principle that followed: the interface should speak the language of the administrator, not the data engineer. Every screen was evaluated against Gavin's vocabulary — sources, columns, jobs, runs — not JSON keys or schema objects.
Six phases, one north star — zero JSON for routine work.
The redesign was too large to tackle as a single effort. I structured it into six sequential phases, each one buildable and shippable independently, with each phase reducing JSON dependency at a different point in the admin workflow.
Study the JSON Tool Manager
Deep audit of the legacy MCPS-Manager. Documented every screen, every JSON structure, every pain point. Mapped the complete data flow from source detection through column mapping to ingestion execution. This phase established the baseline — every future design decision was measured against it.
Navigation & Information Architecture
Designed the Admin Console's structural hierarchy — Dashboard → Job Configurations → Configuration Detail → Execution Logs. Created navigation patterns that surfaced the four primary actions (Create, Search, Edit, Delete) without requiring knowledge of the underlying data model.
Data Ingestion Design
Built guided workflows for source selection and connection. Forms and step-by-step wizards replaced free-form JSON editing. Source detection — previously a manual JSON lookup — became a one-click schema discovery from connected data sources.
View, Edit, Create pages
Comprehensive CRUD interfaces following Oracle Redwood style guidelines. Inline editing, confirmation dialogs, and real-time validation. List views with filtering, sorting, and search — giving Gavin visibility across all 500+ configurations for the first time.
Rebrand the Query Builder
Transformed the text-based JSON query editor into a visual query composer. Drag-and-drop column selection, operator dropdowns, and real-time query preview. Admins build complex queries without writing a single line of JSON.
Shell & Integration
Implemented the unified shell and navigation framework. Integrated all six modules into a cohesive Admin Console with consistent styling across every component. The Redwood design system applied end-to-end — from form labels to error states.
Admin Console — visual configuration, drag-and-drop mapping.
The final Admin Console replaced every JSON interaction with a visual equivalent. Three interfaces formed the core of the redesign — each one eliminating a specific category of JSON dependency from Gavin's daily workflow.
Admin Console — Job Configurations overview · Image coming soon
The design decision that changed the engineering contract: source schema detection had to happen live in the browser, not as a separate developer step. This required a new API surface from the backend team — a design requirement that reshaped the data architecture.
Visual column mapper — drag source columns to Unity attributes.
The most impactful single interface. Admins see source columns on the left and Oracle Unity Data attributes on the right. Drag a column to map it. Unmapped columns flag in amber. Conflicts flag in red. The configuration is validated before it is ever saved.
Visual column mapper · Image coming soon
Three outcomes. Measured against the old JSON workflow.
"With the new console, I can focus on the data itself — which sources to connect, which columns to map — instead of wrestling with JSON syntax."— Gavin, System Administrator · Post-launch validation
Workflow efficiency — 70% time saving per configuration
Before: creating a job configuration took 45 minutes — JSON file creation, manual column mapping, syntax verification, test import. After: 12–15 minutes using the Admin Console — guided workflow, visual mapping, instant validation before save.
Error elimination — visual interfaces catch mistakes before import
JSON syntax errors dropped significantly. The visual column mapper with real-time validation flagged unmapped columns and type conflicts before configurations were saved — eliminating the class of errors that only surfaced at import time. Failed imports due to configuration errors reduced by 85%.
Self-service scale — 2.5× more configurations, same team
New System Administrators became productive in days instead of weeks. The team confidently managed 2.5× more job configurations with the same headcount. The Redwood-based Admin Console became the foundation for Oracle Unity's self-service data integration strategy across product lines.
Download the full case study PDF — wireframes, research findings, and Redwood implementation detail.