Open the report
Go to General → SDLC, then choose the date range and scope you want to investigate. The report compares the selected period with the previous period. SDLC is available when thesdlc feature is enabled for your account. Access also depends on your organization’s report permissions and settings. Sidebar customization can change where the report appears.
Follow work through the lifecycle
The Tickets tab starts with tickets from completed iterations that closed during the selected period. It traces each ticket through its linked pull requests and shows where work progressed or stopped:- No PR — no linked pull request was found.
- PR opened — at least one linked pull request entered implementation.
- Awaiting review — an open pull request has not received a first review.
- In review or Awaiting merge — a reviewed pull request remains open.
- Closed unreviewed or Closed after review — a pull request closed without progressing to a merge.
- Merged — at least one linked pull request was merged.
- No deploy recorded — the pull request was merged, but no production deployment was recorded.
- Deployed — at least one linked pull request reached production.
- Bug traced — Bug Finder traced a later bug fix back to the deployed work.
- No bug traced — no later bug fix was traced back to the deployed work.
Understand phase health
The report organizes health signals into five phases:
The By phase table lists each phase with its stages, ticket count, health score, and coverage. Open a phase to see its drawer, and use the pager to move between phases. The drawer shows the phase’s health, the tickets at each of its stages, and charts for the stage. For Verification, the drawer is split into Review, Mechanical validation, and Awaiting merge sections. Expand Scoring details to see component values, scores, configured and effective weights, targets, benchmarks, sample counts, and coverage explanations.
A phase score combines eligible signals using configured weights. If some signals cannot be scored, the remaining eligible weights are adjusted. If there is not enough eligible coverage, the phase score is withheld. The report explains common reasons a signal is unavailable, including insufficient samples, missing benchmarks, unsupported scope, or unavailable source data.
Verification currently measures review effectiveness. It does not assess automated tests, test coverage, flaky tests, or security checks.
Find review bottlenecks
The Review bottlenecks tab breaks pull-request time into steps, including Merge → deploy, and shows where merged pull requests spend the most time. First-review and approval timings include linked pull requests that remain open or closed without merging. Review rounds and changes requested cover merged pull requests only. The tab compares these measures with the previous period. Review time by step plots each step’s median against the previous period and shades the Middle 50%, the range between the 25th and 75th percentile hours. The table beside it lists each step’s median, middle 50%, change, and how many pull requests are still waiting, and marks the slowest step. A wide middle 50% means the step’s timing varies a lot between pull requests, so the median alone can hide slow cases. Use this view to identify a step worth investigating, then inspect the underlying pull requests and context before deciding what to change.Explore AI efficiency
The AI efficiency tab compares pull requests linked to planned tickets and shows agent session outcomes. The pull-request views cover planned work; the session view includes all sessions matching the report filters, even when they are unrelated to planned tickets. It includes:- AI-assisted vs human PRs — compare PR counts, merge and deploy rates, and time to merge by authorship category. Bot and unclassified PRs are shown together when present.
- AI spend by PR outcome — see recorded AI spend and PR counts by outcome.
- Agent sessions by outcome — explore recorded agent sessions and their outcomes; sessions aren’t linked to tickets, so this view isn’t limited to planned work.
