In a data warehouse, data granularity is the level of detail in a model or decision making process. It tells you how detailed your data is: Lower levels of detail equal finer, more detailed, data granularity [1, 2]. Finer, more granulated data will allow you to perform more precise data analysis.
For example, time-series data for sales volume can be measured in years, months, weeks, or days — with days as the lowest level of granularity. Performing data analysis on the more granular daily data will result in better insights on sales than yearly data.
Data Granularity in Computing
Data becomes more granular as it is subdivided. For example, an address might be contained in a single field or subdivided into individual parts: street, city, and zip code. The more subdivided the unit, the more granularity exists.
Data granularity also reflects the amount of data exchanged in a service:
- Data sent to a service is called input data granularity.
- Data returned by a service is called output data granularity.
References
[1] Bellahsène, Z. (2008). Advanced Information Systems Engineering. 20th International Conference, CAiSE 2008 Montpellier, France, June 18-20, 2008, Proceedings. Springer Berlin Heidelberg.
[2] Ponniah, P. (2004). Data Warehousing Fundamentals: A Comprehensive Guide for IT Professionals. Wiley.