Match-Quality Comparison Dashboard (Old vs. New Engine)
A media-monitoring / data-orchestration platform
Overview
A dashboard for comparing article-matching quality between the legacy production system and the new boolean query service, so the team can validate the new engine before rollout.
The Challenge
When replacing a matching engine, you must prove the new system is at least as good as the old one. That means quantifying overlap, precision, and divergence per subscription, and being able to inspect exactly why each article matched.
What We Built
A React frontend backed by an Express API on Cloud Run, exposing endpoints for subscriptions, old vs. new matches, per-id comparison, and aggregate metrics. The UI offers a subscription overview with match counts from both systems, side-by-side article comparison, keyword highlight visualization showing why each article matched, prev/next navigation with keyboard shortcuts, charts for overlap/precision/distribution, and category filters (All / Only in Old / Only in New / In Both). Real comparison datasets (e.g. exported CSVs) drive the analysis, and the app deploys via Cloud Build.
Technologies & Approach
React + Express in TypeScript, containerized for Cloud Run with Cloud Build CI. The emphasis is on explainability, surfacing keyword highlights and category breakdowns rather than just aggregate numbers.
Outcome / Impact
Gave the team a concrete, visual way to evaluate the new boolean matching service against production, building confidence in the migration by exposing precisely where the two engines agreed and diverged.
Capabilities Demonstrated
- A/B evaluation tooling for a core algorithm change
- Explainable match visualization with keyword highlights
- Metrics dashboards (overlap, precision, distribution)
- Cloud Run + Cloud Build delivery