Another use that none of the previous answers seem to have mentioned is easier deployment of table structure changes.
Say, you wish to retire a table (T_OLD) containing data for active users, and instead use a new table with similar data (named T_NEW) but one that has data for both active and inactive users, with one extra column active.
If your system(s) have gazillion queries that do SELECT whatever FROM T_OLD WHERE whatever, you have two choices for the roll-out:
1) Cold Turkey – Change the DB, and at the same time, change, test and release numerous pieces of code which contained said query. VERY hard to do (or even coordinate), very risky. Bad.
2) Gradual – change the DB by creating the T_NEW table, dropping the T_OLD table and instead creating a VIEW called T_OLD that mimics the T_OLD table 100% (e.g the view query is SELECT all_fields_except_active FROM T_NEW WHERE active=1).
That would allow you to avoid releasing ANY code that currently selects from T_OLD, and do the changes to migrate code from T_OLD to T_NEW at leisure.
This is a simple example, there are others a lot more involved.
P.S. On the other hand, you probably should have had a stored procedure API instead of direct queries from T_OLD, but that’s not always the case.