Data Platform from Zero and AI Pilot

ATB · Product Analyst

Data Platform from Zero and AI Pilot

5 days → <4 hours

Insight Turnaround

5

Teams Unblocked

+8%

Conversion Lift

Data PlatformProduct AnalyticsAI PilotSelf-Service AnalyticsConversion

Overview

Built a data lake, event schema and self-serve dashboards from zero, then used the resulting data to find a validation defect behind funnel drop-off and support a measured AI virtual-assistant pilot.

The Challenge

The initial mandate was to improve product velocity, with a belief that backlog prioritisation was the issue. Observing teams showed product decisions were instead blocked by data requests that took an average of five days and required a data engineer.

The Approach

Reframed the work around insight latency and built the event schema, pipelines and self-serve dashboards with one data engineer. Designed the data for legibility by PMs and analysts. Used the new measurement foundation to find a validation logic defect and developed a business case for the AI virtual-assistant pilot.

Key Outcomes

  • Built a data lake, event schema and dashboards in three months for five cross-functional teams.
  • Reduced insight turnaround from five days to under four hours.
  • Found a data-validation edge case; fixing the logic lifted conversion 8% and reduced time-to-activation 20%.
  • The ATB virtual-assistant pilot achieved 20% case deflection, worth about $150K annually.

The Result

Five teams could answer product questions independently, with insight turnaround under four hours. A validation fix increased conversion 8% and reduced activation time 20%; the ATB AI pilot achieved 20% case deflection.