Moving important business data into a data model for analytic reporting can often be a two-edge sword. Data retrieval is fast and can support all kinds of analytic trending and comparisons. But, data in the model may be one or two layers away from the original source data, making it more challenging to compare with familiar user reports. Often the first validation effort after transforming and loading data into the model and then visualizing the initial results is having a business user say “yeah, that looks about right.” Then, sometime later after more development and extensive user testing, the feedback might be “hmm, that looks a bit off.” …not exactly scientific.
I have been doing a lot of data validation work lately – both formally and informally. Informally: Validating calculation results from a Power BI data model is just a normal part of the development process. Formally: After developing and delivering an entire business dashboard solution, a formal validation process is used to validate the ongoing results after future data refresh cycles and to certify reports so that business leaders know they can trust them.