Have you ever opened your banking app and seen this?
£657.63 spent this month.
Useful? Absolutely.
But then you start asking questions.
How much did I spend on coffee?
How much went on groceries?
Did I really spend that much eating out?
That single number suddenly isn't enough.
To answer those questions, your banking app needs every individual transaction it has recorded over the month. Each coffee, supermarket trip and online purchase is stored separately, allowing the app to group them together in different ways whenever you want.
That idea is one of the most important concepts I've come across since starting my journey in data analytics.
It's called granularity.
So, What Is Granularity?
Granularity is simply the level of detail at which data is recorded. Every row in a dataset represents something. The categorical fields describe what that record is, while the measures capture the values associated with it.
A way I like to picture it is with sand.
Fine grains of sand represent highly detailed data. Every grain is separate and tells its own story.
Now imagine replacing those grains with pebbles.
Then rocks.
Eventually, the whole beach becomes a single object.
The information has become less detailed because individual pieces have been grouped together.
High granularity means lots of individual records. Low granularity means those records have been summarised. Neither is right or wrong, it depends entirely on what you're trying to achieve.
Why It Matters
The questions you can answer are limited by the detail you choose to collect.
Imagine you own a café. At the end of every day you record only one number:
Today's sales: £2550.
A month later you decide to find out your best-selling drink.
Unfortunately, you can't. You never recorded what customers actually bought.
Now imagine recording every transaction instead. Each coffee, tea and pastry becomes its own record.
Suddenly you can answer questions like:
- Which drinks sell best on weekends?
- What time of day is busiest?
- Which products are usually bought together?
Exactly the same business. Completely different insights. The only thing that changed was the granularity of the data.
Increasing and Decreasing Granularity
Granularity can change in two ways.
The first is by adding more detail.
For example, instead of recording just total sales, you might also include the store location, product, employee, payment method and customer type. Each additional category makes the dataset more detailed.
The second is through aggregation.
Instead of looking at thousands of individual sales, you calculate daily totals, weekly averages or monthly summaries. The dataset becomes much smaller, but also less detailed.
The Trade-Off
This is where things become interesting.
Highly granular data gives you flexibility. You can filter it, group it, investigate trends and answer questions you hadn't even considered when the data was first collected. The downside is that detailed datasets require more storage, more processing power and can take longer to analyse.
Less granular data is simpler, faster and often easier to understand. But there's one major catch, you can always turn detailed data into summaries, you can never recreate detail that wasn't captured in the first place. Once you've combined hundreds of records into a single total, those individual records are gone.
Why This Matters More Than Ever
Years ago, businesses often had no choice but to summarise data. Storage was expensive, processing power was limited and querying huge datasets could take hours. Today, cloud computing and modern databases have changed that completely. Therefore capturing every website click, every purchase and every customer interaction is now not only possible, but often expected.
The challenge has shifted from Can we store it? to Should we store it?
Choosing the right level of detail has become just as important as collecting the data itself.
My Biggest Takeaway
Before joining the Data School, granularity sounded like another piece of technical terminology. Now, I see it as one of the foundations of good analytics. It influences every chart you build, every dashboard you create and every question you can answer.
Whenever I open a new dataset now, the first question I ask myself isn't What can I calculate?
It's much simpler.
What does one row represent?
Because once you understand that, you're already well on your way to understanding the data itself.