Research talk:VisualEditor's effect on newly registered editors/Work log/2015-04-17

Friday, April 17, 2015

Today, I'm looking into some more data cleanup issues.

In my meeting with User:DAndreescu, he mentioned that (for wikitext):

Edit completion rate (ready --> success) is very low (much lower than VE)

So, I wrote a query to look at these rates.

> SELECT event_editor, COUNT(*) AS sessions, SUM(readys > 0) AS readys, SUM(attempts > 0) AS attempts, SUM(successes > 0) AS successes FROM (SELECT event_editor, event_editingSessionId, COUNT(*) AS events, SUM(event_action = "ready") AS readys, SUM(event_action = "saveAttempt") AS attempts, SUM(event_action = "saveSuccess") AS successes FROM Edit_11448630 WHERE timestamp BETWEEN "20150416" AND "20150417" GROUP BY event_editor, event_editingSessionId) AS foo GROUP BY event_editor; 
+--------------+----------+--------+----------+-----------+ 
| event_editor | sessions | readys | attempts | successes | 
+--------------+----------+--------+----------+-----------+ 
| visualeditor |    67261 |  63482 |    15252 |     13833 | 
| wikitext     |   775704 | 442426 |   103337 |     96599 | 
+--------------+----------+--------+----------+-----------+ 
2 rows in set (1 min 21.18 sec)

It looks like 20% of wikitext sessions with a "ready" have an "attempt". It looks like 24% of VE sessions with a "ready" have an "attempt". It's not clear to me that this is an issue. I'd expect some subset of Wikitext sessions to have the simple goal of "copy-paste" and that could explain the 4%. --Halfak (WMF) (talk) 14:47, 17 April 2015 (UTC)