How I decide what actually matters in a data analysis project

When I first started learning data analysis, I thought that good analysis meant analysing everything.

Every metric.

Every chart.

Every possible angle.

More data felt like better work.

Over time, I realised something quite uncomfortable: analysing everything often means understanding nothing.

Not every metric deserves attention

One of the hardest things I had to learn was that not all metrics are equally important  -even if they are easy to calculate or look impressive on a dashboard.

At the beginning, I focused a lot on what I could analyse, not on what actually mattered.

If the data was there, I felt almost obligated to use it.

Now I try to ask myself a very simple question first:

If I could only explain three things from this analysis, what would they be?

If a metric doesn’t help answer a real question, I’m okay with leaving it out - even if it means doing less, not more.

Context comes before tools

Another mistake I used to make was jumping straight into tools.

SQL queries. Power BI visuals. Calculated measures.

Only later did I realise that tools don’t create insight - context does.

Before I open Power BI or write a single line of SQL, I try to understand:

• who might use this analysis

• what decision it could support

• what problem someone might actually be trying to solve

Sometimes this means the analysis ends up being much simpler than I expected - and that’s not a bad thing.

What I consciously choose not to analyse

In some of my projects, especially e-commerce ones, there were many directions I could have gone in:

• detailed customer segmentation

• complex time-based comparisons

• advanced forecasting

Instead of trying to do everything, I focused on a smaller set of questions that felt meaningful:

• Where is revenue really coming from?

• Which products actually drive profit, not just sales?

• Are there patterns that are stable, not just temporary spikes?

There were things I deliberately skipped, not because I couldn’t analyse them, but because they didn’t add much value at that stage.

Learning to say “this is enough” was surprisingly difficult - but also very freeing.

Clarity over complexity

I’ve learned that clarity is often more valuable than sophistication.

A simple chart that clearly answers a question is better than a complex dashboard that requires explanation before it makes sense.

I no longer aim to impress with complexity. I aim to communicate with clarity.

And honestly, this shift changed the way I think about data much more than learning any new tool.

What I’m still learning

I’m still learning how to balance depth and simplicity. How to decide when more analysis adds value - and when it just adds noise.

But if there’s one thing I’m confident about now, it’s this: Good analysis isn’t about showing everything you know. It’s about showing what actually matters.