Records are the rows of a table in the Relational Data Service database. These store information on your customers. With the Relational Data Service, it is possible to load more data, and of a higher complexity, to the database.
For example, you can create a dynamic, personalized email which contains products related to specific contacts. You can then upload a .csv file with relational data and, based on the logic you define to determine which contact receives which product the email is generated automatically.
Each new .csv file uploaded either overwrites or appends existing data.