Does Data-Driven Storytelling Need to Be Objective?
I've always wanted to learn how to solve a Rubik's cube. I had one when I was around 10 to 12 years old. I never succeeded. When I did, it was only by "adjusting" the stickers between the blocks. So, it did not really count.
What was the issue? I thought solving the Rubik's cube was an intellectual, logical exercise. I was wrong. Thanks to some YouTube tutorials, 30 years later, I finally learned how to solve it. It turns out you can do it by following a sequence of algorithms. There is no magic – just remembering some rules and steps (and luckily still some mental effort, at least for myself).
Magic in data-driven storytelling
I used to have a similar issue with data-driven storytelling. Let's put it into perspective, akin to me being 10 years old trying to solve the cube. You would think that data-driven storytelling is used to tell the audience an objective story about data and analysis – conclusions, assumptions, positives, and negatives. The audience should know all the facts to make fully informed decisions. This might also involve sharing analysis scenarios that proved wrong. Ultimately, the audience is supposed to be fully educated and capable of making data-driven decisions. But what if they end up replacing the stickers from block to block?
We must acknowledge that humans have a relatively short attention span, often as brief as 15 to 20 minutes. Add to that various distractions, such as incoming emails or mobile notifications, along with daily stressors busy professionals face [1]. This leaves you with a minimal time window to convey your message effectively. And the real challenge is that you might not realize when you lose your audience's attention as they keep nodding their heads and saying "Yes".
You need the shortest and the simplest algorithm to solve this Rubik's cube!
Your storytelling must be highly efficient. You must convey all the information your audience must know to make an informed decision within 15 to 20 minutes. Factor in time for questions or potential technical issues, and your window becomes even smaller. You need the shortest and the simplest algorithm to solve this Rubik's cube! Unfortunately, unlike the old cubes with stickers, the new ones often have often solid colored blocks, so there's no bypassing the process.
Consider another analogy from bestselling author Cole Nussbaumer Knaflic. She uses a metaphor of oysters and pearls to explain Data Visualization. On average, you need to open 100 oysters to find one with a pearl. Imagine a presentation where we show photos of all 100 oysters, discuss patterns on each, and finally present the one with the pearl. We contrast this pearl with the previous 99, explaining what indicated there was a pearl inside.
The same goes for analysts working with data. We sift through multiple sources, use various tools and methods, and finally find the most promising insight thanks to diligent work. This allows us to make a concluding analysis and base our recommendations on it. But when it's time to present, do we need to walk our audience through every step of that process?
No, we don't. Here's why.
Data-driven storytelling is subjective.
Period.
But what does that really mean?
To answer this question, let's revisit the oyster example with the following scenarios. Let's present the same story, but according to two different scenarios.
Scenario 1: We collect a sample of oysters, some with pearls and some without. We analyze the features of their shells, locations where they were found, and other characteristics to determine if any clues indicate the presence of a pearl. Then, we will present the summarized findings to the management of our company. The goal is to develop suggestions that optimize the process of oyster exploration, helping our company minimize search costs and reduce harm to oysters.
Scenario 2: It basically begins with the same procedure, but during the results' presentation, instead of showing a summary, we examine each oyster in our sample individually, one by one, presenting one or two oysters with pearls in the middle and discussing their specific features. Then, we continue with the remaining oysters, which do not contain pearls. We have one slide for each oyster!
Interestingly, both scenarios are valid approaches. However, only the first scenario is suitable for data-driven storytelling. The second scenario is more focused on data discovery or discussion. It may be used to form the story, which can be presented as in Scenario 1.
Is the lack of objectivity an issue?
Let me rephrase this question: Why do I believe that Objectivity is not a hallmark of data-driven storytelling?
As I have argued on multiple occasions (e.g., in the post linked below), data-driven storytelling typically presents a single, selected narrative.
Following this approach, we do not guide the recipients through the nuances of our research or show all the experiments we have conducted. While we strive to remain objective, we inevitably select one scenario, interpret it, and form recommendations. The analysis, scenario selection, and interpretation can be influenced by subjectivity, from cognitive heuristics that affect our perception to our knowledge or experience levels.
Considering all this, it is clear that we cannot label our data-driven storytelling as entirely objective. However, does this make it dishonest? The answer is certainly not!
Unless we conduct exploratory research where the audience must understand all options, it is unnecessary to delve into every nuance with our business audience. We do not have the time, and doing so would detract from our primary goal – securing approval (or rejection) of our recommendations. Therefore, it is optimal not to burden them with unnecessary details.
Despite this, we still consider ourselves (data) scientists. So, what tactics can we apply to maintain efficiency without sacrificing objectivity in presenting our research results, even to business stakeholders?
How to be subjective and stay objective at the same time?
We can consider the following techniques to ensure your data-driven presentation retains professional objectivity while maintaining a subjective narrative. Balancing these elements is crucial for delivering a compelling and trustworthy story. Objectivity ensures that your data is presented accurately and credibly, while subjectivity allows you to highlight the most relevant and impactful insights. Combining these approaches creates an informative and engaging narrative, helping your audience make informed decisions based on clear, well-structured information.
Technique 1. Structure your narrative
Follow a storytelling arc to present your data insights and recommendations. Start with setting the scene and providing context (see technique 2), build up with insights, and culminate with the essential findings and recommendations. This approach helps logically guide the audience through the data.
Technique 2. Never, never forget about the context
As I argued in one of my posts (link below), context is king of data-driven storytelling.
Context is essential because it provides background information that helps the audience understand the data's significance. Context clarifies the data by explaining the "who, what, and how," ensuring the audience interprets it correctly and preventing misinterpretation. It enhances credibility by establishing the reliability of methods and sources, reducing skepticism, and fostering trust. It also aids decision-making by aligning data with benchmarks or goals, clarifying its relevance and implications. Finally, it supports narrative flow, making the data more engaging and accessible.
Technique 3. Don't cut off your audience from the omitted research parts
Always describe the research methodology you used to set the context. Remember that not all audience members will be as familiar with data science tools and techniques as you might be. Therefore, keep your descriptions relatively straightforward but without being overly simplistic. Minimize technical jargon to what is necessary. Additionally, mention that some experiment results were omitted and briefly explain why this omission is justified. Conclude by indicating your willingness to provide further details if any audience members are interested. If available, provide a link to an extended version of the material for those who want to delve into the details that couldn't be covered in your brief presentation. It is your responsibility to say: Hey, I am highlighting this, but I also want you to know that… [3].
Technique 4. Use honest visualizations
Choose the correct type of visualization to communicate your data. I advocate for this in almost every post. Avoid overly complicated visuals, especially when discussing complex matters. Use bar charts for comparisons, line graphs for trends, pie charts for structures, and scatter plots for relationships between variables to make your data more accessible and understandable.
It's crucial to maintain honesty in your visualizations. In one of my older posts (link below), I outlined practices to avoid to prevent unintentional misleading visualizations. Ensuring clarity and integrity in your visuals helps build trust and effectively convey your message.
Technique 5. Be transparent
Clearly state any assumptions made during the analysis and acknowledge the limitations of your data. This openness fosters trust and demonstrates a commitment to objectivity, even when presenting subjective interpretation. Here are some examples of what you can say.
Assumption statements:
- This analysis assumes that the sample accurately represents the larger population. This assumption is based on the random sampling method, although we recognize that some inherent biases may still exist.
- We assume that all self-reported data provided by the participants is accurate and truthful. While self-reporting can sometimes lead to over- or underestimation, it was the most feasible method for data collection in this context.
- We have assumed that the economic conditions during the data collection period remained stable and did not significantly influence the results. This assumption is crucial for isolating the variables of interest.
Limitation acknowledgments:
- The data used for this analysis was collected over three months. As a result, any seasonal variations or long-term trends may not be fully captured.
- Our analysis relies on historical data, which may not account for recent market conditions or consumer behavior changes. Therefore, the results should be interpreted cautiously and not solely relied upon for future forecasting.
- The dataset has some missing values, which were attributed using the mean of the available data. While commonly used, this imputation method may introduce bias and affect the robustness of the results.
Technique 6. Include contextual benchmarks
Present your data within a meaningful context by comparing it to relevant benchmarks or historical data. These benchmarks might include:
- Plan: Such as budget, forecast, or business case projections.
- Past performance: Historical data from previous periods shows trends or changes over time.
- Competition: Data from industry peers or competitors to highlight our position in the market.
- Different classes of analyzed metrics: For example, comparing behaviors across various age groups, geographic regions, or customer segments.
By contextualizing data with these comparisons, you help the audience grasp the significance of the numbers. This approach highlights trends and deviations and aligns the data with concrete goals or standards. Consequently, our audience can better understand how current performance measures against expectations, past results, industry norms, or specific demographic behaviors. Be mindful, however: having too many benchmarks won't be effective [4]. The benchmarks you choose must be relevant and helpful to your audience or yourself to ensure your story is legitimate. Lastly, we should be sure we understand the deviations against these benchmarks, especially if the changes are significant (see technique 9).
Technique 7. Encourage audience interaction and feedback
Encourage the audience to ask questions and provide feedback throughout or after the presentation. This approach allows you to address any ambiguities or concerns in real-time, ensuring clarity and fostering a sense of involvement. Interaction also provides a platform to elaborate on subjective interpretations and gather diverse perspectives, which can enrich the overall analysis. Incorporating audience insights can help balance the narrative, making it more comprehensive and inclusive. Below, I crafted a small "cheat sheet" with suggestions for the questions you could ask to encourage such feedback.
Interaction and feedback cheat sheet
Open-ended questions:
- What are your thoughts on this finding?
- Do you see any other factors that might be influencing these results?
- How does this data align with your experiences or observations in this area?
- Are there any additional insights or perspectives you think we should consider?
Clarification requests:
- Is there any part of this analysis that needs further clarification?
- Do you have any questions about our methodology or data sources?
Encouraging diverse perspectives:
- I'd love to hear from different departments or teams. How do these findings impact your area of work?
- How do these results compare with your observations in your respective fields?
Prompting follow-up discussions:
- What additional information would be helpful for you to take the decision?
Feedback on presentation style:
- Is the pace of this presentation comfortable for everyone? Should we slow down or dive deeper into any section?
- Do you prefer more detailed data or a higher-level overview for this type of presentation?
And the most dreadful…
- Can you see the numbers on the slide? (typically they don't…)
Technique 8. Integrate qualitative data
Supplement quantitative data with qualitative insights to provide a more rounded perspective. Qualitative data, such as customer testimonials, case studies, or expert opinions, can add depth and context to the numbers. This integration helps to humanize the data, making it more relatable and understandable while still grounded in objective evidence. Balancing complex data with personal stories or expert commentary allows you to create a more engaging and persuasive narrative that resonates with your audience.
Technique 9. Seek feedback & and be prepared
Before presenting your insights to high-profile executives, take the time to consult with peers, business partners, or academic connections. Seek out accessible individuals who can maintain discretion and will provide objective and honest feedback. Conduct a self-audit of your results, checking for any unusual trends and ensuring you can explain them thoroughly. When preparing yourself, avoid brushing over anomalies with vague statements like, "I see, that's probably that. I will handle it." Such an approach will not suffice, especially if you face tough questions and get stressed.
My advice: rehearse your presentation thoroughly. Write down and practice explanations for any complex or tricky parts. While this doesn't guarantee success, it can significantly enhance your confidence and performance.
What is the takeaway from all that?
Learning to solve a Rubik's cube and mastering data-driven storytelling share surprising similarities (in my subjective case). Both initially seem to require pure intellectual prowess. However, they are sometimes more about following structured methods and straightforward steps. The key takeaway is that data-driven storytelling is not about overwhelming the audience with every detail but about presenting a succinct, engaging narrative that leads to informed decision-making. By structuring the narrative, providing context, using honest visualizations, fostering discussion, being transparent about assumptions and limitations, and investing time in preparation, we can create compelling stories that resonate with our audience. This approach ensures that while the process may involve subjective choices, the resulting tale remains credible and impactful.
This post was drafted using Microsoft Word, and the spelling and grammar were checked with Grammarly; I reviewed and adjusted any modifications to reflect my intended message accurately. All other uses of AI (image and sample data generation) were disclosed directly in the text.
Did you like this post?
Consider a subscription to get notified about my new stories, follow me, or leave a
