Page MenuHomePhabricator

Tranche burnups are incorrect
Open, NormalPublic8 Story Points

Description

The forecast lines for both the tasks completed and the work remaining are unconnected to the historical data.

After some inspection, this is because they pull from different data sources, and the burnup areas are probably the wrong ones.

Event Timeline

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptJun 12 2017, 10:16 PM

First round of digging suggests that the forecast lines are correct and the burnups and backlog are doubled for some reason. probably caused by https://github.com/wikimedia/phlogiston/commit/e3fdbc14b53cf46d5c98d7435acfd5666d2c6919, the last big change to touch this stuff.

JAufrecht renamed this task from Phlogiston forecast lines are misplaced to Tranche burnups are incorrect.Nov 3 2017, 9:08 PM
JAufrecht triaged this task as Normal priority.
JAufrecht updated the task description. (Show Details)
JAufrecht set the point value for this task to 8.Nov 3 2017, 10:56 PM

One hypothesis for why one of the counts (either the burnup line or the backlog areas) is wrong is that the recently_closed table may be mixing data. It has one column for date, one column for (first day of) week, one column for (first day of) month, and one column for (first day of) quarter, plus two dependent data fields. However, the week, month, and quarter should each have a separate row, since they each represent different time periods and presumably have different aggregade data. So, there could be mingling of week, month, and quarter data, which would explain variation in the discrepency.

Solution would be

  1. re-understand and document the relationship between the date field and the week/month/quarter date fields
  2. probably make separate rows for each of the three time periods.