Blog articles

Introducing Advanced Outlier Controls

August 01, 2018

Today, we’re announcing advanced outlier controls within GitPrime, which allows you to set custom thresholds and manually flag individual commits as outliers to be ignored from calculations within the platform.

The goal with everything we do at GitPrime is to provide an accurate picture of the ground truth in software development—the truth as it exists and is experienced by your engineering team on the ground.

Inevitably, very large commits can distort the ground truth when we roll up metrics into summary metrics. For example: when Warren Buffett walks into a football stadium, the average net worth of everyone in attendance instantly rises to over a million dollars. Neither Warren nor the majority of the fan’s ground truth has changed. The same is true for very large commits.

Commit outliers interfere with our ability to understand the ground truth. Sophisticated controls are required to screen out distortion from things like:

  • Importing libraries
  • Wholesale formatting changes from prettifiers
  • Large copy and pastes
  • Inconsequential touch-ups

Total control, customized for your needs

Previously, we used our data-science expertise to define what a general-purpose outlier was. It solved for large open source libraries, but came up short, for example, when users couldn’t correct false positives (e.g. “Actually, Bob wrote that huge library and should receive full credit”).

Now we’ve leveled that up.

Today you can define an Outlier however you want. You now have the ability to programmatically set global thresholds for your entire org (both minimum and maximum), and individually mark commits directly in the UI.

Here’s how it works:

1. Set global outlier thresholds for your organization

To get started with customized outliers, you’ll want to visit the configurations page for your organization and turn on Advanced Outlier detection.

You now have a new set of tools to define what “an outlier” is. Once you customize the settings, the new outlier detection will happen automatically. Just set it and forget it.

The first set of controls give you the ability to screen out commits based impact based on an upper and lower bound. Using either percentage or absolute values, set the values to automatically exclude calculations from the largest and smallest of all commits.

Additionally, you can fine-tune settings based on size to include or exclude:

  • Semantic awareness of comments
  • Whitespace, and
  • Min and max edit length

2. Manually override any commit

Once you have configured with your global thresholds, you can fine-tune individual commits at any time directly in the UI from any report within GitPrime that displays commit objects.

If you identify a commit that you would like to exclude it from showing up in the platform calculations, simply click on it and select “Mark as outlier commit.”

The commit will still remain in view, but it will be removed from any further calculations within GitPrime. If you make a mistake or change your mind about a commit, you reverse the process in the same way.

We hope you enjoy this powerful new capability to define your own rules for Outliers within GitPrime.

Advanced Outlier settings are a powerful way tune your data to bring the most relevant information into focus. You now have more control, more accuracy, and more understanding than ever before. All of this is aimed at providing engineering leaders with exactly what they need to see the ground truth in engineering.

Get started in your configuration dashboard, or read more in the documentation.