As the “data age” emerges day by day, data analysis abilities become more needed for most companies. If you work in a data-drive organization – or even if your company is thriving on it – you probably have had contact with a dashboard or a data visualization tool.
Data visualization is the ability to translate data and information into graphic representations, which can then be interpreted easily by people with no background in data. It is a significant connection between product-oriented and business-oriented roles inside a company, which can ease decision-making in a well-founded way. Data scientists also use data visualization (known as datavis) as part of Exploratory Data Analysis, generating insights from the data to improve the model’s performance.
Even though data visualization tools have been developing in the past few years, they can’t sustain themselves without human interference. The human eye with a specific business perspective plays an essential role in a data-driven culture.
Data always has a story behind it that the tools you use can’t comprehend, and we are responsible for translating it. It’s like giving life to a story.
For a good analysis, we must take this simple quote as a rule:
A beautiful chart no one understands is just abstract art.
The main focus of data analysis must be communication, which means being able to express an idea with no need for inventing extraordinary and overloaded graphic representations. The founder and author of Storytelling with Data and former Google manager Cole Knaflic once said:
“Beyond annoying our audience by trying to sound smart, we run the risk of making our audience feel dumb. In either case, this is not a good user experience for our audience.”
Furthermore, for you to start, you can try answering some simple self-questions:
By answering these questions, you will be able to start your analysis and guarantee your message will be understood by whoever you’re talking to. After all, it’s all about communication.
Some tips that you can use for good data visualization are:
It’s all about communicating. Take some time to understand your audience and your goals with this analysis. Depending on them, you should decide your approach: a more technical one, or a straight-forward one – like using quantagrams and typography. For example, you can show a pretty confusing heatmap to present a correlation between features, or you can stand out what matters.
You are not supposed to be an artist, but you need to be a good communicator. When in doubt, stick to the simple (you don’t need to show the entire table, bar charts can get the job done).
Preattentive attributes like color, size, and position play a significant role in a chart’s readability. Good usage of these attributes is key to success since you can quickly draw a logical order for the audience’s eyes. Take a look at these two representations from the same data and compare all preattentive attributes:
If you want to have some fun, look at this blog post about bad data visualization to see what NOT to do.
After you’ve read about some tips for a good data analysis, take a look at this excellent representation of Manhattan at working and non-working hours throughout a workday, created by Dark Horse Analytics:
Besides presenting the population’s movement concisely inside Manhattan on a workday, this graph contains many design elements called “preattentive attributes” that make it easier to read and understand. During work hours, you can see that Manhattan concentrates a considerable part of the “work” category in central areas such as Times Square. Meanwhile, the “Home” category shows up off the work hours in more peripheral areas.
This representation requires good data-wrangling abilities, but it stands out for its simplicity and communication effectiveness. Most of the time, the information consumer is not interested in knowing all technologies you have used to build it; he is most interested in the information itself.
With this article, you should have a good comprehension of an exemplary data analysis. You have seen relevant tips to analyze data and some good examples. Furthermore, you have learned some of the abilities you should have to perform satisfying analysis. By the way, one of the best books on datavis – and a must-read for data analysts – is “Storytelling With Data” by Cole Knaflic. I strongly recommend it if you want to dive deeper into data visualization.
Indeed, there are plenty of tools and ways to perform this job. Many of them can suit a specific necessity so that the tool will depend on the user and his need.
I can even list some that don’t require much coding experiences, such as Power BI, Tableau, and Google Charts. As some of my experience with exploratory data analysis with Python, I can list some beneficial python libraries, like Matplotlib, Seaborn, Plotly, and Folium.
One of my favorite Python libraries for exploratory data analysis is Matplotlib, which I will show you in detail in the next blog post – where we can build up together some excellent visual representations of data using these tips.
I hope you enjoyed reading a bit more about datavis and hope to see you in the next blog post.