North Star for Memory
I would like every group to have a single “north star metric” that your group is trying to move with your development activities. In most cases this will be AMAU/SMAU, but I’m open to other north star metrics. Once you align on the metric, then we need to baseline it, and set a quarterly goal for it. Then we’ll have regular reviews to track performance against the metric, and review activities underway to move it further. We will use this data to inform future investment decisions, so it’s important to lean in on this.
Introduction
GitLab needs to become a more data driven organisation to decide where to make future investments. In order, to support this effort every group has been asked to create a north star metric that should measure the thing that matters most to us. We can then track against that metric. In many cases, this will be related to Stage Monthly Active Users (SMAU) but alternatives are possible.
Further information
Proposal
Requests per Hour (Rolling 7 day average) / ( (GB of RAM * Cost per GB of RAM Hour) + (Cores * Cost per Core Hour) )
North star - Since the memory team primarily focuses on the efficiency of our service, we can quantify that as the number of requests we process per dollar of compute.
We should include CPU and Memory costs of the GitLab service in the cost computation (include Sidekiq and other services), and ignore storage since that is not a focus of the Memory team.
Supporting metrics
- Apdex - ensure we are not negatively impacting UX
- Minimum CPU/RAM requirements - work to reduce the minimum requirements of GitLab