Include cycle time in usage ping
We want to be able to show that a higher adoption of GitLab features, which is the same as using concurrent DevOps, which is reflected by a higher ConvDev Index, results in a lower cycle time. We don't have the needed data to show this for on premises GitLab installations.
We should add cycle time to the usage ping.
Proposal
In the usage ping, send the average cycle time of the last 7 days of each project on the instance.
The main problem with sending just a single value is that it's very often missing. We should consider to take all values and send the average, sd and any missing of each. So this would mean in the usage ping:
{ avg_cycle_analytics:
{
issue: {
average: 123,
sd: 23,
missing: 23 }
},
{
plan: {
average: 321,
sd: 12,
missing: 21,
time_since_start: 50 }
},
total: 543
}
}
Each value would report the average in seconds, rounded to a full second.
Edited by Job van der Voort