Ethics in Data Storytelling

[MeganBowers]

Data storytelling is an essential skill for analysts. Whether painting a picture of the past or illustrating likely future outcomes, analysts provide value to the business with their data stories. 

Sometimes there is pressure to tell a certain story--but data analysts are biographers, not fiction writers. This is where data ethics come in. How do you ensure you are telling an impactful story with your data without compromising integrity?

According to the Harvard Business School, there are three elements of data storytelling: data, narrative, and visualizations [1]. In this article, I will share some tips for an ethical approach to each of these elements of data storytelling.

1. Data: understand your data quality 

No data is better than bad data. There might be times when you are under pressure to deliver analytics on a tight timeline, but if you are unsure of the accuracy of the data inputs to your process, then you have to push back.   

How do you assess your data quality? With exploratory data analysis (EDA). Here are some things to check on: 

• Null values 
• Data mismatch/spelling errors
• Outliers 
• Summary statistics 
• Data types 

There are multiple tools in Alteryx that are good for EDA! The Browse tool will give you data on null, empty, and not ok data, along with summary statistics:  

The Field Summary tool can give you even more details, including plots for each numeric variable and remarks with suggestions for you to review: 

Finally, you can check out the sample workflow in the Help menu in designer for data file quality assurance:  

Completing data quality checks and cleaning the data appropriately before telling your data story to the world is important. Read this article for an in-depth look at how to tackle exploratory data analysis. 

2. Narrative: be transparent & thorough

When KPIs and charts are shown on a dashboard, you can’t see all of the calculations behind the numbers. For example, if specific data is excluded from the dashboard logic, it could be important to share that, depending on your audience. If you present to executives, they probably won’t need to know the details. But when reviewing with your business stakeholders, sharing the inclusions, exclusions, or formulas helps ensure the business logic built into the dashboard is correct.  

Any good story will spend time setting the scene—a.k.a. providing context for the plot line. In the same way, it’s the analyst's job to provide adequate context for their data stories through narration when they review findings with their audience. 

Documenting business logic and change requests inside your Alteryx workflow is a great way to maintain transparency around what the workflow is doing and why. (And if you don’t believe me, hear it from an ACE.) This can be helpful context for you to look back on when you are ready to report on the findings of your analysis. 

Example documentation inside the Formula tool 

If your analysis includes a dashboard, I have seen developers increase transparency by: 

• Adding metric explanations when the user hovers over a data point 
• Designing an “FAQ” page to answer common questions about the data and the methods 
• Documenting the code behind the dashboard 

As the owner of the analysis, you get to control the narrative. Leaning on documentation to provide the history and context of the analysis can help you craft an accurate and compelling story with your data.

3. Visualizations: be aware of data fallacies 

In written and verbal storytelling, logical fallacies undermine arguments because they don’t stand up to reasoning. These logical traps exist when working with data as well! I wrote about common data fallacies in this previous article.  

You’ve probably seen misleading charts before. Or maybe you’ve encountered studies where the sample population didn’t seem diverse enough to make the conclusions the author reached.  

A misleading y-axis on a chart (source). 

These are just a few examples of how data fallacies can mislead your stakeholders, whether intentionally or not. Understanding concepts like cherry-picking data and false causality can help ensure you are telling a compelling data story grounded in the facts. 

Correlation does not equal causation! (Source) 

Conclusion 

Data storytelling is an important skill in the business world today. As more companies seek to make data-driven decisions, data analysts need to be well-versed in how to tell stories with data in an ethical way.  

As an analyst, the data you present could show that stakeholders are getting farther away from their goals or that the gut feelings held by teams are wrong. This can lead to some... uncomfortable conversations. But at the end of the day, these conversations are necessary—a successful business isn’t run on gut feelings alone.  

To learn more about best practices for working with data, I encourage you to review the data stewardship interactive lesson by our SparkED team.  

---

Sources: [1] https://online.hbs.edu/blog/post/data-storytelling