​
Pull Request Metrics

​
Productivity Score

An AI-generated score (0–100) that estimates the relative effort and complexity involved in completing each pull request. This is calculated using contextual metadata (code changes, issue data, etc.) and LLM-based analysis. Learn more about how we calculate the Productivity Score here.

​
Rework Rate

The percentage of lines that engineers modify or delete from their own recently written code, calculated as (Total Reworked Lines / Total Changed Lines) × 100%. A reworked line is one that was originally written by the same author within the last 30 days. This metric can be analyzed at the pull request, engineer, team, or organization level to identify patterns of inefficiency or areas where requirements may be unclear.

​
Total Lines of Code

The net number of lines of code added, modified, or deleted across all merged pull requests. This includes both additions and deletions.

​
Cycle Time

The duration between the first commit and when a PR is merged. This reflects how long it takes for code to move from draft to production-ready.

​
Review Time

The time taken between the first review comment and the final merge of the PR. This measures how quickly PRs are reviewed once submitted.

​
Time to First Comment

The duration between when a pull request is created and when the first comment is posted on it. This metric measures how quickly team members engage with new pull requests, indicating responsiveness and collaboration patterns. A shorter time to first comment typically suggests active team engagement and faster feedback cycles.

​
Review Rate

The percentage of pull requests that received a code review before being merged. A key signal of collaborative development and code quality checks.

​
Merge Frequency

The number of pull requests merged per engineer (or per team) per day. It reflects how frequently changes are being integrated into the main branch.

​
AI Metrics

​
Adoption Rate

The percentage of engineers who used AI tools (e.g., Copilot, Cursor) during the selected time period. This helps measure how widely AI is being integrated into daily workflows.

​
Productivity Score Boost per Engineer

The average increase in Productivity Score for an engineer on days when they use AI tools compared to days they don’t. This reflects the measurable impact of AI on engineering output.

​
AI Requests per Engineer

The number of AI tool activations, completions, or suggestions made per engineer. This helps quantify the intensity of AI usage.

​
Most Used Model

The AI model most frequently used across the organization (e.g., GPT-4, Claude, Codium). Useful for understanding model preferences and potential standardization.

​
Most Used Language

The programming language most frequently associated with AI-assisted development during the selected period. Helps identify where AI is most leveraged across your codebase.

​
PR Categorization

​
PR Categories

Each pull request is automatically classified using an LLM into one of the following categories:
  • Bug Fixes – Code changes that resolve bugs or errors.
  • Tech Improvements – Refactors, performance optimizations, dependency upgrades, or infrastructure changes.
  • Feature Work – Code that implements new user-facing functionality or business logic.

​
State

The current status of the PR: Open, Merged, or Closed.

​
PR Opened At

The timestamp of when the pull request was initially created.

​
PR Merged At

The timestamp of when the pull request was successfully merged.

​
Deployment Metrics

​
Deployment Frequency

How often new changes are deployed to production. This is tracked using deployment metadata from your CI/CD pipelines.