What Granularity Means?
Granularity has many different meanings, but in data it refers to the accuracy of data categorisation. Granularity in data refers to the level of detail of the data. Data with a high level of granularity would have a large number of rows with information, such as individual records or measurements. Data with a low level of granularity would have a small number of individual pieces of information, such as summary data or aggregated data.
When asked “What is the granularity of this dataset” you should be thinking what is the level of detail in this dataset in terms of what each row represents. For example: if each row was to represent one transaction, then the data is at transaction-level granularity. If each row represents one customer, it is at customer-level granularity and so forth.
Another way to think about it is: “what is the lowest level of detail in this dataset?” This can help you work out whether the data has already been grouped or summarised. It also helps you spot if a join will work as expected or if you’ll need to aggregate one side first.
Why does this matter?
When joining multiple datasets together, mismatched granularity can cause problems like:
- Duplicate rows
- Incorrect aggregations
- Inflated values (row explosion)
Granularity becomes especially important when you are working with dates. For example, if one table is broken down by day and another is broken down by month, joining them without adjusting the date fields can cause rows to duplicate and totals to be wrong. This is where people often get caught out.
You can also get granularity issues in areas like people data (e.g. per employee vs per training session), product data (e.g. per product vs per product and supplier), and geography (e.g. postcode vs region).
So when looking at data (especially when joining) and asking “what is the granularity” you need to understand first what one row equals in each dataset. Understanding the granularity of your datasets will help you join data more confidently, avoid surprises, and create cleaner outputs in your dashboards or reports.