Engineering Health

Individual PR scores tell you about individual PRs. Engineering Health zooms out and answers the team-level questions: is our change quality getting better or worse? And which habit, if we fixed it, would help most?

The top stats

Four numbers headline the page:

• Average Future Debug Cost — the project's north star. If everything Firetrail nudges you about is working, this drifts down over time.
• High / Critical PR count — how many genuinely risky changes are landing. A team can have a fine average while still shipping a scary tail; this catches the tail.
• Bugfixes traced to risky code — how many real bugfixes traced back to a PR that scored High or Critical. This is the "it wasn't a false alarm" number.
• Missing relevant tests % — the share of changes shipping without tests for the behavior they touch. Usually the most fixable number on the page.

The charts

• Most common risk pattern — which classifier says "yes" most often on your PRs. This is your highest-leverage finding: if Multi-Intent tops the chart, your team's biggest win is smaller, more focused PRs — a habit, not a heroic effort.
• Future Debug Cost over time — a weekly average. Read the slope, not the individual points: any single week is noise, but three months of drift in either direction is a real signal about how the team is changing.
• Risk band distribution — how your PRs split across low / moderate / high / critical. Healthy projects skew heavily green, with high-band PRs as deliberate exceptions rather than routine.
• Pattern incidence — the top classifiers by number of PRs flagged; the ranked list of your team's structural habits, and effectively a coaching agenda.

Using it well

This page is at its best answering "where should we focus?" in a retro or planning conversation — pick the one pattern topping the charts, agree on the habit that addresses it, and watch whether the trend responds over the next month or two.

For the follow-up questions, drill into Classifier Analytics (how each check performs, with example PRs), Team Comparison (how habits vary across the team), or just ask Historical Chat directly.