Similar to other professionals working within data analytics, I have reverted to third party research with the hope to become more efficient at storytelling with data. To that end, I have recently participated in a webinar on “Using Contrast for Effective Data Visualization” led by Cole Nussbaumer Knaflic, the author of the bestselling book ‘Storytelling with Data.’ While I expected to learn more about effective storytelling using the support of robust data, the webinar mainly concentrated on how to use different color schemes for graphs and bar charts in order to make the data more digestible. Certainly she must address strategies for good storytelling with data in her book, and she just ran out of time in the webinar, but she is not the only one in this particular space who is focusing on such questions that should be secondary. While working in data analytics for the past three years, I learned that there are many questions to be asked while using data to support an argument and a few particular questions are more critical than simple aesthetics.
What is the argument? Every story must have a basic structure that includes a good introduction, an argument, supporting evidence for the argument, and a conclusion that will circumvent back to the introductory thoughts. When starting a research project that preludes a dataset, or not, the starting point should be constructing a hypothesis. A good analytical researcher will then conduct a preliminary background check on the hypothesis and any preexisting research and will be able to form an argument even before developing and analyzing the dataset. The data should serve as support for the argument and never actually tell the story by itself. A similar process applies to creating a client presentation. The visualization of the presentation is determined by the message that is being conveyed to clients, be it to reduce costs or adopt best practices in supply chain and procurement. When helping clients to reduce indirect spend, and before analyzing data, we perform our due diligence research, interview stakeholders and attempt to understand the issue at hand. And only then, after we have done such preliminary research and have gotten to know the spend story of our client, we analyze and present the data to prove or disprove the initial hypothesis.
Who is the audience? Storytelling, with or without data, is dependent on the audience. Knowing the audience, will empower an analyst to better understand not only the story to be conveyed but also the manner in which it should be presented. Changing the color scheme of a chart or bar graph just to appeal to the eye will not covey the message more effectively if there is no regard for who is sitting in the audience. The size of the audience, the background, and the expectations will all shape the presentation of the story and the data supporting it. Presenting data to a group of bankers, a group of supply chain officers, or a group of professional women requires three different strategies that will lead to different visuals respectively. Knowing the audience also indicates if and how much time should be spent on visuals. When communicating to a client a strategy that will save a substantial sum of their spending, a visual might not be needed at all - besides the raw pivot table summarizing the savings (and it is irrelevant if the pivot is purple, green or blue). On the other hand, if we are presenting to a group of graphic designers, the color scheme might be relevant. Thus, instead of starting a presentation with questions on data visualization, we should first understand who we are speaking to.
What data is relevant? Even when there is an argument and the audience and its expectations are widely known, data should not be doing the storytelling. Data cannot be relied on solely to relay a message simply because too much is left to individual interpretation. There are very few and exceptional cases when after analyzing the data, one sees a clearly cut and indisputable trend. Most datasets can be cut in too many ways and thus tell very different stories just by changing a small variable. Even when giving the same data, and posing one single question, to different data analysts, results may vary drastically. Consider the recent debacle exposed by The New York Times in the article “We Gave Four Good Pollsters the Same Raw Data. They Had Four Different Results.” When the same data leads to different results, how can we let the data do the storytelling? It takes years of research to refine a dataset and reduce to a minimum any margin of error. Even then, the margin of error will always persist. However, that does not mean that we cannot tell our story and support it with the data as evidence. We just have to be honest about who is telling the story, what is important in our story, and not let the data do the storytelling on its own.
What is the argument? Every story must have a basic structure that includes a good introduction, an argument, supporting evidence for the argument, and a conclusion that will circumvent back to the introductory thoughts. When starting a research project that preludes a dataset, or not, the starting point should be constructing a hypothesis. A good analytical researcher will then conduct a preliminary background check on the hypothesis and any preexisting research and will be able to form an argument even before developing and analyzing the dataset. The data should serve as support for the argument and never actually tell the story by itself. A similar process applies to creating a client presentation. The visualization of the presentation is determined by the message that is being conveyed to clients, be it to reduce costs or adopt best practices in supply chain and procurement. When helping clients to reduce indirect spend, and before analyzing data, we perform our due diligence research, interview stakeholders and attempt to understand the issue at hand. And only then, after we have done such preliminary research and have gotten to know the spend story of our client, we analyze and present the data to prove or disprove the initial hypothesis.
Who is the audience? Storytelling, with or without data, is dependent on the audience. Knowing the audience, will empower an analyst to better understand not only the story to be conveyed but also the manner in which it should be presented. Changing the color scheme of a chart or bar graph just to appeal to the eye will not covey the message more effectively if there is no regard for who is sitting in the audience. The size of the audience, the background, and the expectations will all shape the presentation of the story and the data supporting it. Presenting data to a group of bankers, a group of supply chain officers, or a group of professional women requires three different strategies that will lead to different visuals respectively. Knowing the audience also indicates if and how much time should be spent on visuals. When communicating to a client a strategy that will save a substantial sum of their spending, a visual might not be needed at all - besides the raw pivot table summarizing the savings (and it is irrelevant if the pivot is purple, green or blue). On the other hand, if we are presenting to a group of graphic designers, the color scheme might be relevant. Thus, instead of starting a presentation with questions on data visualization, we should first understand who we are speaking to.
What data is relevant? Even when there is an argument and the audience and its expectations are widely known, data should not be doing the storytelling. Data cannot be relied on solely to relay a message simply because too much is left to individual interpretation. There are very few and exceptional cases when after analyzing the data, one sees a clearly cut and indisputable trend. Most datasets can be cut in too many ways and thus tell very different stories just by changing a small variable. Even when giving the same data, and posing one single question, to different data analysts, results may vary drastically. Consider the recent debacle exposed by The New York Times in the article “We Gave Four Good Pollsters the Same Raw Data. They Had Four Different Results.” When the same data leads to different results, how can we let the data do the storytelling? It takes years of research to refine a dataset and reduce to a minimum any margin of error. Even then, the margin of error will always persist. However, that does not mean that we cannot tell our story and support it with the data as evidence. We just have to be honest about who is telling the story, what is important in our story, and not let the data do the storytelling on its own.
While there are many articles, blogs, webinars, and even books on manipulating data and data visuals in order to become more persuasive storytellers (and all of those strategies are indeed important and valuable to keep in mind) we cannot forget about the human element in all forms of storytelling. No matter how sharp and accurate the data is, alone it cannot tell stories and persuade a group of people since people are emotional beings no matter how much we try to revert to rationality. In order to tell persuasive stories data and data visuals are necessary but insufficient on their own.
Post A Comment:
0 comments so far,add yours