Photo by the author
# Entry
Why do people misread your data? Because they are data illiterate. This is your answer. Done. End of article. We can go home.
Image source: Tenor
Yes it’s true; in many organizations, even those that are “data-driven”, data literacy is still low. Our job, however, is not to go home, but to stay around and try to change that through the way we present our data. We can only improve our own data-driven storytelling skills.
If you want to improve how you wrap your data in a narrative using structure, anecdotes, and visual appeal, check out this guide creating an impressive analyst portfolio. Provides practical tips for creating data stories that really resonate with your audience.
Knowing all this, we can be sure that our data will be understood the way we wanted, and this is really the only thing that matters in our work.
# Reason #1: You assume that logic always wins
This is not the case. People interpret data emotionally, through personal narratives, and pay attention selectively. The numbers will not speak for themselves. You need to make them speak without any ambiguity and room for interpretation.
Example: Your chart shows a decline in sales, but your sales manager dismisses it. Why? They believe that the sales team has worked harder than ever. This is a classic example of cognitive dissonance.
Fix it: Before you show the chart, show this takeaway: “Despite increased sales activity, sales declined 14% this quarter. This is likely due to reduced customer demand.” Gives context and clearly states the possible cause of the sales decline. The sales team doesn’t feel attacked to accept the frosty fact of dwindling sales.
# Reason #2: You’re relying on the wrong chart
A flashy chart may attract attention, but does it actually present the data clearly and unambiguously? Visual representation is exactly that: visual. Angles, lengths and surfaces matter. If they are distorted, the interpretation will be distorted.
Example: A 3D pie chart makes one budget category appear larger than it actually is, changing the perceived priority of funding. In this example, the Sales slice appears the largest for perspective, even though it is exactly the same size as the HR slice.
Fix it: Stick to chart types that are effortless to interpret, such as a bar chart, line chart, 2D pie chart, or scatter chart.
In the 2D pie chart below, the budget allocation size is much easier to interpret.
Only exploit fancy plots when you have a good reason for it.
# Reason #3: Correlational causation
You understand that correlation is not the same as causation. Of course; you analyze data. The same is often not true for your audience, as they are often not very adept at math and statistics. I know, I know, you think the difference between correlation and causation is common knowledge. Trust me, this is not the case: the two indicators go hand in hand and most people assume that one is the cause of the other.
Example: A edged boost in brand mentions on social media (40%) coincides with an boost in sales (19%) in the same week. The marketing team doubles its advertising spend. However, the surge was driven by a free review from a popular influencer; the extra expenses had nothing to do with it.
Fix it: Clearly label relationships as “correlated,” “causal,” or “no proven relationship.”
If you want to prove causality, exploit experiments or additional data.
# Reason #4: You present everything at once
People who work with data often think that the more data they cram into a dashboard or report, the more reliable and professional it will be. This is not. The human brain does not have unlimited capacity to absorb information. If you overload your dashboard with information, people will skim, miss critical data, and misunderstand context.
Example: You can show six KPIs at once on one slide, such as customer growth, customer churn, acquisition cost, net promoter score (NPS), revenue per user, and market share.
The CEO focused on the tiny NPS drop, which derailed the meeting, while completely missing the 13% drop in premium customer retention, which was a much bigger problem.
Fix it: Be a slide Nazi: “One slide, one chart, one main conclusion.” For the earlier example, the conclusion might be: “Premium customer retention decreased 13% this quarter, primarily due to service interruptions.” This keeps the discussion focused on the most critical issue.
# Reason #5: You are obsessed with precision
You find that showing detailed breakdowns and raw numbers with six decimal places is more reliable than rounding numbers. Basically, you think that more decimal places show how complicated the calculation is. Well, congratulations on this complexity. However, your audience focuses on round numbers, trends and comparisons. Sixth decimal precision? Confusing. Dispersion.
Example: Your report says: “Defect rate increased from 3.267481% to 3.841029%.” WTF!? People will get lost and miss the fact that the change is significant.
Fix it: round the numbers and frame them. For example, your report might read: “Defect rate increased from 3.3% to 3.8%, an increase of 15%. A clean and easy-to-understand change.”
# Reason #6: You exploit unclear terminology
If the terminology you exploit is unclear or the metric names, definitions, and labels are not clear, you leave the door open to multiple interpretations. Also the wrong one.
Example: Your slide shows “Retention Rate”.
Detaining who or what? Half the team will think it’s about retaining customers, the other half that it’s about retaining revenue.
Fix it: Say “customer retention” instead of just “retention.” Exactly. Additionally, when possible, exploit concise and precise definitions of the metrics you exploit, such as: “Customer retention = % of customers active this month who were also active last month.”
You’ll avoid confusion and also facilitate those who may know what metrics you’re talking about but aren’t entirely sure what it means or how it’s calculated.
# Reason #7: You’re using the wrong level of context
When presenting data, it is effortless to miss context and present data that is excessively zoomed in or out. This can distort perception; insignificant changes may seem significant and vice versa.
Example: During your monthly planning meeting, you present the 10-year revenue trend. Well, kudos for showing the bigger picture, but there’s a smaller, much more critical picture behind it: there was a 17% decline in the last quarter.
Fix it: enlarge the relevant period, e.g. last 6 or 12 months. Then you can say, “Here’s the revenue for the last 12 months. Notice the decline in Q4.”
# Reason #8: You focus too much on averages
Yes, averages are great. Sometimes. However, they do not show the distribution. They hide the extremes and therefore the history behind them.
Example: Your report shows that the average customer spends $80 per month. Frigid story, bro. In fact, most of your customers spent $30-$40, which means only a few high-spending customers push the average up. Oh yeah, that campaign created by marketing based on your report, targeting $80 customers. Sorry, this won’t work.
Fix it: Always show the distribution using histograms, boxplots or percentile distributions. Apply the median instead of the mean, e.g., “The average spend is $38, with 10% of customers spending over $190.” With this information, you can significantly improve your marketing strategy.
# Reason #9: You’re overcomplicating visuals
Too many colors, too many shapes, too many labels and legend categories can turn your chart into an unsolvable puzzle. Visual elements should be visually appealing and informative; finding the balance between the two is almost a work of art.
Example: Your line chart tracks 13 products (that’s 13 lines!) over 12 months. Each chart has its own color. By the third month, no one will be able to follow a single trend. Additionally, data labels have been added to make the chart easier to read. Well, you failed! Data labels have started to resemble Jamie and Cersei Lannister – they are disturbingly intimate.
Fix it: Simplify your charts. Show the three or five most critical categories, group the rest as “Other”. Provide only critical information; not all the data you have deserves visualization. Leave something for later when users want to delve deeper into the topic.
# Reason #10: You don’t tell me what to do
Data is not an end in itself. It should lead to something, and that something is action. You should always make recommendations for next steps based on your data.
Example: You show that customer churn has increased by 14% and end your presentation there. OK, everyone agrees that employee growth is a problem, but what to do about it?
Fix it: You should link each critical piece of knowledge with an actionable recommendation. For example, say: “Customer churn increased by 14% this quarter, mainly among premium customers. Recommend introducing a retention offer for this group within the next month.” This has achieved the ultimate goal of data storytelling – making data-driven business decisions.
# Application
As a person presenting data, you sometimes have to be an amateur psychologist. You should think about the people you are presenting to: their backgrounds, biases, emotions, and how they process information.
The ten points I talked about show you how to do this. Try applying them the next time you present your findings. You will see how the risk of misinterpretation decreases and your work becomes much easier.
Nate Rosidi is a data scientist and product strategist. He is also an adjunct professor of analytics and the founder of StrataScratch, a platform that helps data scientists prepare for job interviews using real interview questions from top companies. Nate writes about the latest career trends, gives interview advice, shares data science projects, and discusses all things SQL.