How Visuals can Mislead Data Interpretation

Bad data visualization is everywhere — here’s how to fix it. Understanding the essentials of effective data visualization is one thing, but witnessing poor data visualization in practice offers the real lessons. Take a look at this chart, “How Baby Boomers Describe Themselves,” which had some fundamental errors. The major problem? It disregards the rule of relativity. The design implies the data forms a complete whole, yet the percentages total 243%. This clearly indicates the wrong visual format was selected. If respondents could choose multiple answers, the data should be shown as a grouped bar chart rather than being forced into a single human figure. Additionally, contrast is mishandled: • Size contrast is deceptive – Larger sections don’t correlate with larger values. • Color contrast is excessive – Every section demands attention, causing nothing to stand out. • Shape contrast is absent – The chart depends solely on color to distinguish categories, reducing clarity. • Annotations cause confusion – Instead of providing clarity, extra design elements divert attention from the main insights. So, how to fix it? Opt for the correct visual structure, use proportional sizes, apply color contrast wisely, introduce meaningful shape variations, and ensure annotations are purposeful. Bad data visualization doesn’t just appear cluttered. It misleads. Correcting it involves directing the audience to the right insights without making it a struggle. Art+Science Analytics Institute | University of Notre Dame | University of Notre Dame - Mendoza College of Business | University of Illinois Urbana-Champaign | University of Chicago | D'Amore-McKim School of Business at Northeastern University | ELVTR | Grow with Google - Data Analytics #Analytics #DataStorytelling

𝗙𝗼𝘂𝗿 𝗱𝗮𝘁𝗮𝘀𝗲𝘁𝘀. 𝗦𝗮𝗺𝗲 𝘀𝘁𝗮𝘁𝘀. 𝗪𝗶𝗹𝗱𝗹𝘆 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝘀𝘁𝗼𝗿𝗶𝗲𝘀. Visual inspection is 𝘯𝘰𝘵 optional. Anscombe's Quartet is a classic reminder of why plots matter: Each of the four datasets has: 👉The same mean for X and Y 👉The same variance for X and Y 👉The same correlation between X and Y 👉The same linear regression line But when you plot them? 🚨Completely different shapes: ✅A linear relationship ✅A clear curve ✅An outlier dominating the trend ✅A vertical line with a single influential point Same stats. Different stories. 𝗪𝗵𝘆 ���𝗵𝗶𝘀 𝗺𝗮𝘁𝘁𝗲𝗿𝘀 𝗶𝗻 𝘁𝗵𝗲 𝗿𝗲𝗮𝗹 𝘄𝗼𝗿𝗹𝗱: 👉KPIs may hide anomalies 👉Descriptive stats can misinterpret patterns 👉Decision-makers might rely on misleading summaries What looks like a tidy trend could actually be noise. Or worse: a trap. In data science, context is everything. And 𝘃𝗶𝘀𝘂𝗮𝗹𝗶𝘇𝗶𝗻𝗴 𝘆𝗼𝘂𝗿 𝗱𝗮𝘁𝗮 is often the fastest way to: ✅Spot errors ✅Identify outliers ✅Understand relationships Before trusting any model, always ask: 𝗛𝗮𝘃𝗲 𝘄𝗲 𝘀𝗲𝗲𝗻 𝘁𝗵𝗲 𝗱𝗮𝘁𝗮? 🎯 Plot first. Analyze second. Let's make this a norm: No summary statistics without visual context... ... especially in low-dimensional data. Curious to hear from others: Have you ever been fooled by stats that looked perfect on paper but broke down when you visualized them? Drop your favorite example below. #statistics #datascience #dataviz #analytics

3 small design choices that turn your charts into liars Good data visualization is more than just aesthetics or making your takeaway obvious - it's about communicating data honestly. And whether intentional or not, small design choices can distort insights, mislead your audience, and even call your credibility into question. Here are three common pitfalls to watch out for: ➤ Not starting x- or y-axes at 0 on bar charts Bar charts rely on bar length (or height) to compare values. If the x-axis (for horizontal bar charts) or y-axis (for vertical bar charts) don’t start at zero, differences can be exaggerated, making a small change look more dramatic than it really is. Rule: Always start bar chart axes at zero to ensure accurate comparisons. ➤ Using inconsistent time intervals on line charts Line charts show trends over time, but uneven time spacing can distort those trends. For example, if one gap represents a month and the next represents a year, the chart might suggest a sudden shift or flatten a real change. Rule: Always use equal time intervals and label them clearly. ➤ Ignoring margin of error If the margin of error between two or more values overlaps, the difference might not be meaningful - but a bar chart’s crisp edges can suggest otherwise. Rule: If the margin of error overlaps, use error bars or annotations to highlight uncertainty. To recap: - Start axis at 0 for bar charts - Use consistent time intervals - Mind the margin of error —-— 👋🏼 I’m Morgan. I share my favorite data viz and data storytelling tips to help other analysts (and academics) better communicate their work.

Graphs tell many stories, most of them misleading by design. 📈 Graphs are good visual aids for communicating our “understanding” of how a process or product is affected by factors we can control. Just about everyone understands an XY graph but its most common design, the line graph, is typically misleading. This is because line graphs only show the average (mean) of the model built with the data. Plotting the model average alone leaves out essential information, which causes your audience to make wrong conclusions about what is being shown or how much to trust it. The animation below shows a couple of alternatives that I’d like to share my own opinion on: 1. The graph you see all the time: the model mean. By design, it includes no information about uncertainty. This makes most audiences take the line as gospel, meaning it causes them to expect all future results to land exactly on the line. This inevitably sets everyone up for disappointment. 2. One step up is showing that line shape is uncertain, using a confidence interval for the mean. This assumes your model was built with good data coverage, using e.g. #DesignOfExperiments. In my example, the graph makes it very evident that we don’t really know what the system does near 0. Statisticians might pat themselves on the back and think “Problem solved!”, but in my experience most people ignore the extra information and still take the mean as gospel 🤷 A second, much bigger, problem is that this graph only shows how uncertain we are about the 𝒂𝒗𝒆𝒓𝒂𝒈𝒆 of the results. When your colleagues run a follow-up experiment and get a result outside the confidence interval (for the mean), they’re often very unpleasantly surprised. I think the graph author, not audience, is to blame for this! 🫵 3. One way to align audience intuition with your point is by creating a (95 %) credibility interval and shading it uniformly. The resulting band shows us where we would expect 19 out of 20 future data points to appear. In my experience, even non-technical audiences correctly see what the graph is meant to convey, namely: a) our process has variation! and, b) while we can’t know the exact result at X = 2.5, but it’ll be somewhere between 15 and 25. 4. In the last graph, I’ve made use of a gradient to highlight that we are more likely to see future data closer to the model average than near the borders of the 95 % credibility interval. While I personally like this sophistication and think it appropriate for a statistics loving audience, it really confuses everyone else and I would caution against using it, even though it’s pretty 😉 Which graph do you prefer? #data #dataviz #graphs #statistics

If you’re using multiple charts to compare related data, ensure that the y-axis scales are consistent across all charts. It’s a very minor, yet critical element that often gets overlooked by many. Inconsistent y-axis scales can distort the perception of change or magnitude. This can lead to confusion, and worse, mislead your audiences into drawing inaccurate conclusions. For instance: A small increase in one chart might appear more significant than a larger increase in another simply because the scale was compressed. This misinterpretation can snowball into some wrong decision. Plus, if you have a skeptical audience, they might actually perceive the scaling inconsistency as a deliberate attempt to manipulate data, even if it is unintentional. You need to be wary about these situations, because, once trust in the visualization is lost, it can hurt the credibility of the data (and you, as well). If using different scales are totally unavoidable, then at least explicitly label them and include visual cues, such as annotations or markers, to clarify the differences. Above all, make sure your scaling decisions reflect the data honestly and transparently. Don’t turn visualizations into a rush job. Take the time to align scales properly, and clarify any differences where necessary, so that audiences can draw accurate insights. 

How not to visualize a ranking list. I stumbled across this chart on social media, and it’s a great example of what not to do when visualizing rankings. This looks like an AI generated chart. Polished, busy, and confidently wrong. And that’s exactly why data literacy still matters. ❌ 𝘿𝙤 𝙣𝙤𝙩 𝙢𝙞𝙭 𝙫𝙚𝙧𝙩𝙞𝙘𝙖𝙡 𝙖𝙣𝙙 𝙝𝙤𝙧𝙞𝙯𝙤𝙣𝙩𝙖𝙡 𝙧𝙖𝙣𝙠𝙞𝙣𝙜. People read lists top to bottom. Splitting ranks into two columns forces zigzag eye movement and breaks order tracking. ❌ 𝘿𝙤 𝙣𝙤𝙩 𝙪𝙨𝙚 𝙗𝙖𝙧𝙨 𝙬𝙞𝙩𝙝𝙤𝙪𝙩 𝙖 𝙨𝙝𝙖𝙧𝙚𝙙 𝙖𝙣𝙘𝙝𝙤𝙧. Bars must start from the same baseline and match the values. Here, equal numbers produce different bar lengths. That makes the visual misleading. ❌ 𝘿𝙤 𝙣𝙤𝙩 𝙪𝙨𝙚 𝙘𝙤𝙡𝙤𝙧 𝙬𝙞𝙩𝙝𝙤𝙪𝙩 𝙢𝙚𝙖𝙣𝙞𝙣𝙜. Colors should encode a variable. These do not segment by sport, category, or anything else. That’s visual noise. ❌ 𝘿𝙤 𝙣𝙤𝙩 𝙖𝙙𝙙 𝙙𝙚𝙘𝙤𝙧𝙖𝙩𝙞𝙫𝙚 𝙤𝙧 𝙞𝙣𝙘𝙤𝙣𝙨𝙞𝙨𝙩𝙚𝙣𝙩 𝙞𝙘𝙤𝙣𝙨. Icons should clarify, not confuse. LeBron gets a crown instead of a basketball. Magic Johnson’s icon is unclear. The Rock is shown as a football player, even though football isn’t how he built his wealth. ❌ 𝘿𝙤 𝙣𝙤𝙩 𝙧𝙖𝙣𝙠 𝙬𝙞𝙩𝙝𝙤𝙪𝙩 𝙙𝙚𝙛𝙞𝙣𝙞𝙣𝙜 𝙩𝙝𝙚 𝙢𝙚𝙩𝙧𝙞𝙘. Is this career earnings or net worth? Business value or brand equity? The source is vague, so the comparison is undefined. 𝙏𝙝𝙚 𝙩𝙖𝙠𝙚𝙖𝙬𝙖𝙮: AI can generate visuals. It can’t replace understanding what you’re measuring, how people read, or why design choices matter.

📉 Common AI Errors in Data Visualization You see that your sales have tripled in a month! Excited, you look at the numbers in your beautifully AI-generated report—wow, the ROI is through the roof! Your bonus this month must be huge… Awesome, but wait... It looks like the AI simply multiplied the values instead of calculating the actual difference, creating an illusion of incredible success... AI makes insignts and data visualization faster, but it’s not perfect. It misinterprets data, distorts charts, and hides key insights, leading to misleading conclusions. Here are the most common issues to watch for: 📊 Data Processing Issues - Phantom Data – AI fills gaps with made-up values. - Wrong Grouping – Misclassifies categories. - Faulty Aggregation – Summarizes data incorrectly. 📉 Chart Distortions - Incorrect Scaling – Over- or under-emphasizes trends. - Swapped Axes – X and Y mix-up distorts meaning. - Poor Chart Choice – AI picks unclear visualizations. 🚫 Filtering & Visibility Issues - Hidden Data – AI removes "irrelevant" values. - Over-Simplification – Key insights get lost. 🎨 Over-Stylization - Excessive 3D & Color – Looks good, confuses users. - Distracting Animations – Harder to read, not easier. AI assists, but it doesn’t replace human judgment. Always review auto-generated insights before acting on them. What AI visualization issues have you encountered? #DataViz #AI #Analytics #MachineLearning #businessintelligenceAI

Anscombe's quartet is a group of four data sets that share identical statistical properties like mean, variance, correlation, and regression lines. However, when plotted, these data sets look dramatically different. This shows how important it is to visualize data instead of relying only on summary statistics. ✔️ Better Understanding: Visualizations help reveal patterns, outliers, and trends that might be hidden in the numbers. ✔️ Improved Decisions: Seeing the data helps understand relationships more clearly, leading to smarter decisions. ✔️ Model Validation: Plotting data can help assess if statistical models represent the data accurately. ✔️ Error Detection: Visualizations can quickly reveal data entry errors or unusual patterns that summary statistics might miss. ❌ Misleading Conclusions: Ignoring data visualization can cause wrong interpretations, even if the numbers look right. ❌ Limited Insight: Relying only on summary statistics risks missing crucial information. ❌ Bias Risk: Poorly designed visualizations can lead to biased interpretations. ❌ Overfitting Risk: Misinterpreting patterns in visualizations may lead to models that fit the training data too closely without generalizing well. The image below shows four scatter plots with identical statistical summaries but very different patterns. This makes it clear why data visualization is crucial for a complete understanding of data. Image adapted from Wikipedia: https://lnkd.in/eJPuBaCa 🔹 In R: Libraries like ggplot2 for plotting and dplyr for data manipulation are helpful. The datasauRus package has similar data sets for practice. Using broom can tidy model outputs for better analysis. 🔹 In Python: Use matplotlib and seaborn for plots and pandas for data handling. The statsmodels library is useful for visualizing how well models fit, while scikit-learn helps with building and evaluating models efficiently. Want to explore more about Statistics, Data Science, R, and Python? Subscribe to my email newsletter! For more information, visit this link: http://eepurl.com/gH6myT #ggplot2 #database #research #tidyverse #rprogramminglanguage #package #dataviz