please use rolling averages for statistics
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| ubuntu-sponsoring |
New
|
Undecided
|
Unassigned | ||
Bug Description
The average graphs seem to use all time average algorithm. The the page started, that was a useful metric.
= number of items in queue =
good, as it effectively shows the subjective rate of reviews & submissions. There is obvious correlation with release interlock.
= Hours in queue =
is this the average of a total time a single sponsorship request stays in the queue? If so, it should be converted to a rolling average, e.g. the average total time a single sponsorship stayed in the queue which got submit between the data point's date and 10 days[1] in the past from that data point. This way you get the trend, showing how responsive reviews were in and about certain time period.
[1] 10 days because the peak was 175 hours, which is approximately 7.3 days + some extra time in case of a national holiday and/or a world-wide event (olympics, world cup, etc).
= udd hours until first response =
confusing, is this an accumulating average? Again this should probably be converted to a rolling average with a limit of what looks like 14 days?! [2]
[2] Seems odd that the average time to first response is higher than hours in the queue. Does this mean that some requests are sponsored without a response and not calculated correctly? If that is the case, sponsorship timestamp should be counted as a first response as well as the end time.
= udd number of contributors =
Is this rolling total? In that case again it should be converted to a rolling average. After some time people gain upload rights and leave the sponsorship process all together. It is more interesting to see the trend of current udd contributors. E.g. sum of contributors at the datapoint, if they did an udd contribution between datapoint's timestamp and 40 days prior to that (4 times more than hours in the queue)
= udd number of contributions per person =
This is actually interesting. Does it take 7 udd contributions to gain upload rights? =))))
= udd number of contributions =
Should be converted to the same rolling average as 'udd number of contributors' which should have strong correlation.
ps. please add small paragraphs explaining what the graphs are showing and what is the meaning of the datapoints.
