Why not to try to reduce data in the Database and increase in the app?

i am new to MongoDb and quit frankly just data modeling. My question is, with the increase in cheap memory and space with the devices like smartphones why not outsource a portion of what maybe used to be stuff that had to be in the centralized database and put it into the device and aggregate with this data.
Like, if i had a document with a list of 1000, seldom or maybe even never changing, entries that i would have to query with a lookup in the mongo database, why not store this in the app hard coded or if regular updates are needed, why not store the data on the devices database with sqlite or isarDb etc, and just update this local dataset with an initial, say daily, query to the remote database? What would be the cutoff that makes sense 1000 local entries, 10.000, 100 i wonder?
With the computing power today i could also save space in the mongoDB (16mb limitation) with saving just representing numbers and decode these numbers on the device with pre saved value pairs on the device instead of saving direct human readable data(*) like “address”, “street”, “name”, etc. if i am sure about the easy and “eternal” data why not shorten everything an map on the device, with the ORM or data model?
Sorry for the what i suspect stupid questions but i really wonder and i can not find answers to this on the web, maybe because of the language barrier.

Thank you for the help and/or pointers! Best wishes!

*EDIT: well, of course human readable is better for debugging and obviously less technical dept but that could be handled with mapping in the debugging “tool” and you would not have to overboard with reducing the size of entries. Would there not also a performance benefit in addition to the saved space if one would choose less human readable raw data? But maybe MongoDb does not store redundant/repeating “key” names internally but replaces this internally with something shorter, it would make much less sense to replace that with something less human readable…