Nike · Data Insights Platform
Nike Data Insights Hub.
A one-stop shop for people of every technical background in the data insights space. It replaced constant cross-geo handoffs with a single place to find dashboards, build queries without SQL, and automate recurring work with AI. It was scoped for one team, shipped in one quarter, and expanded to 10–15 teams.
Role
Lead Product Designer & Lead Front-end Engineer
Team
10 engineers
Timeline
One quarter, delivered on time
Executive Summary
The Data Insights Hub is an internal platform that gives everyone working with data at Nike one place to discover dashboards, assemble queries, and automate recurring analysis, whether or not they write SQL.
I was the lead product designer and lead front-end engineer on a team of 10 engineers. I used AI to accelerate both the wireframes and the interactive prototypes, and paired that speed with research analysis, edge-case testing, and A/B testing to make sure we were building the right product, not just building it quickly.
“We were given one quarter to build a tool for one team. We delivered on time, and because we designed for every technical background, the product grew to serve 10 to 15 teams.”
The Problem: Knowledge Lost in Handoffs
Data work at Nike spans teams in different geos and timezones. Every question that crossed a team boundary became a handoff, and every handoff leaked context.
1. The Timezone Relay
Workflow FrictionA question asked in one geo often waited overnight for a team in another timezone. Even simple requests, like what a metric means or who owns a dashboard, could take a full day to resolve.
2. Persistent Knowledge Gaps
Information ArchitectureDefinitions, table locations, and query logic lived in people's heads and scattered docs. Each handoff dropped some of that context, so the same questions kept resurfacing.
3. The SQL Barrier
AccessGetting an answer usually meant writing SQL against unfamiliar schemas. Anyone without that background had to queue behind a data engineer, which added yet another handoff.
Process: AI Speed, Research Rigor
With one quarter on the clock, I used AI to compress the design loop while keeping every decision grounded in evidence.
AI-Assisted Wireframes & Prototypes
Wireframes and clickable prototypes were generated and iterated with AI, turning stakeholder conversations into testable screens in days instead of weeks.
Research Analysis
Research analysis mapped where handoffs happened and which questions repeated most, shaping the information architecture.
Edge Testing
Empty groups, invalid queries, conflicting filters, and failed runs were designed and tested on purpose, so the product stayed trustworthy for less technical users.
A/B Testing
Competing layouts and flows were A/B tested to confirm we were building the right product before committing engineering time.
The Pivot: Designing for Every Technical Background
Early research showed the hub could serve far more than data engineers. We widened the target personas and designed each surface so less technical users could succeed without a handoff.
Data Engineers
Original audienceOwn the pipelines and tables. Needed a single place to publish, document, and monitor data so questions stopped arriving as late-night handoffs.
Data Scientists & Senior Data Scientists
ExpandedComfortable with data but not every schema. The Admin Panel lets them assemble metrics, dimensions, filters, and joins, then take a natural-language prompt or SQL skeleton.
Software Engineers
ExpandedNeed reliable numbers inside their own products. Automations let them schedule pulls and alerts in plain language without learning the warehouse layout.
Directors & Senior Directors
ExpandedNeed answers at a glance. Dashboards and Favorites give them a curated, cross-team view without waiting on another region to come online.
Product Surfaces
Four core surfaces cover the path from finding data to acting on it. You can try each one in the prototype above.
Dashboards & Favorites
DiscoveryEvery team's dashboards and automations live in one grouped home. Each group previews four items with a count badge that expands in place, and Favorites gives each person a cross-team shortlist.
Create Dashboard
Drag-and-Drop BuilderStart from a blank canvas and drag chart types (bar, line, area, donut, KPI, table, and more) onto it. Charts can be reordered by dragging, resized, or removed, and the finished dashboard saves directly into a team group.
Admin Panel
No-SQL Query BuilderPick metrics and dimensions, click a dimension to preview and filter its values, add joins and a date range. The panel validates the query and outputs both a natural-language prompt for an LLM and a SQL skeleton.
Create Automation
AI-Assisted WorkflowsDescribe a task in plain language and the assistant drafts it as a workflow step. Users confirm each step, choose a compute warehouse, set a schedule, then run and save, with no pipeline code required.
Impact
Designing for every persona turned a single-team tool into shared infrastructure.
1 quarter
Delivered on Schedule
Shipped within the single quarter originally allotted to build the product.
1 → 10–15
Teams Served
Scoped for one team; expanded persona coverage brought it to 10 to 15 teams.
Org-wide
Adoption Path
Adoption is expected across all of those teams, from data engineers to senior directors.
The biggest win wasn't a single feature. Once a director, a data scientist, and a data engineer could all get answers in the same place, the handoffs that slowed every geo began to disappear.