Imagine going to a restaurant and ordering a meal.
It arrives beautifully presented. Perfect plating. Looks incredible.
Then you take a bite and realise the ingredients are completely wrong.
Suddenly, the presentation doesn't matter quite as much.
I've started to think about dashboards in a similar way.
We spend a lot of time thinking about charts, colours, KPIs and interactivity. But before any of that happens, there's something much less glamorous that can make or break the final result:
Preparing the data.
And that's where Tableau Prep comes in.
Why does data need prepping?
Real-world data isn't always analysis-ready.
You might find duplicate records, missing values, inconsistent spellings, incorrect data types or information spread across several different files.
If those problems aren't spotted before analysis, they can work their way into our calculations and eventually into the dashboard. And a beautiful visualisation displaying the wrong number is still displaying the wrong number. So, before we start cooking up the charts...
Let's prep the ingredients.
1. Get your ingredients: Connect
First, we need the data.
Tableau Prep can connect to different data sources and bring the information we need into a flow.
But connecting to the data isn't just about opening a file and immediately starting to clean it.This is my chance to understand what I've actually been given.
What does each row represent? What fields are available? What should the data types be? How many records am I expecting?
Think of it like checking the ingredients before you start cooking. You need to know what you're working with first.
2. Check what you've got: Profile
Now it's inspection time.
One feature I find particularly useful in Tableau Prep is the Profile pane, which gives you a visual overview of the values and distribution within your fields. This can quickly reveal things that deserve a second look.
Imagine a Region field containing:
London
LONDON
london
To us, they're clearly referring to the same place.
To the data?
Those can be treated as separate values. Profiling can also help uncover nulls, unexpected categories and unusual values. It's essentially the data equivalent of checking your ingredients before putting them in the pan. Because it's much easier to spot something questionable before you've built an entire dashboard around it.
3. Start chopping: Clean
Now the actual preparation begins.
A Clean step in Tableau Prep gives us somewhere to make changes to the data.
Depending on the dataset, that might mean:
- Renaming fields so they're easier to understand
- Correcting data types
- Removing fields we don't need
- Grouping inconsistent values
- Handling nulls
- Identifying and dealing with duplicates where appropriate
The important part is that cleaning isn't about changing every imperfect-looking value.
Every change should have a reason behind it. A null isn't automatically wrong. A duplicate isn't automatically a mistake. Sometimes the messiest-looking ingredient is supposed to be there.
Understand first. Clean second.
4. Get everything into shape: Reshape
Sometimes the individual values are perfectly fine, they're just organised badly, that's when we might need to reshape the data.
A pivot, for example, can turn data stored across several columns into a structure that's easier to analyse. And if our information is spread across multiple datasets, we might need to bring them together.
A join combines related information using matching fields. A union stacks datasets with compatible structures together.
I like to think of it as getting everything into the right shape before it goes into the final dish.
The ingredients might all be good individually. They still need to work together.
5. Taste before serving: Validate
This might be the easiest step to overlook.
You've cleaned everything. The flow runs. Surely we're done?
Not yet. We need to check that our preparation hasn't accidentally changed something it shouldn't have.
Did a join unexpectedly increase the number of rows? Did records disappear? Have new nulls appeared? Do the totals still make sense? Are the values within the ranges we'd expect?
Validation is the data equivalent of tasting the food before it leaves the kitchen.You don't want the person eating it to be the first one who discovers something went wrong.
6. Ready to serve: Output
Once the data has been connected, understood, cleaned, reshaped and checked, we can create an Output step.
Now our prepared dataset can move on to the next stage of the process and be used for analysis and visualisation.
And then I can start thinking about the fun stuff.
Charts. KPIs. Filters. Dashboards.The presentation.
The dashboard might get the compliments...
...but the preparation happening behind it deserves some credit too.
Every decision we make while cleaning and reshaping data can influence the numbers that eventually appear in front of a user.
So yes, I still love getting to the point where everything comes together in a dashboard. But I've learned not to rush to the serving plate.
Because just like cooking:
A great final result starts with good prep.