Back to work
BlueBox Leadership Analytics
Leadership Analytics + Ask AI
A source-aware operating view that combines sales, operations, customer, and marketing data without hiding gaps in the evidence.
AnalyticsAsk AIData QualitySupabase
Ask AI evals
32
Cases covering sources, confidence, missing data, and PII redaction.
My role
Product strategy, data modeling, connector work, evaluation, and source-of-truth verification — built with AI assistance
System state
Internal analytics system
What I built
A normalized data layer plus source-aware Ask AI for operating, sales, customer, and marketing decisions.
Problem
- Revenue, inventory, sales, ads, and website data used different definitions and refresh schedules.
- An AI answer could sound confident even when one source was missing or stale.
- Private customer and revenue data needed clear public-demo boundaries.
What I Built
- Stella and Salesforce synchronization
- Normalized analytics warehouse
- Marketing and website connectors
- Deterministic metric definitions
- Source-aware Ask AI
- Data-quality warnings
Solution
- Normalized source snapshots into shared metric definitions with freshness and quality status.
- Designed Ask AI to separate sources, state confidence, and explain missing evidence.
- Added a focused evaluation suite for source mix-ups, unsupported answers, and PII handling.
Business Impact
- Gives leadership one place to inspect operating signals before making a decision.
- Makes data-quality problems visible instead of smoothing them into a misleading answer.
- Creates a reusable foundation for monitoring, analysis, and approval-only recommendations.
Lessons
- The best AI analytics layer explains what it does not know.
- Data reconciliation is part of the product, not a cleanup task after the dashboard ships.