GitHub Copilot ROI Section: What the Three Card Metrics Mean

This article was researched, verified against primary sources, and written by AI agents. It is not a hands-on review.

Bottom line: one output metric, and a cost figure GitHub calls an estimate

According to GitHub’s changelog of 7 August 2026, the Copilot impact dashboard now includes a “Potential return on investment” section that, in GitHub’s words, “connects what you spend on Copilot to the pull request output you get back”. GitHub says the section is available at both the enterprise and organization level.

Three things decide how far this screen can be pushed. Of the three card metrics, only one sits on the output side: pull requests per month. The cost figure is not an invoice amount but an estimate derived from AI credit consumption. And GitHub itself adds the note “Treat these metrics as directional.”

Note that the spend/output split used here is this article’s framing: GitHub does not use numerator-and-denominator language, and says the section puts spend and output side by side.

What is on the ROI cards

Per the changelog, the section has two cards comparing developers by adoption depth: those working primarily in chat and code completions (Passive users and Phase 1) against agent-first developers (Phase 2 and Phase 3). Each card shows three metrics.

What these cards do not show

Review time, defect rates and lead time are not among the three. That is not the same as saying GitHub measures Copilot only by pull request counts. Per the changelog of 22 July 2026, the cohort cards in the same dashboard show average pull requests merged per user per month, median pull request merge velocity, users in the phase, share of overall users per phase, and average lines of code per day per user. Copilot usage metrics also cover acceptance rate and pull request lifecycle metrics separately (GitHub Docs, checked 9 August 2026).

The two changelogs word the PR metric differently

The 22 July cohort card is described as “average pull requests merged per user per month”; the 7 August ROI card is “Average pull requests per developer per month”, without “merged”. Neither source states whether the two definitions are identical. When quoting the ROI card, keep GitHub’s wording and do not add “merged”.

Steps before putting this screen in a budget deck

1. Check access. GitHub states the “Copilot usage metrics” policy must be enabled, and the section is available to enterprise owners and billing managers, organization owners, and anyone with a custom organization or enterprise role granting the View Copilot Metrics permission.
2. Annotate what the cost is. The changelog says: “Cost figures are estimates based on AI credit consumption, and the salary selector is a modeling input rather than actual payroll data.” Picking a compensation band recalculates the cost-derived metrics instantly, but the band is an assumption, not payroll data.
3. State that there is one output metric. If quality or speed numbers are required, they must come from the cohort cards or usage metrics instead.
4. Do not place counts taken before 7 August 2026 next to later ones. The counting method changed (see below).
5. Cite the 7 August changelog. As of 9 August 2026 no equivalent description was found in GitHub Docs, so it is the only source for this section.

Caveats

Cohort counts break continuity on 7 August 2026

According to the changelog, the same release changed cohort user counts to reflect every user active during the full 28-day reporting window, rather than only users active on the window’s final day. GitHub notes that a report ending on a weekend or holiday could previously show sharply lower counts in each phase.

The only thing GitHub states as changed is how users are counted. It says nothing about how metrics such as average pull requests per developer behave under the change, so claiming that no numbers are comparable after 7 August would overstate the source.

Phases are recalculated daily

The 7 August changelog’s phrase “those working primarily in chat and code completions” summarises how the cards are grouped, not the Docs definition of Phase 1.

Credits and included amounts

GitHub Docs state that each token is priced by model and converted into AI credits, where 1 AI credit = $0.01 USD, and that as of 1 June 2026 GitHub replaced request-based billing with usage-based billing based on model and tokens consumed. As announced in April 2026, premium request units were to be replaced by GitHub AI Credits from 1 June.

Credits are pooled per billing entity. On the promotional amounts, the usage-based billing page says “the first three months of usage-based billing (June 1 – September 1, 2026)” while the plans page says “June–August 2026”, so GitHub’s own two pages word the period differently. Only the usage-based billing page states that included usage returns to the standard amounts afterwards; the plans page does not address it. The end date should not be asserted — check your own billing screen.

The formula behind the ROI card is not published. GitHub says only that the cost is derived from actual AI credit consumption; the treatment of included credits and whether licence fees are counted are not stated, so the figure cannot be reproduced by hand.

A single-company source

Every claim here rests on GitHub’s own changelog and documentation. As of 9 August 2026 no independent coverage of this ROI section was found in English, and no retraction or correction either. All figures and quotations were checked against the primary sources on 9 August 2026; see GitHub’s official pages for later updates.