Graph for understanding peer-weakness

[mgledi]

Why? The peer strength page is very helpful to detect strong peers. But it is not so helpful when looking for week peers or even isolated groups. It is important for engineering teams to build an almost fully connected graph. I experienced in the past that the connectivity is a very good indicator to see who will leave a project or even resign soon.

What? If a group in the search request is defined then show a graph on the peer strengths page. The nodes represent the people in the group. The vertices indicate the collaboration, e.g. commenting on PRs, approving or rejecting a PR. The thickness of the vertices reflects the strength of the collaboration.

[morucci]

Thanks you for the feedback and suggestions.

As of now when a group is specified then the computation is done on each group members (each member is treated as a reviewer or commenter). This means that the authors column might includes authors outside of the group.

So what about having:

• A switch button such as "exclude non group members" to ensure authors/reviewers/commenters are only part of groups selected in the query filter.
• Swap columns - move the "Peer reviewer" column to the left

However group members w/o review activities won't be displayed in the graph (I guess rendering will skip those nodes). So then what about an option to reverse the listing on the left side to display peers with low strength ?

Something else that come in mind is that we get two lines in the list for each peers (A <-> B) and (B<->A) where this can give more insight, a user might want to see both lines squashed. wdyt ?

[mgledi]

1. The non-group members are not the problem. The problem is that I can't see by a first glance if someone in my group contributed or not. What is also not easily possible is to recognize groups (e.g. UI devs and BE devs)
2. My suggestion is the following: the nodes of the graph are all users in the defined group + all users in the peers. The vertices in the graph are the peer strengths you have calculated where (for simplicity) you should sum A<->B and B<->A. The outcome should look like https://www.researchgate.net/profile/Prohollad-Sarkar-2/publication/321846685/figure/fig1/AS:572036225683456@1513395472028/Graph-theory-analysis-in-social-network-Image-courtesy-of-5.png (without the colors)

Again, the important information for me as a manager is who is not in the graph or almost outside, i.e. the people with weak peers.

BTW: This holds for all statistics. The mean and average stuff is the first step. But what is truly helping is finding the negative outliers. E.g. PRs whose cycle time are over the average.