Problem
What made this hard
The existing setup required repeated saves and page changes. At enterprise scale, one channel can contain 10 to 40 sources, many views, and more than 100 fields, so continuity could not come at the cost of visibility.
All case studies03 / 11
Data platform · System design · 2025
Unified Data Configuration
Turn channel setup and multi source mapping into one continuous system.
RoleLead Product DesignerStatusValidated conceptContextTapClicks · Data managementEvidenceThree saves reduced to one in the design target · 10 to 40 sources · More than 100 fields · Three documented iterations
Outcome
What changed
A single configuration workspace with one final save, inline creation, grouped mappings, sticky field context, and protection for unfinished work.
The thesis
One validated concept that joins Channel Configuration with Unified Data Configuration, from creating the channel to mapping equivalent views and fields across sources such as Meta, Google, and TikTok.
Case-study proof
The decision record
Enough context to judge the choice, inspect the work, and separate evidence from ambition.
Why this, not that
I grouped mappings around each unified field and kept its context sticky instead of sending administrators through separate source pages. That choice adds repetition at very large scale, but it makes every equivalent mapping auditable in one place.
Why this case is specific
The concept was stress-tested for enterprise channels with 10–40 sources, many views, more than 100 fields, and an existing minimum journey of three saves.
What you can inspect
Validated prototype · product and iteration evidenceThe page shows the before-state save gate, the consolidated workspace, and field-centred grouping. It also records three iterations and clearly labels the work as a validated concept rather than a shipped product.
Decision to outcome
The selected model reduced the design target from at least three saves to one final save while preserving mapping visibility. It did not receive an engineering slot, so no launch or customer-impact metric is claimed.
Interface evidence
Design Iteration History · 3-Version Evolution
Evolution of the Unified Configuration Workspace
The workflow evolved through 3 distinct iterations based on stakeholder reviews with product managers and engineering lead feedback.
Iteration A · Baseline
3-Save Gate Flow
Configuration forced administrators through 3 distinct save gates across separate pages. Leaving the browser silently removed unsaved field mappings.
❌ Rejected: High context loss & repeated context switchingIteration B · Source Workspace
Source-First Accordion
Grouped fields by connector source (Meta, Google, TikTok). Improved single-connector setup but required jumping across accordions to audit a single metric.
⚠️ Trade-off: Fast setup per source, hard to audit cross-source metricsIteration C · Validated Final
Field-Centric Grouping
Grouped mappings by unified field with sticky headers. Reduces 3 save steps down to 1 final save while keeping all 100+ fields auditable in one workspace.
✅ Selected: Validated by stakeholders for enterprise data auditabilityRelated configuration work · 2025
Channel Configuration
Replace repeated saves with one continuous setup flow.
RoleSolo Product DesignerStatusValidated conceptContextTapClicks · Enterprise data managementEvidenceThree saves reduced to one in the design target · Three documented iterations
The thesis
A mature channel setup flow required at least three saves, spread configuration across pages, and could discard unfinished work. I mapped the failure, consolidated the journey, and kept the necessary detail visible.
Problem
What made this hard
Creating one channel required repeated saves. Every Data View and Field created another loop, while leaving the page could silently remove unfinished changes.
Outcome
What changed
A validated single page concept with one final save, inline creation, grouping around fields, and protection for unfinished work. It did not receive an engineering slot, so I do not claim launch impact.
01 / Context
Find the problem behind the screens
With no dedicated research or analytics function, product signals arrived through years of client feedback held by product managers. I audited areas with low adoption and visible friction, then made the repeated save cost concrete.
02 / Reframe
Turn a vague request into a testable question
The brief asked for easier screens. I reframed it: how can a dense setup remain continuous and auditable without hiding the mappings administrators need?
- Administrators need continuity
- Data operators need every mapping visible
- Engineering needs a pattern the product can support
03 / Evidence
Show the failure before drawing a solution
A task audit exposed the minimum three save journey and every repeated loop. Recordings compared page jumps, context loss, and the proposed consolidated flow. Three built iterations captured how the direction changed with stakeholder input.
04 / Principle
Continuity, auditability, recovery
Three rules guided the work.
- Keep creation and configuration together
- Group mappings around the field being inspected
- Warn people before unfinished work is at risk
05 / Choice
Organise around the task that happens most
I grouped by field because auditing a metric was the more frequent job and every related source mapping stayed in one scannable block. The cost is more repetition for accounts with many sources, and the case study makes that tradeoff explicit.
06 / Evidence
Claim only what the work supports
Stakeholders validated the concept, but it did not ship before the project ended. The supported evidence is a design target of at least three saves reduced to one, three documented iterations, and no invented product metrics.
01 / Audit
Expose the cost of a fragmented setup
The task audit revealed a minimum three save journey, repeated page changes, and the risk of losing unfinished configuration. I reframed the work around continuity, auditability, and recovery.
02 / Scale
Design for the real maximum
I tested the structure with realistic connector and field counts, not tidy examples. That changed navigation, sticky context, collapse behaviour, and grouping decisions.
03 / Model
One definition, many mappings
Each unified field has a common name and type, then maps to metrics within grouped source and view combinations.
- Create sources, views, and fields inline
- Select many sources together
- Keep field and type visible
- Group related mapping rows
- Support bulk add
- Preserve context through overflow
04 / Guardrail
Explain what unlocks the next step
Mapping becomes available only after the channel and sources are defined. Disabled states explain what is missing, and leaving the workspace with unfinished work triggers protection.
05 / Evidence
Claim only what the concept supports
Stakeholders validated the consolidated direction, but it did not receive an engineering slot. The supported evidence is a design target of three saves reduced to one, three documented iterations, and realistic testing at enterprise data scale.
Supporting evidence
Bring me the difficult part.
Complex workflows, AI trust, and enterprise systems.
Discuss a product challenge