CDT vs Record Type

1. CDTs. All data in the world of programming needs to be "typed". You have "primitive" types (String, Boolean, Integer etc) that cannot be decomposed into anything smaller. And then you have "complex" types which are assemblies of primitive and/or other complex types. An "Address" might be a complex type, made up of, say, 4 lines of the address, a zip code and a country, all of which can be of a string type. You can use CDTs to map to database entities (tables, views). Note: not all CDTs HAVE to map to database entities. You can have a CDT in your application layer that is juts used locally to hold a set of related data for the purposes of processing locally.
2. Records. I think of these as "entry points for processing". This is where you can surface you key business entities for the purposes of conducting some form of processing against them. Records can have different sources of data. They can be mapped directly to a database entity, or be sourced from a "service"; or can be sourced from in-flight process instances in Appian I've never seen this last one used but I include it for completeness). Most commonly they are database entity-backed. Note that not all of you entities needs to be surfaced as records. Consider which make most sense to surface for the purposes of finding something and then launching a process against that entity.

Thank you Stefan and Stewart for the explanations.

Thank you Stefan and Stewart for the explanations.