We’ve updated the name of the “Features” metric to “New Deliveries” to avoid confusion with project management terminology. This change is purely naming-related and does not impact how the metric is calculated or displayed.

We’re making an adjustment to improve the accuracy of our data — specifically, how we track the amount of code that is New, Churn, and Refactor. These updates provide a more accurate view of changes over time, especially for teams that ship new code iteratively.

You may see adjustments to your metrics and benchmarks—most notably, an increase in New Units per Developer (formally Features per Developer) and a higher percentage of pull requests categorized as New. Since these improvements apply to historical data as well, your data will be uniformly adjusted upward, maintaining all relative differences and trends.

By tracking a more detailed history of file line changes—including both their original position and where they appear after edits—we are now more precisely differentiating between newly written code and modifications to existing work. This added granularity gives a more refined view of how code evolves over time, ensuring that new work is accurately captured and modifications are correctly attributed.

You can learn more about how pull requests are categorized in our documentation. If you have any questions, feel free to reach out to us at support@software.com.

We’ve updated our algorithms to more precisely track your lead time and productivity metrics.

1. Lead Time Stages

   Lead Time is incrementally updated when a PR completes a stage, rather than waiting for the PR to complete all stages. For instance:

   • Previously, Time to Review was updated after a PR was merged.
   • Now, Time to Review is updated when a PR is reviewed.
   • This change ensures that your Lead Time metrics are more timely and reflect an ongoing historical trend.

2. Excluded Pull Requests

   Our algorithm automatically excludes pull requests that are not representative of development productivity, such as those authored by bots and backmerges from main. We also exclude pull requests that are part of your release process (e.g. promoting changes from a release branch to main). We’ve refined our algorithm to more precisely detect these pull requests, such as by detecting re-used branches. Depending on your development process, you may notice a slight increase in your pull request metrics.

If you have any questions about how our algorithms work, feel free to reach out to us at support@software.com.

Based on feedback from our community, we’ve improved our Features per Developer metric to better standardize industry benchmarks and provide better accuracy when comparing groups within your organization, such as developers using vs. not using GitHub Copilot.

As part of this update, the developer count will include contributors who created a pull request in the last 90 days. Previously, the developer count included any contributor who created or reviewed a pull request in the last 30 days.

Extending the activity window from 30 to 90 days accounts for seasonality, providing a more stable representation of your organization’s size over time. Developers who leave your organization are counted towards historical activity, but immediately removed from your developer count going forward.

Cost per feature still includes contributors who created or reviewed code in the last 90 days—to more precisely capture the costs associated with developing, reviewing, and delivering features. Development costs are calculated by multiplying a group’s cost per contributor by the count of developers and reviewers within the group.

You may notice small changes to your metrics and global benchmarks. You can read more about this change in our documentation, or feel free to reach out to us with any questions at support@software.com.

We’ve improved our feature calculations to automatically ignore pull requests in development workflows—such as dependency updates, back merges, and releases—that are not representative of developer productivity. These improvements have been requested by customers. As a result of this update, you may see a slight decrease in features per engineer, and industry benchmarks will be more accurate.

A pull request is excluded from your metrics if it meets any of the following criteria:

• It is authored by a bot (e.g. a dependency update)
• The head branch is the default branch (e.g. a back merge to sync changes from default into a development branch)
• It does not include any unique commits that haven’t already existed in a previous pull request (e.g. merging a release branch into main)
• More than three authors merged a pull request into the head branch (e.g. a long-lived feature branch)

You can review the list of pull requests that are included/excluded in our definitions by hovering over a data point and viewing more pull requests in the flyout.

Based on feedback from our community, we’ve made several improvements to our active engineers algorithm:

Improvements

• We’ve updated our active engineers definition to include users in your Git organization who have contributed or reviewed code in the last 30 days. Previously, only users who created pull requests were included. Users who exclusively review pull requests — such as team managers — will now be included in the active engineer count.

Bug fixes

• We’ve updated several queries to ensure that a 30-day rolling average of active engineers is always used. Previously, an active engineer may have been excluded if they did not create a pull request in a specific week.
• We’ve added better automatic detection for pull requests authored by bots, such as Dependabot updates. As a result, you may notice a decrease in your features per engineer metric. As a reminder, you can always manually filter out bots under Settings / Git Users.

Today we’re releasing a small but helpful improvement to our app for Slack. In our latest update, you can now send notifications — including daily and weekly digests — to private channels in your workspace.

To get started, you’ll need to approve new permissions for our Slack app (it takes just a few seconds). Head over to your team’s Connected Apps in your organization’s Settings and click Upgrade Slack.

⚡ How to send notifications

To send notifications to a private slack channel, you will need to first add our bot to the channel.

1. Select the channel where you want to send notifications.
2. In the message field, type “@Software.com” and hit send. Follow the prompts in Slack to invite our bot to this channel.
3. Visit your organization settings in the Software.com web app. In the Notifications section, refresh the dropdown menu and select your private channel.

You can learn more about our app in the Slack marketplace.

We recently released Feature Breakdowns to give you more insight into what work is categorized as new, churn, and refactor during a given time period. You can access breakdowns by clicking on a data point in the Feature Breakdown chart.

💡 As a quick refresher, features are broken down into three types:

• New: new code added
• Churn: code reworked less than 21 days after being merged
• Refactor: code reworked more than 21 days after being merged

Based on feedback from our community, we’ve slightly modified how these calculations work. Instead of aggregating line changes over a time period and then multiplying the ratios by the total number of features, we are now averaging the breakdown per feature. As a result, you will likely see a lower number for % New Features vs. Rework. Our industry benchmarks have been adjusted accordingly.

If you have any questions, please don’t hesitate to reach out to us at support@software.com.

Today we're releasing a small improvement to our definition of 90-day rolling averages for active code time. 

To better understand how much time you spend coding during a typical day, your 90-day rolling average now includes days when you did not spend time coding.

As developers, we know that we are often responsible for a variety of non-coding tasks at work, such as meetings, code reviews, and project planning. As a result, most developers don’t code every day. We hope to better reflect in our metrics the reality that many developers face at work. 

For example, if you coded for several hours on Monday, but had long meetings on Tuesday and were not able to code, both days will now be included in your 90-day average. Previously, your 90-day rolling average only included days when you spent time coding. 

Note that we're updating our algorithm for rolling averages, but your total active code time each day will remain unchanged. As always, you can browse your historical data in your dashboard.

You can read the full definition of active code time.