Request for Impact Metrics

[ccerv1]

We just published our first Request for Impact Metrics covering a number of open source and onchain related impact metrics. This discussion thread is for sharing your impact metrics! Contributions should include the following:

• Name of impact metric
• Brief description of your calculation steps
• Link to your code (eg, Colab or Jupyter notebook)
• Chart or table showing the metric applied to a set of projects

[ccerv1]

Project Health: Bus Factor

Description

My approach grabs all OSS projects with merged pull requests over the last 6 months. It calculates the number of unique contributors, the number of total contributions, and the number of contributions from the top contributor. Finally, it derives bus factor by taking the ratio of contributions from the top contributor to total contributions. A project with the highest bus factor would have a score of 1.0, effectively meaning it's a team of 1. A low bus factor approaches 0.0. There's an interesting relationship to explore between bus factor and unique contributors.

Code

https://colab.research.google.com/drive/11JLRlZEjHHCCzFtB7V7pVLsZVOCryVWs?usp=sharing

Results

[ryscheng]

Project Health: Acceleration

Description

My approach starts with the starter notebook for project velocity and looks at the the second-order acceleration of a project. This basically takes all velocity metrics (e.g. commits and PRs merged etc) and looks at the difference between 2 consecutive weeks. This will give you a sense of whether the project is speeding up or slowing down over time.
I anticipate this to be noisy over shorter intervals (days), but more interesting over longer windows (e.g. months).

Code

https://colab.research.google.com/drive/1KFW8y0okMPU5jf_K9VxtJLe3_Qyggote?usp=sharing

[ccerv1]

Very cool! Agree that this one that becomes more meaningful over time and in the context of other projects' / some baseline. Copying the image from your analysis in here.

[ccerv1]

Project Health: Burstiness

Description

My approach is inspired by the CHAOSS description of burstiness and a very simple Bollinger Band implementation forked from here. The key variables to play around with are value_col (I used code commits) and period (I analyzed over 12 week periods). I also created a metric for Burstiness derived as the number of periods where the amount in the value_col is greater than the upper Bollinger Band over that period, ie:

data['BurstPeriod'] = data[value_col] > data['UB']

Sample output:

Code

https://colab.research.google.com/drive/1B5Oyra71CNpk_Q7ieyaVOjdCsYm9IWLc?usp=drive_link