Not all semantic models are born equal. Some achieve greatness, others die young. Organizations with multiple Power BI developers should consider certifying their best semantic models to improve the overall level of their developments.
A certification gives users confidence in the quality of the data they are looking at. It also gives the manager confidence in the quality of the work done by its team. These benefits remain the tip of the iceberg.
The certification process is above all an opportunity to identify, promote and even impose good development practices. Whether or not the analysis of a semantic model leads to its certification, discussing good practices with its developer is a opportunity to learn and to identify improvements that will raise the general level of quality of developments.
How to certify a semantic model
Firstly, organizations implementing a certification program must identify the good practices they want to promote and formulate them into criteria. Then, a certifier uses them to analyze existing semantic models and to certify the best ones.
It is possible to attribute points each criterion. This overall score can be monitore at the semantic model level, as well as company wise.
My two cents
Here is the list of the semantic model certification criterias that I use. You may complete them with the specificities and preferences of your organisation.
Your comments and suggestions are more than welcome!
Organization
The owner of the data or process desires the certification
The semantic model is monitored, maintained and improved over time
.pbix files are saved in a secure location that logs versions
The semantic model is independently tested before being put into production
Reporting and semantic model development tasks are separated
Security groups manage access to the semantic model in Power BI Service
The data source is authoritative
e.g.: Avoid Excel
Development
Columns and measures have appropriate formats
Tables, columns and measures are named with business rather than system terminology
Unnecessary columns and data are purged
Data transformation is done as much upstream as possible and as much downstream as necessary
Data is refreshed adequately and efficiently
Ex.: incremental refresh, DirectQuery, composite, etc.
The tables, columns and measures of the semantic model are documented
Row-level security RLS and OLS enable secure use of data for anyone who needs it
A measurement table and folders help classify data and measurements
Learn more about certification and best practices
Comments