# Automation coverage, measured from source

- Period: Oct 2026 – present
- Domain: National telecom carrier
- Role: Designed and built the pipeline and reports
- Key result: 81.7%, regression automation coverage across 21 teams, now measured weekly
- Stack: Node.js, git, GitHub CLI, Playwright, TestNG, HTML email
- URL: https://farhanbaluch.com/work/coverage-insights

Replaced spreadsheet tracking with a zero-dependency pipeline that reads 20+ team test repositories and publishes a weekly executive report.

### Problem

Leadership reported test-automation progress from hand-maintained spreadsheets. Every team organized its repository differently, used different tags, and counted "automated" in its own way. The numbers in the steering deck couldn't be traced back to code.

### Approach

Measure from the source of truth: the test code itself. Mirror every team repository, parse the specs, and map each team's conventions through one config instead of asking 21 teams to change.

### What I built

- A shared scanner and spec parser that collapses copied tests and understands per-team folder and tag conventions, later extended with a Java-aware parser for mobile TestNG suites.
- Four report generators: regression coverage, build-verification coverage, functional-flow coverage, and an open-PR dashboard (age, AI-review status, mergeability).
- Week-over-week change markers, a release filter, and an email-safe variant so the report reads correctly in Outlook.
- Weekly publishing to an internal CDN plus an executive email.

### Outcome

- 21 teams measured from code, not self-reporting: 81.7% regression automation (2,352 of 2,879 cases).
- 272 automatable manual cases identified, setting a 91.1% target.
- Per-team spread (66.9% to 98.5%) made it obvious where to invest.