Before you start

Please ensure you’ve already connected the tools before reviewing this doc.

Tools Connect: GitHub

Why should I measure GitHub Copilot usage?

As more engineering teams incorporate GitHub Copilot into their daily workflows, the ability to measure the usage and adoption rates of GitHub CoPilot, along with its impact on development teams, becomes more important. Allstacks allows you to understand the ROI of this generative AI coding tool across your teams.

Learn how you can track, view adoption trends, and optimize your Copilot development cycle in our overview video below.

Where to find the New GitHub Copilot Usage metrics in Allstacks

Go to Configure (left sidebar)
Click Copilot Usage Metrics

If you can’t see this in the menu, please contact your Customer Success Manager, or email support@allstacks.com

The New Tag Behavior

We categorize Copilot usage into four stages based on user activity over a trailing 30-day period. These stages help standardize how we measure and compare adoption across users.

New Copilot Tags

Copilot Heavy

Users with very high and consistent Copilot activity. Copilot is a core part of their daily workflow.

How this is measured:

High average lines of code changed per active day
- Example: |-1000| + |1000| = 2000 total changes
High average number of Copilot requests per day
High percentage of AI-active days

Copilot Regular

Users with moderate and consistent usage. Copilot is used regularly but not for all development work.

How this is measured:

Moderate lines of code changed per active day
Moderate number of Copilot requests per day
Moderate percentage of AI-active days

Copilot Light

Users with low or occasional activity. Usage is inconsistent or limited.

How this is measured:

Low lines of code changed per active day
Low number of Copilot requests per day
Low percentage of AI-active days

Copilot None

Users with:

No Copilot activity in the past 90 days, or
Not invited / not enabled for Copilot

Copilot Agent Intensity

The Copilot Agent Intensity tag measures how heavily users interact with AI agents over the past 30 days.

How this is measured:

Average agent intensity per day over a rolling 30-day period
- Total agent-based code generation activity count is used

Users are divided evenly into percentile groups:

Tag	Percentile

Tag	Percentile
Light Agent Intensity	0% – 33%
Medium Agent Intensity	34% – 66%
Heavy Agent Intensity	67% – 100%

How This Is Measured

All Copilot usage stages are calculated using the same logic:

Time Window

Based on activity over the last 30 days

Activity-Based Counting

Days with no activity are completely ignored
We only look at days where something actually happened

Weekend Handling

If no activity on weekends → ignored
If there is activity → treated like a weekday
In simple terms: weekends don’t matter — only activity does

Metrics Used

We evaluate usage based on:

Lines of Code Changed (per active day)
- Measures total code impact
- Example: additions + deletions = total change
Average Number of Copilot Requests per Day
- How often Copilot is actively used
Percent of AI-Active Days
- How frequently Copilot is used across active days

Copilot Usage Metrics

This metric measures all usage activity from active users, code acceptance, contributions, and individual user usage:

Top Banner:

Total Active Users: Count of distinct users who had any Copilot activity in the selected date range
Daily Active Users: Average number of unique active users per day (sum of daily distinct users / number of days with any activity).
User Retention Rate: Percentage of users active in the prior 30-day window (days -61 to -31) who were also active in the recent 30-day window (days -30 to 0). [0 = end date selected, today in case of trailing days]
Total AI-Accepted Code: Total lines of code changed from accepted AI suggestions (lines added + lines deleted).
Average Acceptance Rate: (suggestions accepted + tabs accepted) / (suggestions total + tabs total) x 100. Includes both inline completions and tab completions.
Requests per Active User: Average of each user's daily request rate — for each user compute total_requests / active_days, then average across all users.

Code Acceptance Rate Over Time

Tracks how often developers accept AI code suggestions (tabs/applies) daily.
Why it matters: Shows trends in AI suggestion quality and developer trust.

Net Code Contribution by AI

Compares AI-generated lines added vs deleted to measure true code impact.
Why it matters: Reveals if AI is driving new features or just refactoring.

Feature Usage Distribution

Monitors daily use of different Copilot features.
Why it matters: Understand which AI tools developers rely on most.

Usage Breakdown (User Table):

Total Suggestions Accepted: sum of suggestions accepted + tabs accepted for the user
Suggestions Acceptance Rate: (suggestions accepted + tabs accepted) / (suggestions total + tabs total) x 100. for the user
Daily Requests: total_requests for the user / active_days for the user.

All data comes from AI_USAGE service items belonging to services Github, GithubApp, and GithubCopilotBusiness.

Where to find Legacy GitHub Copilot metrics in Allstacks (Deprecating)

All Copilot Adoption Rate and Suggestion Acceptance metrics are being deprecated and will be removed once all customers migrate to the new Copilot Usage model.

Go to Configure (left sidebar)
Click Metrics
Scroll down to Copilot Usage

You’ll see:

Adoption Rate
Suggestion Acceptance

How Legacy metrics are generated

Usage metrics are generated by pulling data from Copilot API.
A “suggestion” by Copilot is defined by either a prompt or autocomplete.

The Legacy Tag Behavior

These tags behave like any other tags. They can be used to filter or group data in any metric. Specifically, this set of tags will help users understand the differences in behavior and output between copilot and non-copilot using developers.

Existing Copilot Tags

“Copilot Activity 0-7 days ago.”
“Copilot Activity 8-30 days ago.”
“Copilot Activity 31-60 days ago.”
“Copilot Activity 61-90 days ago.”
“No Copilot Activity in past 90 days.”
“Not Invited in Copilot” (never invited, therefore never accepted)

Copilot Adoption Rate

The adoption rate metric measures the effectiveness of the rollout of Copilot at your organization. You will be able to drill down to see the contributor list in each category.

Active - A user who took action that week.
- When you actively use Copilot to generate code suggestions.
- When Copilot processes your context to provide inline suggestions.
- When you accept a Copilot suggestion.
- When you interact with Copilot chat.
Inactive - A user who took no action that week.

Copilot Suggestion Acceptance Rate

This metric measures the number of suggestions accepted and suggestions rejected over time. This data is pulled from users that are defined as “active.”

Additional Reports

With the copilot integration, your contributors will get an auto-generated tag that can help you take a new perspective on the metrics your team uses to measure productivity, code quality, and coding productivity. This will allow you to compare the output of your cohort that uses copilot against those who are not yet using copilot to speed their development.