For the Western Name Authority File, one way of evaluating changed metadata was developing a script that delivered baseline statistics about name representation at an early stage of the project, which was then repeated at a later stage of project development. The script uses the python requests library, along with a csv of representative names. Consider taking snapshots of your data right after you've gathered it, and after you have had time to work with it more to document improvements and change.
Simple DPLA Data Gathering Script
If gathering statistics over an API doesn't work with the nature of your project, having csvs or other files of your vocabulary metadata and using a Diff visualization tool is another option to compare the changes in your vocabulary.
Check in with the people who might be using your new vocabulary: