9to5 coverage · Guides
How 9to5 software verifies coverage changes
An internal engineering example: make changed-line evidence explicit while keeping test quality and deployment status separate.
The problem in our own marketing release
While updating this application’s marketing pages, we found that the page controller and page helpers were excluded from coverage reporting. Rendering tests passed, but an aggregate coverage score could not prove execution of those excluded modules. This is a 9to5 software internal workflow example, not a customer testimonial.
What we changed
- Removed the page controller and page-helper exclusions from both coverage configurations.
- Moved page-specific structured data into a dedicated SEO module with direct tests for software identity, page types and breadcrumbs.
- Rendered every public page and checked metadata, navigation, duplicate IDs and the pricing content.
- Compared added executable source lines with the coverage artifact, including new files that were not yet tracked by Git.
What the first local validation established
| Evidence | Observed result |
|---|---|
| Application tests | 241 tests passed |
| Changed executable Elixir lines | 54 of 54 measured lines executed |
| SEO module | 100% measured module line coverage |
| Vitest documentation example | A real test run generated a non-empty LCOV file |
These are results from the first local SEO-foundation validation on 7 September 2026, not live service metrics. The wider application’s total coverage was 94.3%; the changed-line result was 100%. The figures describe that snapshot and will not track future changes automatically.
Keep the proof boundary visible
The coverage artifact attributed generated HEEx execution to the page module rather than separate template lines. We therefore paired executable-line evidence with rendered-page tests and mobile/desktop inspection. We did not treat local success as proof of hosted CI, deployment or Google indexing.
This example does not claim that an agent produced every test or that MCP was used to obtain these local measurements. It demonstrates the evidence standard we use before connecting CI reports and agent workflows.
Reproduce the Elixir report setup · Apply the same verification sequence to a PR