This is a principle that you will become familiar with when using Feed Donkey - but its certainly worth taking some time to understand this ahead of time if you're are a new user.
Building a data feed in its most headline sense involves building columns of data - when you first create a new header in the Feed Donkey builder first actions you will take are:
a) give the column a name
b) choose the method by which the data is populated within that column - Map or Process.
Map - in other words this means that the value that is returned in this column against the product record rows - is 'based upon' or 'derived' from store data.
The 'Map' description comes from the way that you construct this data population type. You link (or map) a value to a match condition (or set of match conditions).
In your Google Merchant Centre feed for each product you need to include a 'google product category' column.
Say you are an IT equipment retailer, and you sell a number of Micro SD Memory Cards.
Google requires that you return a value of 'Electronics > Electronics Accessories > Memory > Flash Memory > Flash Memory Cards' for each of these products.
Now depending upon how your data is set up in your store you can set rules to cover all of the products that need to be assigned this 'Google defined' category in the feed.
In the builder this may look something like this:
So in short if product is in the product collection 'SD-Memory-Cards' or it has SD Card in its title will be assigned the right value.
Every time new products are imported they are tested against these match conditions - so as long as your rules are correct your products in the feed will always have the correct Google category.
Whereas 'Map' is returning controlled values 'based up' your store data - when you select 'Process' - this is your store data, in its original form or transformed (Processed) in some way.
You are a retailer of mobile phones. You are creating a feed to populate an international channel. You hold a range of phones and you want them to appear as 'Cell Phones'. This column construct might look something like this:
There are lots of options of how to 'transform' your column data - find an replace, extract certain portions, truncate to certain number of characters. You string these 'processes' together as well.
Note: the Process column type also includes 'STATIC' which as the name suggests in a simple static value per product record.