An inefficient Content Review Process
By Piyush Lohana
A social networking site called Google+ owned and operated by Google Inc., an American multinational technology company specializing in Internet related services and products has been running its content review process across various locations worldwide, one location among them is Hyderabad, India.
During my service over a period of 1 year, I encountered and experienced an observable event which affected the efficiency, accuracy and user experience of the product at large. The content review process run by the Content Review Team (CRT) department is responsible for manually reviewing photos, profile pictures, posts, comments, pages, names and hashtags of its users from their day to day activities. The tool has a separate queue for each type of content as mentioned above. The business requirement for most of the queues to be reviewed is from few minutes to an hour and the expected quality is 100%. Also, the tool follows a FCFS (queue) priority approach for the reviews which mean the next photo or posts in the queue will be reviewed only after reviewing the preceded review and also a single review is reviewed by at least three reviewers before passing it on to the second stage or final judgment.
Due to not having inspected 100 % coverage on some days, especially on weekends and public holidays, when only half of team works and the priority of work (queues) on those days is very specific, i.e., the team is made to work on high priority queues like posts and photos as a result many reviews in other queues like profile pictures and comments keep getting accumulated in the queue bucket and are only worked on or cleared (emptied) on Mondays when the team lead realizes the need and pushes his entire content review team to put extra efforts to clear the queues, delaying manual review process by some hours which ultimately affects the quality of product and user experience as the review is still alive(visible) to public even if it might have/had violated the social networking site acceptable/allowable content policy.
For example, a user posts a nude image or say a hate speech as his profile picture, as per the general content policy it should be reviewed and blocked so that it is not visible on the site but due to not having enough coverage on weekends and some special days the review would be still alive and visible to public until it is taken down on the full coverage day like Monday after weekends.
The direct implication of this is over a period of time the user migrates to a more clean, well-managed and efficient social networking site which gives him a good user experience so he don’t have to see unwanted bad content while using the site.