Blog

How to Tell Stories with Data: 5 Steps to Make it Work 

13 minutes read

We all share a love for data and the clarity it brings. Some use data to make autonomous decisions about the future of our teams or company; others make sure that data is governed, compliant, and fit for purpose; others still build data pipelines to bring that data to those who need it.

Whatever our role is, we often analyze data, create charts and reports, and make presentations. Unfortunately, traditional slide decks are hard to maintain, while BI reports could be overwhelming for a new audience. Both can fail to keep your audience engaged.

There is a better way, though: data storytelling. It combines the precision of data with the art of storytelling to present facts in the context of a larger narrative. This article will walk you through the steps you need to prepare your data and best practices to create and tell an engaging data story.

What is data storytelling?

Data storytelling visualizes insights from your data in the context of a narrative. It uses classic examples of data visualization like charts and tables but is presented in an intentional order to guide the viewer along with a chronicled account of what data you collected, what it showed you, and what you think your viewer should take away from the presentation.

A data story can be as simple as two charts stacked on top of each other or a complex network of interactive graphs, connecting text, different versions of the same chart, comparison images, etc.

Data storytelling vs. data visualization

What separates data storytelling from a typical data visualization is that there will be a theme connecting the images (sometimes with text explaining the connection) that is clearly stated and understood by both the presenter and the viewer. With data storytelling, you are trying to convince your audience of something, while with data visualization, you are just presenting data for it to be more easily understood.

Data visualization:

This month every office reached its target. We have enough money to open a new office.

Data Storytelling:

As we can see, every office reached its target this quarter after we put out our new incentives program. Here is what we expected, and here’s how much we made. The money we made went far beyond our expectations. Now, we have the resources to open a new office. Here is how much we can expect to grow our revenue with this new office…

Step 1: What question are you trying to answer?

Start a data analysis storytelling project with a research question. Ask yourself: what gap in the world (or my audience’s knowledge) am I trying to fill? Then you can begin searching for data that can answer it. Alternatively, as a data analyst, maybe you were tasked with answering a specific question.

For example, you may want to know why there was a considerable increase in sales last year. You can then search through your data catalog to find all data sets containing sales reports, new product launches, department performance numbers, etc.

Who is your audience?

Another critical factor in starting a data storytelling project is finding out who you’ll be presenting. You can ask yourself the following questions about your audience:

  • Who is my audience?
  • What is their relationship to the topic?
  • What type of charts are they familiar with?.
  • Are they from another business? Are they customers? Investors?
  • Did they respond well to previous presentations? Why or why not?

While these suggestions aren’t universal, understanding these factors will help you tailor your presentation to fit its purpose and audience. However, you’ll still need to understand your data before moving forward. This way, you can ensure you have all the necessary information to answer your research question and that you answered it accurately.

Step 2: Understand your data

Having a solid grasp of the data will help you expand upon your initial research question and find data sets that can answer it. It’s essential to have both a structural and a business understanding of your data.

Structural understanding

The first step to understanding your data is data profiling. This will give you helpful information about your data’s characteristics like its:

  • Structure
  • Format
  • Patterns
  • Completeness

For example, you might find some data sets with null values during profiling. Null values mean the data wouldn’t be usable because of many empty rows. You should not include that set in the story, or it must be fixed/enriched. Or, you need to find an alternative set (which you can do in your data catalog).

If you have a data quality tool, it’s even better because you can also check for:

  • Validity
  • Accuracy

After you’ve gathered all of this data quality information, you will understand which data can be used as-is and which needs data preparation. Using high-quality data for your analysis will ensure that it is accurate and representative and that your data story is valid and trustworthy.

Data Quality Basics

If you are new to data quality management and want to learn the basics, we suggest you start by reading this blog post.

Learn data quality basics

Business understanding

However, to truly understand its purpose and generate valuable insights, you need to understand the business meaning of data. For example, when the number 100,000 is written in the "sales" column, a handful of questions arise:

  • Does it refer to the yearly sales? Monthly? Weekly?
  • What is “sales” anyway?
  • If “sales” = revenue, which regions does it cover?
  • In which currency is it accounted for?

Having a business glossary is a massive help because you can assign terms to data, making it easier to sort and understand. Once you know the structure, quality, and business perspective, you can begin preparing the data for your story.

Step 3: Prepare your data

After profiling, if some of your data does not meet your expectations (invalid data, incorrect formatting, missing entries, etc.), they will need data preparation. It usually includes processes such as:

  • Format standardization
  • Parsing
  • Enrichment
  • Deduplication

Here are some examples of data preparation for data storytelling:

  • Data in one or several data sets contains inconsistent formatting in the same column, for example, 80% and 0.8. You would need to standardize them to perform analysis and build charts.
  • When working with multiple data sets, you may discover that one of them is incomplete. For example, a monthly spending report is missing one country or region. You may have to find missing data in a different data set or request this information from finance and fill in the gaps manually.
  • You might have to join several data sets together to have complete data for your analysis and presentation.

Step 4: Build your data story

Before you can build the best data visualizations, you need to understand a little bit about how most people consume charts and other data visuals. The common belief is that everyone reads bar graphs according to their details and characteristics, focusing on the metric measurements of a bar graph or the angles of a pie chart.

In reality, people have much more of a “big picture” mindset regarding charts. They want to know what the chart is saying and how it’s saying it almost automatically. Data visualization expert, John Burn-Murdoch, provided us with some data storytelling tips on how to achieve that.

#1: Present information in the right order 

The order you choose is vital and will be driving the message behind your visualization. You don’t want to send out individually isolated charts in no particular order. The first thing you can do is infer follow-up questions your viewer might ask once they’ve been given a piece of information.

  • For example, your previous chart was about how introducing a new office impacted revenue for that quarter. Maybe people will ask, “Well, what did that office do differently than others that led to this increase?” Your following chart could then depict that office’s regular activities vs. those of other offices.

Another way to strategize your story is by following the pattern of your research. After you found your research question, which data sets did you analyze to get your answer, and in what order? The chances are that whatever path leads you to your conclusion is also a good path for your story to follow.

  • For example, your question was: Which office had the best sales this quarter and why? So, you dug into the sales reports: each office’s reported hours, staff's measurables, resources, and previous sales numbers. Then you can just set up charts for those tables, 1, 2, 3, 4.

#2: Use layering to unveil your story gradually

You don’t want to overload your audience with too much information. People can appreciate the visual aesthetics of a chart and still not understand it, especially when presented with a misleading data visualization. That’s why it’s sometimes better to have three charts appear consecutively instead of mapping them all onto the same graph.

It’s essential to hold the viewer’s hand through the presentation. For dramatic effect, you can even use different slides or animations to indicate significant movements of the data. You can begin with a line, add the annotation to show why that line dipped or peaked in a specific place, and move on to the following key takeaway or area of interest.

You also don’t want to have a big jump in the charts without explanation. Think of your charts as frames in an animation. Each needs to flow into each other and progress with the story as a whole. It might be better to move incrementally through a 50-year gap than jumping from a five-decade-old set of data to a set from this year.

One of the most powerful data visualization tools is data change over time. Remember that you can show the difference more dramatically by:

  • Changing Charts to visualize drastic contrast

  • Showing one visual first and then introducing the second line (or variable) later on.

  • Finally, you can guide your audience into the drama by creating an expectation (see the question on the chart below) and promptly shattering it.

#3: Use text to add context

One of the best places to start building your visuals is with text. A study by American data scientists pointed out that the first place people’s eyes go when looking at a chart is the title. The following two places are the labels and any paragraph text or annotations included in the graph. The actual type of graph is only recognized fourth.

So what kind of text should you include in your story? Here are some examples.

  • The title. Your title should illustrate the primary purpose of the chart immediately. Simply calling it “Data Quality Reports from 2020” will do no favor to your viewer. Instead, you should call it “Data Quality Declined in 2020 and will Decrease More in 2021.” This way, the people already know what the graph is telling them, and they can move on to observing the data that verifies it.
  • Subtitle. This is where you can describe the graph's data to avoid any confusion about what is being measured.
  • Annotations. Annotation is an excellent way of explaining why a change happened in the data or why certain factors look different from others. You can label a dip in a line graph with “Financial Recession” or mark the smallest bars explaining why they’re so low, like “This branch lost 50% of staff after COVID-19.”
  • Linking Text and Hero Images. Sometimes text in between charts can be equally helpful. Whether it’s a meaningful average or a connection you don’t think is well displayed on the charts, you can use text and a variety of shapes and colors to single it out and make sure nobody misses it.
    • If you’re unsure what your key insights should be, Ataccama’s Data Stories can suggest them to you through AI. It might notice that one of the averages is higher than other similar sets or recommend supplying the maximum and minimum records for a particular data set.

#4: Do not disregard visual aesthetics

While every chart doesn’t have to be comparable to the Mona Lisa, there are ways to add visual elements that can help build your narrative:

  • Changing colors in a chart. You can use color-coding to improve your charts in many ways. You can color-code various data segments to show differences between them or use a change in color to indicate if a result was positive or negative.
    • E.g., changing a section of the graph from blue to red to indicate a problem occurred.
  • Visual Components. Symbols are another great way to communicate your narrative. You can add them at the end of a line graph, on an additional axis, at the top of bars, or anywhere you like to indicate something the audience might not be able to see. For example, you can put a question mark at the end of the line graph to show the future is uncertain.


#5: Choose the right visuals

The people who receive your data story will scroll through your thought process step by step. Whichever charts you choose will be interactive, and viewers can drill down, cross-filter, check exact values, and even follow the values on a line by hovering over it.

But how do you choose the type of visuals you want? Our data visualization experts say that a bar or line graph is sufficient up to 80% of the time.

Selecting which visual is best for your story is really up to you. Some factors you’ll need to consider are:

  • The story you want to tell. Did you collect your data based on trends, correlations, or something else? Do you want to show change over time? You may consider using line charts to illustrate correlations and changes over time.
  • Who you’re presenting to. Is your audience very knowledgeable? Have they seen this data before? Is there a newer/better way to explain it this time? For example, if you don’t have an expert audience, you need to prioritize simplicity and text over complex layouts.
  • The volume of data you have. Some data volumes are too big for certain chart types. If you want to use a pie chart (but we recommend you don’t), you probably shouldn’t choose figures with too many data sets because it will be hard to see all of them.


source

  • The data type. There are many different types of data, and you can present each of them with a particular chart type. For example, line charts will be better than a bar if your data is continuous instead of categorical. The reverse is also the same because you need a finite number of categories for categorical data, and line charts are continuous.

You can read more about the seven most used chart types, what they’re best for, and when to avoid them.

Step 5: Share

Now that your storytelling data visualization is finished, you probably want to share it. You can upload the story to social media, embed it on your website, or send a link to it. You can also record it as a video to make it even more accessible.

Tell your data story with Ataccama

If you want to illustrate your stories vividly, give Data Stories a try at datastories.ataccama.com. It connects directly to your sources, has a wide assortment of visual options and customizable animation, offers AI insights to help you tell your story, and has an easy and secure publishing option.

Start telling stories with data today

Ataccama Data Stories are free.
10 public stories • 10 datasets • 10 million data rows

Start telling data stories