The majority (70%) of the GLAM implementations examined lasted less than 10 days while 30 percent lasted more than 30 days.
Many activities may be involved in implementing a GLAM implementation: working with the GLAM institution to freely licensed material, standardizing metadata, uploading the images with volunteers, and distributing those images on Wikimedia projects. The length of reported implementations varied widely; more implementations fell in the shortest and longest implementation length categories than the mid-length categories. The shortest implementation lasted ten hours and the longest lasted 365 days. As illustrated in the graph below, out of 20 implementations for which we have dates, most implementations (75%) lasted less than 164 days and the most common lengths were between 5 and 164 days.
The GLAM implementations reported had few participants, the typical implementation had two participant and most (61%) had fewer than three participants.
Participant counts were obtained for 23 of the implementations (50%). The number of participants in each implementation ranged from 1 to 525, with an average of only two participants  Of the total participant base (740 users), we found 1 (.001%) were newly registered user.
↑We use “GLAM implementations” here to mean time-bound implementations of partnership agreements. For example, if a partnership has been ongoing for two years, but had two agreements between August 2013 and January 2015 under which new media were uploaded to Commons, we count that here as two implementations.
↑If the participant counts were not reported, where possible the count of users that contributed to the Commons category associated with the GLAM partnership was used as proxy.
↑Program leaders had reported participation for eight implementations and we mined on-wiki data to produce counts for ten more. Where the number of participants was not reported, we used the count of people that contributed to the implementation's category on Commons during the implementation period.
↑This may be lower than the actual number of newly registered users, because some implementations reported number of participants but no participant list, so we were unable to determine the number of new users for those implementations.