GitHub Copilot: PR Throughput and Time-to-Merge Now in Usage Metrics API

New PR Outcome Metrics in the Copilot Usage Metrics API

GitHub has extended its Copilot usage metrics APIs to surface new enterprise-level data points focused on pull request performance and outcomes. The update, released on February 19, 2026, shifts the measurement focus from raw feature adoption toward concrete engineering impact — making it easier for organizations to evaluate whether Copilot is meaningfully accelerating their software delivery workflows.

What the New Metrics Measure

Three new categories of data are now available through the Copilot usage metrics API:

Pull request review suggestion acceptance: Organizations can now measure how often Copilot's code review suggestions are acted upon and turned into real changes, providing a signal of review quality and developer trust in Copilot's feedback.

Coding agent PR merge rates: When the Copilot coding agent authors a pull request autonomously, GitHub now tracks how frequently those PRs achieve merged status. This gives engineering leaders a way to assess the agent's practical effectiveness — not just whether it produces output, but whether that output passes human review and lands in production.

Pull request cycle time: The API can now surface time-to-merge data, measuring how long it takes from PR creation to merge. When combined with Copilot adoption data, this allows organizations to draw correlations between Copilot usage and delivery speed.

Access and Availability

The new metrics are available to Copilot customers who have usage metrics enabled. Access is restricted to enterprise administrators or users who have been granted the "View Enterprise Copilot Metrics" role. The data is surfaced through the existing Copilot usage metrics API endpoints, requiring no new integration work for teams already consuming the API.

Why This Matters

The shift from activity metrics to outcome metrics is significant for enterprise Copilot subscribers evaluating the ROI of their investment. Rather than reporting how many completions were accepted, organizations can now track whether Copilot-assisted code actually ships — providing a more defensible picture of productivity impact when presenting to engineering leadership or justifying license renewals.