Research:Editor milestones/Lifecycle

From Meta, a Wikimedia project coordination wiki
Jump to: navigation, search

Questions[edit]

  1. All of these editors made their 1,000th article edit in the same few months. Of editors who began contributing in the past 4 years (2008 and later), what do their editing velocities look like? (E.g., slow and steady, rapid and condensed?)
  2. (Optional) How different is the velocity of editors who began contributing before 2008?
  3. What is the typical/average velocity? What do the outliers look like?
  4. What does the velocity of editors who made their 100th edit to articles look like in comparison?
  5. What are the most useful temporal metrics for defining an editing session? (1 hour of inactivity? more/less?)
  6. What does a typical editing session look like...
  • in the first few hours of editing?
  • in the first day?
  • subsequent days?
  • for these editors as compared to a random sample of current ones?

High-velocity editors[edit]

Run the analysis for editors who started post Jan 1, 2008

Include the following stats for each time period:

  • Mean
  • Median
  • High
  • Low
  • 25th percentile edit count
  • 75th percentile edit count
Generate a CSV that has each editor and their edit counts per time bucket and lifetime edit count.

E.g.:

User 0-1 1-2 2-6 6-12 12-24 24-48 48-168 Lifetime # of sessions
Run a similar analysis for 100+ editors that are post 2008.

Editing sessions[edit]

Estimate the number of edit sessions within the first 7 days.
  1. We can use a simple definition of "count the number of >=1 hour gaps within the first 7 days".
  2. Assumption: if a user stops for >=1 hour, they've restarted an editing session (e.g., went away for a cup of coffee, switched contexts, etc.).
  3. The industry standard measurement for an internet session is 30 minutes of inactivity (e.g., that's what comScore uses to identify sessions). But since editing activity is a very time-intensive thing, 1 hour feels more appropriate.
  4. Example: if a user makes 3 edits from 1:00-1:15, and their next edit is at 3:00, there's a gap between the first and second clusters. So # gaps + 1 would equal number of sessions.
Rough estimate for the above on the order of
  1. 1-2 hours
  2. half a day
  3. a day
  4. 1+ days

Charts[edit]

1st week of editing[edit]

Mean daily edits after first day of editing
Total edits for all editors, 1 week
  • Mean: 23.61
  • Median: 5
  • Min: 1
  • 25th percentile: 2
  • 75th percentile: 16
  • Max: 1117