Jobright.ai / Agent Analytics
Platform analytics dashboard modeled on the kind of metrics an AI-native job matching platform actually needs to track. Match score distribution, agent recommendation accuracy, user funnel from discovery to application, and A/B test results on AI agent suggestions — all on synthetic data.
RoleSelf-directed project
Duration4 days
Year2026
StatusShipped
The brief.
AI-native consumer platforms have a unique analytics challenge: you're measuring agent performance, match quality, and user conversion all at once, and they all influence each other. I built this project to work through how I'd structure analytics for an AI matching platform from both a user funnel and a model performance angle.
The work.
USER_FUNNEL / DISCOVERY → APPLYLIVE
Discovery100%
Match Shown84%
Click-through58%
Apply Started31%
Applied22%
Interview7%
Variant A: BaselineVariant B: +18% lift ★
How I approached it.
- Modeled the discovery → match → click → apply → interview funnel on synthetic data calibrated to public job platform benchmarks.
- Built an agent performance layer that flags underperforming match recommendations for product team review.
- Designed A/B test analysis showing an 18% lift in apply rate for variant B.
- Architected the dashboard so it could plug into real production data with minimal rework.
The outcomes.
- Surfaced where agent quality and user funnel diverge — a non-obvious diagnostic for AI matching platforms.
- A/B framework reusable for any product team running agent recommendation experiments.
- Connected user-facing metrics to model-facing metrics in one operator view.
The stack.
Python
Funnel simulation and statistical testing
SQL
Schema design for production-equivalent reporting
Recharts
Interactive funnel and A/B comparison viz