As part of APP, SD team submitted the following hypothesis:
If we make improvements to the Commons upload wizard that minimize one of the most common problems that cause future deletion requests, we will decrease moderator burden as measured by a 2% decrease of the ratio of newly uploaded media that become deletion requests (as per KR WE1.2). One improvement will be to encourage users to select the right option when uploading not their “own work”. We will identify other improvements and measurable goals based on an analysis of a sample of 1000 deletion requests.
The goal of this task is to investigate data available, calculate baselines to enable Product validate the choice of success metrics and goals.
This ticket will lead to defining a dashboard for commons moderator workflows to monitor metrics on a regular basis in next FY.
Some decisions:
- We focus on desktop (As the majority of upload traffic comes from desktop - see https://phabricator.wikimedia.org/T337417)
- Focus on individual uploads (exclude bulk uploads)
- We focus on upload wizard
Requirements
Step 1
Calculate the baselines (with absolute numbers) and create a report table. While we focus on upload wizard, it would be interesting to see comparison between other upload methods:
- Total number of upload media within a month (through upload wizard) (filter by own work and not own work)
- Total number of filed deletion requests within a month and total number of speedy deletions (filter by own work and not own work)
- Main metric: the deletion rate of upload media (where we can filter by own work and not own work) within a month. This does not include speedy deletions.
- Metric of interest: deletion rate of upload media (where we can filter by own work and not own work) which is speedy deletion within a month (as it was suggested that it might go up)
Additionally look at the deletion request queue: How long does it take a DR to get resolved?
- e.g. the ratio of not closed DRs after x months (3 months as per suggestion from Legal). Look at distribution, bring to discussion with the PM and team.
Note:
- Data Engineering team might help with some instrumentation: https://phabricator.wikimedia.org/T336955 (although seems like this won't be prioritised)
Step 2:
- Success metric: calculate % of media uploaded within a given month (calculate for the past 12 months) that is flagged for deletion (DRs) within 30 days of being uploaded.
Data sheet https://docs.google.com/spreadsheets/d/1qR6yPFktt-DTfETFJD50a2ooVbPe6Ad9m1S3fEdK2kE/edit#gid=0