What is data storytelling?
Harvard defines data storytelling as the ability to effectively communicate insights from a dataset using narratives and visualizations.
Why is data storytelling important?
A study by Stanford’s professor Chip Heath found that 63% of people could remember stories, while only 5% could remember statistics. Another study found that humans are mainly visual learners, retaining 80% of what we see, compared to 20% of what we read.
These findings suggest that by combining stories with visuals we can achieve strong retention of our message, a critical factor when wanting to inspire action from an audience.
How to tell a story with data
For effective data storytelling, you will need the following:
- Data
- Analysis
- Insight
- Story
Data
First off, you want to select the data sources that you will analyze. These will be the foundation of your data story and many times, getting your hands on a dataset in itself can be very exciting in anticipation of the types of analysis you will be able to do.
Below are a couple of examples of where you can find data sources:
- External data sources. You can find datasets that have already been built, such as those found for free on Kaggle or Google’s Dataset Search. Alternatively, you could also build your datasets by using 3rd party tools, APIs, or web scraping.
- Internal data sources. If you are doing a project at work, you could make the most of the tables found in your company’s database, use the BI tools already set up at the company, or a mix of both.
If you already know what question you are trying to solve or what hypothesis to test, that may give you ideas on what data sources are critical to the analysis and which could be interesting to add. These could be a mix of internal and external sources.
For example, if you work at a boat rental company and want to create a report on sales, it will be a basic requirement to have access to a historical sales table. Additionally, it may also also be interesting to add an external data source, such as historical weather, to assess the possible impact on sales, especially if you are in a market where the weather changes frequently.
Data is at the heart of every great data story, but you have to tread carefully and make sure your data sources are trustworthy and that you fully understand them to avoid problems down the road. A great story built on top of bad data is a recipe for disaster. On the internet, an error can get amplified because of how easy it is to share information.
It is always good practice to get familiar with any available documentation related to the data source you will use and to also do proper data cleaning before starting your analysis.
Analysis
This is where the real fun starts. Let’s assume you have a dataset ready, it has been cleaned, you are familiar with its structure, have a question you want to answer or a hypothesis to test, now is the time to start your analysis.
The four main types of data analysis and the main questions they seek to answer are the following:
- Descriptive – What happened?
- Diagnostic – Why did it happen?
- Predictive – What is likely to happen in the future?
- Prescriptive – What’s the best course of action?
There are multiple data analysis techniques that can be used in each of these types of analysis.
In my recent essay, How to Learn Data Analytics, I mentioned some of the educational content and tools I used to learn to analyze data and enter the exciting field of analytics.
Eventually, you’ll get familiar with the different techniques and know which is best for the type of analysis you want to do. I enjoy looking at other people’s analyses to get an inside look at how they deep dive into data. Places like Kaggle are great for that type of research.
As I perform my analysis, I like to create charts that I find interesting, save them on a doc or spreadsheet in consecutive order, and write a brief note under each explaining the main takeaway from that chart and why it is important.
This will come in handy later on when it’s time to put together your story.
Insight
If you look up the definition of insight on the internet you will find a myriad of varying definitions. Many analytics companies have their custom-made definition on display, which are often unnecessarily complex.
To make things simple, an insight is a finding within the data that challenges a common belief.
It is finding something new and exciting to tell. Many times, you’ll even feel it physically. Goosebumps may happen if you feel you’ve found something extremely new or status-quo-challenging to tell the world or your company.
Once that happens, try to add more layers of insight such as:
- Cause: Identifying the cause of what you have found
- Effect: Identifying effects caused by what you have found
- Solution: Identifying a possible solution to the problem caused by what you have found
Let’s see what this looks like in an example:
- In this example, let’s assume that you work at an Edtech startup and are in a meeting with management to review the performance of the app’s purchase funnel metrics. A recent downtick in conversion from step A to B worries management and nobody knows what has caused it. The timing of this downtick is similar to when new pricing changes were introduced on the app. Management believes the price increase caused the decrease in conversion and is considering reverting to the previous price structure in the coming days. Proactively, you decide to deep dive into the app’s recent usage data. After thorough analysis, you find that for most phone models there was no downtick, but that conversion rate on phone model XYZ has completely fallen from step A to B since two weeks ago. You go to the XYZ app store and find that user reviews during the last 2 weeks have been pretty good, same as before. You present your findings to the product team and after review, they identify that event trackers had broken for that phone model.
- Insight: The downtick in conversion was not due to the price increase, as others thought.
- Cause: The event trackers being broken on phone model XYZ caused the downtick.
- Effect: If management reverts prices, the company will potentially miss out on increased revenue, as analysis of conversion on other phone models showed that conversion remained stable after the price increase.
- Solution: Fixing the event listeners should bring the conversion rate back to normal in the dashboards and allow the company to maintain the new prices and increase revenue.
Insights aren’t only found in app data, they can be found almost anywhere and relating to any industry. For example, in this article I wrote a few years ago, the main insight was that, against common belief, interest in the surf industry had been severely declining for a decade, along with the financial health of some of the largest surf apparel companies. The finding was so shocking that it became viral and the article was read by more than 20.000 people from around the globe.
Remember the charts you were making in the previous section (see “Analysis”)? Now is the time to select the ones that best help you communicate and support the insight you have found. It might take multiple charts to do so effectively. For example, in the article mentioned in the previous paragraph I had (1) one chart showing the overall decline in search volume, (2) another chart showing how interest in surfing was correlated to interest in surf apparel companies, and (3) another chart showing how surf apparel companies’ stock price had decreased similarly during the same time.
Story
Now that you have your insights and supporting data visualizations ready, it’s time to link them together with a captivating narrative to create a story.
As mentioned at the beginning of the article, “63% of people can remember stories, while only 5% can remember statistics”.
Charts themselves mainly just show data, but narrative is the best way to tell the reader why that data is important to them.
Every time you communicate your analysis to someone –through an article, presentation, podcast, etc.. – you surely have an objective.
For example, your objective could include any of the following:
- Get buy-in from your stakeholders at work
- Convince a prospective client to buy your product
- Get readers to share, like, or comment on your blog post
Whatever your objective may be, you will have a better shot at achieving it by sharing a data story than by simply sharing data. About 12x higher chance, according to the Stanford study mentioned earlier.
The Oxford definition of the word narrative is that it’s “a spoken or written account of connected events; a story” and that’s almost exactly what a data story is, with the addition of effective visualizations to maximize reader understanding and recollection.
The key word in the Oxford definition is “connected”.
Remember the imaginary charts you created at the end of the “Analysis” section and later filtered in the “Insight” section to keep only those that best helped support the insight you found? Well, now is the time to connect them through narrative and turn them into a story.
Before you start writing the story, it’s important to consider that humans have very short attention spans, at an average of 8.25 seconds. Fortunately, there are many writing best practices you can leverage to keep your readers engaged, such as:
- Keep your story centered around one main concept. Use arguments and visualization to support it.
- Start strong. Don’t wait until the end to deliver the value. After all, we have an average attention span of 8.25 seconds and may not make it until the end. Who better to show us how it’s done than scientists? The standard scientific research paper starts with a short abstract that explains the main point and finding of the study, after which the reader can choose to continue reading to learn the full details. Of course, we are not writing scientific papers so maybe we don’t need a paragraph but instead 1-2 strong sentences to kick things off by delivering some value to the reader.
- Use language that is easy to understand and avoid fluff. Understanding the data, statistics, and charts you are presenting may be challenging enough for the reader, so avoid adding complex words and jargon on top of that.
- A picture is worth a thousand words. We’ve already covered the importance of using data visualizations a lot in this article 😉
Additionally, I urge you to not be afraid to tell the reader that a certain piece of information is key, important, or critical for them to consider. Many times, readers will even appreciate this as it gives them a heads up to put special attention to the upcoming piece of information. For example, in my best-performing LinkedIn post, which reached over 90.000 people, I specified in the copy that a certain piece of information was a “must-read” for professionals in a certain industry. Telling readers that something is a must-read might seem too pushy at first, but if you feel strongly that it is true and that they will benefit from it, do consider going for it.
