CDF simplifies the process of identifying the set of records that are updated, inserted, or deleted with each version of a Delta table. It helps to avoid having to implement downstream 'custom' filtration to identify these changes. This makes it an ideal source for servicing changed data to downstream applications.
Some things to consider when working with CDF are around the the type and volume of changes. CDF works well where the changes are a mix of insert, update and deletes; CDF may not be the best option when changes are append-only. Also, CDF works well in situations where small fractions of the data set are being changed per batch; when a large fraction of the data set is being changed CDF may not be a good fit.
