top of page
  • Writer's pictureJohn Kirby

Kimball is here to stay

Kimball is not just a star schema. It's a methodology to capture business requirements and create a star schema around them. There are techniques to the business requirements gathering, data analysis, and technical implementation, which are specifically part of Kimball such as the business matrix. There are other aspects of Kimball which are more generic, like partition switching for fact loads.

Knowing what a star schema is is not the same thing as knowing Kimball. Star schemas can be implemented without going through the Kimball methodology. But, they will most probably be better by using the Kimball methodology.

Star schemas themselves are still part of modern data solutions like a Lakehouse. Albeit it could take the form of the 'gold' layer in a Delta Lake. Dims and Facts are well established, supported and familiar artefacts, by data analysts and analytics tools like Power BI, Tableau, Qlik, and goods old Excel.

Data Science is one use case where a star schema probably isn't the best fit. Neither is a big proportion of semi or unstructured data. Maybe Data Vault 2 is better, or just json or parquet files in a sensible folder structure. Data scientists are usually more technical, so their requirements will be different to a typical end-user who wants some reports and dashboards. The 'silver' layer of a Lakehouse may suit them better. Find out, and create the right solution for the job.

Star schema or no star schema depends on the project's end user requirements, the structure and nature of data, and end user tools. A vast majority of projects still want some reports and dashboards based on structured data.

If a star schema is a good part of the solution, then why not use Kimball to create it? It's a great methodology to identify data points and group them logically, fitting with the business domain. This is the key thing - if it doesn't fit with the business by meeting the business requirements, terminology, logic and rules, and familiarisation, how likely will the business adopt it.

So, if your project has end users creating some Power BI dashboards, whether they're 'business' staff, report writers or data analysts, the familiar star schema is still a great option to include in the data landscape. Consider creating the star schema by researching the Kimball methodology, build a business matrix, and implement the structure with the Kimball approach. It's tried and tested with the business and its users as the focus to maximise adoption.

6 views0 comments


bottom of page