In the case of small data (you can load it all into memory), simple
analysis (maps well to your statistical
package of
choice), and no plans for you or anyone else to repeat your analysis
(nor receive updated data), then keeping your data in text files,
and using a scripting language like Python or R to work with it, is fine.
In the case that you have a large amount of diverse data (cannot all
be loaded into memory) that may be updated, or if you want to share
your data with others and let others easily reproduce your analysis,
then use a DBMS (Database Management System). DBMS are
important for storing, organizing, managing and analyzing data. They
mitigate the scaling and complexity problem of increasing data
in volume and diversity. DBMS facilitate a data model that allows
data to be stored, queried, and updated efficiently and concurrently
by multiple users. In general, as a data scientist your toolkit will
involve using SQL (Structured Query Language) with a database and
something else--python, R, SAS, Stata, SPSS. This tutorial covers the
basics of relational databases and NoSQL databases, the pros and cons
of each type of database, and when to use which one.