Data storytelling and Business intelligence implementation
“Storytelling” is a catchy slogan. Stories are used in various areas of life, including data analysis. What is data storytelling in Business Intelligence?
Norbert Kulski: As with any slogan, there are many definitions. To put it simply, data storytelling is the art of communicating information in such a way that the recipient is not overwhelmed with the amount of data. Data storytelling means creating reports in such a way that the person using them perfectly understands what questions are being answered.
Our role, as report creators, is to engage, grab users’ attention, and make the user interested in both the message itself and the conclusions resulting from it. Storytelling with data means presenting information visually in such a way that the recipient feels engaged, and doesn’t only feel that he is looking at raw tables which are not telling him anything.
Do companies know what story they want to tell with Business Intelligence tools?
On one hand, there are companies that know what data they have and which systems they come from. On the other hand – some companies lack this knowledge and have not used Business Intelligence systems before or do not know the storytelling capabilities of the tools they have. This is why, when implementing Business Intelligence solutions, we talk to everyone individually.
How to start a conversation about the implementation of Microsoft Power BI or another Business Intelligence tool?
I start with a question about the area that the company wants to explore and what it wants to present in the report. Thanks to the experience gathered in many areas, as Business Intelligence specialists, we can also present effects which have been achieved in other companies. To do so, we present the capabilities of the tool or report templates.
At companies that have no idea what they can achieve and which are at the data analysis stage, the opportunity to learn about the capabilities of the tool is very valuable because it directs the conversation. This is when the client says: “This looks good, we would like to have this kind of report!”
What happens next?
We start to think how to download the necessary data and from which systems. In enterprises that know what data they have and what they want to see in the reports, it may be simpler and, in particular, it is a matter of making a report attractive and easy to use.
What is data visualization? Good practices and deadly sins
What characterizes an attractive report?
Today, reporting tools have a number of functionalities that allow you to design a report in such a way as not only to keep the recipient interested, but even to increase his level of interest. These are, for example, mechanisms of moving from general to specific, conditional formatting, i.e. highlighting with color elements that meet certain rules. This way, using only the color, we inform the recipient: “This is OK” or on the contrary – “This is bad”. Business Intelligence tools and frameworks make it possible to lay on trend lines or forecasts. There are prompting options and even automatic data analytics based on Machine Learning algorithms.
How does Artificial Intelligence and Machine Learning facilitate working with reports?
AI and ML are mechanisms that really facilitate working with reports! The Microsoft Power BI tool tries to understand which numbers or observations may be of interest to the recipient, on its own, with the use of algorithms. For example, the Quick Insights functionality makes it possible – this is an automated data set analysis.
In the background, Power BI analyses our data set to present interesting conclusions on dashboard, e.g. outliers. After publishing the extracted data collected in the report, the question of whether we want to see Quick Insights appears on the powerbi.com website. For example, in the case of sales, QI can show us the best- and worst-selling products. Sometimes, thanks to Quick Insights, you can come across very interesting findings.
Despite an increased level of facilitation, many users avoid reports. What are the “deadly sins” in the visualizations?
First of all, information overload and conveying the message in an inconsistent way. If we don’t have a consistent idea of how to present information, but only fill the page with a number of unrelated visualizations, the user will lose track because they simply won’t know which data to look at. They will not be able to draw conclusions about what is right and what is wrong.
Another common mistake: the person who’s creating a data report wants to present the user with the entire data set, the entire story without aggregating the data. Experts in a given field may benefit greatly when they have access to the entire data collection. But for others, however, you shouldn’t start by presenting everything at once, but instead use a top-down approach.
During the pandemic, there were many reports on COVID using aggressive colors.
The use of colors allows you to elicit emotions in the scope of data presentation, change user experience. We highlight undesirable numbers in red, which is a clear signal to users that this is not the expected value. Our brain intuitively perceives colors this way. If someone builds a COVID report using aggressive colors, they hope to build emotions: fear or the belief that it is bad. Using colors also includes a series of practices, but you need to be consistent here. For example, if you use blue in the sales report for the sales value, you should use that color on all pages.
A Business Intelligence specialist should also be a UX Designer, or even a bit of a psychologist then?
Business Intelligence combines knowledge from different fields, including knowledge about how the human brain works. Therefore, Business Intelligence is a wide set of techniques, algorithms, and fields, including in the area of UX Design good practices. All this to convey information and data story as efficiently as possible and to get the most out of it. Do you know why we shouldn’t use pie charts to present values?
It has been proven that the human brain is unable to compare and estimate the areas within a pie chart. If the visual representation is pie chart is sorted from the largest to the smallest value, this is half the trouble because going clockwise, we move from the biggest to the smallest value. Without this, however, it would be very difficult to say straight away which part of the pie is the biggest. Secondly, it may turn out that there is one dominating value and the others become almost invisible – sometimes interesting conclusions are hidden there.
Finally, if there are positive and negative values, it will be very difficult for us to show them at once on a single pie chart. There are visualization selection standards, one of which is the IBCS, which is most often used in accounting and has become more popular recently. It is worth reading about it to know which visualizations not to use according to this standard to avoid misleading recipients. A link to the IBCS standard pointing to other visuals as alternatives to the pie chart.
Self-service data storytelling system
What about self-service tools? Don’t we need to have all of this knowledge to use them?
The “self-service” slogan is very catchy. Behind it, there is a tool with great capabilities hidden behind a simple, and thereby accessible to everyone, interface. It’s true… or actually it is only part of the truth. With simple data, we can create automatically a functional report quickly and easily. But the bigger the data model is and the more dependencies there are, the more difficult it is to work with a self-service tool. Underneath, almost invisible to the normal user, all the layers are hidden. They have a big impact on whether the report is user-friendly or not.
The way we organize data collection and analytical models is important in terms of performance and the visual (meaning visible to everyone) form of the report. The modelling aspect is not always discussed in the case of slogans promoting self-service tools, and maybe this is why some users believe that Microsoft Power BI desktop will do everything for them. In fact, the greater the amount of data and the scope of information, the more knowledge is needed to organize a data model appropriately.