It's possible. In multiple places we use 365 instead of a precise number. Even the metric is only an estimation not a precise number due to using 30.45 as month length (the only explanation for it I can think of). https://github.com/apache/kibble-1/blob/debb80d400f03af69b65cb43ce00978b1177217b/api/pages/mail/retention.py#L198 https://github.com/apache/kibble-1/blob/debb80d400f03af69b65cb43ce00978b1177217b/api/pages/mail/retention.py#L216-L218
While it's a rather simple change it needs time to understand the code. I will try to take a look. Tomek On Fri, 23 Apr 2021 at 17:57, Sharan Foga <sha...@apache.org> wrote: > > Hi All > > I am posting this here so we have a record of it. > > Justin McLean posted a note in our Slack Channel that there seems to be a dip > in the statistics every 4 years. He was wondering if it is a leap year bug. > > To replicate > > 1) Login to Kibble demo > 2) Click on Engagement > 3) Click Contributor Retention > 4) Look at the graph for email retention > > I took a look and can replicate it. The dips do seem to be starting around > the 1st March on 4 leap years 2004, 2008, 2012, 2016 and there is currently > the beginning of a dip in 2020. > > Next step will be to take a look at the data and calculation. > > Thanks > Sharan
