There has been published little research on mobile apps yet, except for Pelle Snickars’ exellent anthology Moving Data.
One initial problem facing any researcher is the difficulty of getting an overview of the offerings. With over 500, 000 apps for iPhone and iPad and a similar number for Android, a qualitative view becomes impossible.
Quantitive studies are also difficult. One could collect numbers for the different genre categories in the App store, as is done by the web site 148Apps.biz with data from Apple’s Enterprise Partner Feed. These categories seems to reflect Apple’s intentions rather than the actual uses of apps, however. The education category, e.g., is subdivided into a total of 29 sub-categories, many of which contain fewer than 10 apps. Many other categories are so wide, however, that there is large variation within each category. Many apps are also placed in categories that seem poor fits, probably because it is the developer of each app who registers the category for the app.
The only study I have found that tries to measure the actual variation is the iLearn study, which coded a sample 200 apps from the more than 70, 000 in the «Education» category. This coding was rather simple, and made solely by the description of the app in Apple’s App Store and elsewhere. I have seen no large studies where the researchers actually installed the apps and looked at them.
Jakob Nielsen reports that his company over the last few years has watched 350 users perform 194 tasks on mobile apps. He does not state how many apps these tasks where performed with, but it may be significantly fewer than 194 (for the first study 34 apps and sites where studied, according to Nielsen).
It seems that a couple of hundred apps is about what researchers have the resources to look at in a coherent study, and obviously this is a limited window into the app world. Apple’s App store and Google’s Android Market have already grown so much that the situation is like the Web. It is to diverse for us to know what is actually there, as even simple metrics are hard to collect