Story finding

Three techniques for structuring your exploratory data analysis

Here’s a dataset. Find me the insights.

When I started working with data, I dreaded this type of request. Of course, we analyse data to extract meaningful conclusions. But how exactly do we find them? Is there a process that can save us from headaches and grey hair? I never found the magic pill that would work 100% of the time, but I did adopt some techniques over the years that make the process less painful. Turns out, exploratory data analysis is as much about intuition and play as it is about structure.

That said, here are three methods that may help you next time you go story finding.

Deduction

Somewhere in the books on logical thinking, you’ll find an introduction to deductive and inductive reasoning. Despite the scary names, we can describe them in a simple, non-scientific fashion:

• Deductive reasoning: following a hypothesis, curiosity, or idea;

• Inductive reasoning: playing around to see what happens.

I find the deductive approach to be quite useful in data exploration for storytelling. Often, we’re asking a question at the early stages of the process (think conflict), and that question can be a wise guide for deduction. What drove the drop in sales last quarter? Is there a seasonal trend in human rights funding initiatives? What are the most impactful ways to mitigate climate change?

If you’re feeling playful and patient, you can also try the inductive approach.

5Ws

The 5Ws method is used in journalism to ensure that news articles cover all essential aspects of a story. All you have to do is ask and answer five questions. Here’s the list, slightly adapted to data work:

1. Who: individuals, groups, customers, stakeholders, etc;

2. What: objects, products, or other types of categories that are appropriate;

3. Where: location or setting;

4. When: timing and its evolution;

5. Why: motivations, reasons, or causes behind the story (not always relevant).

I find this method most helpful when I don’t know the data well. I often make a tab in a spreadsheet for each W and create exploratory graphics around it. Usually, early in this process, certain data points or trends grab my attention, and I end up diving deeper into them.

The 5Ws could also serve as a summary analysis if you like doing descriptive overviews of your datasets.

The Kirk method

I’m giving this method the author’s name as I learned it in Andy Kirk’s book on data visualisation. By the way, Andy has just subscribed to this newsletter, so if you’re reading this, hi 👋 The technique consists of narrowing down your data for visualisation by answering the following questions:

• What’s your angle: spatial, evolution over time, relationship, etc.?

• What’s your frame: how much information will you include? all categories and timelines or just some of them?

• What’s your focus: are there specific data points you want to highlight?

I find that this approach works well for small stories or standalone visualisations. It drives us to choose a message for the graphic and then highlight it using tunable features. Simple and effective!

What’s your favourite method for uncovering insights? Have you found the magic pill?

Thank you for reading!

I’ll see you again next week.

—Evelina

---

“This course is necessary for those willing to acquire the newly valuable skills sought in the labour market: organise, illustrate, and visualise data to build stories that will serve your cause.”

What are we talking about? My three-week online data storytelling bootcamp! The next cohort starts in early June, and a few spots remain. Oh, and not to brag, but the bootcamp does have a 4.8/5 ⭐️ rating.

Use the code THEPLOT10 for your exclusive 10% discount.

---