If you are just now introducing Performance Analytics, I would suggest using it with only the Platform Analytics frontend (data visualizations, dashboards, filters). And if you are using Platform Analytics frontend, I would suggest avoiding technical dashboards if at all possible and sticking to the Visualization Designer and in-line Dashboard editor, even if this means having to compromise on the ideal appearance of your frontend. Doing so also will simplify your roles somewhat.
Having said that, besides the topic on Performance Analytics roles, I would suggest reading the Platform Analytics roles doc, especially the section Role lookup by use case, which says which role you need to do which task instead of describing the roles.
The tl;dr is that a KPI creator who is mostly a business analyst needs pa_power_user or higher and the user with a deep enough knowledge of your tables to set up the KPI score collection jobs needs pa_data_collector or higher. More important than this, they should take the correct trainings from ServiceNow University. Everyone involved should take Platform Analytics Overview (formerly Performance Analytics Essentials and still a Performance Analytics training, not really an overview of all of Platform Analytics), anyone creating indicators should take Platform Analytics Fundamentals (3 days and not free), and anyone managing data collection jobs or with an administer role should also take Platform Analytics Advanced (Also 3 days and not free). These might have been replaced with "new" courses with Platform Analytics in the name, but again are really Performance Analytics courses.
