> **About this file:** Text-only export for editors and AI assistants. **Figures are not embedded** (image URLs only work on the published site). Each figure is summarized below; open the live guide for the actual charts. > > **Web edition:** `/resources/data-visualization-best-practices/` (prepend your site origin). --- ## 1. Start with the message, not the chart Before choosing a chart type, define what you want to communicate. Ask: ``` What is the main idea? Who is the audience? What decision or discussion should this chart support? Where will the chart be used: dashboard, document, or live presentation? How much time will the audience have to interpret it? ``` Avoid starting with: ``` Which chart looks nice? Which chart does the tool suggest? Which chart did I use last time? ``` A good chart often communicates **one main point**. If the audience needs to discover the message by inspecting every label, axis, and legend, the chart is probably doing too much. --- ## 2. Use visual thinking People do not read charts like they read text. In text, readers usually follow a sequence. In charts, the eyes move toward what stands out: ``` Strong colors Large objects Bold text Outliers Peaks and valleys Intersections Labels Highlighted areas Unexpected patterns ``` This means you must actively guide attention. If the main insight is a line crossing another line, make that crossing obvious. If one category matters most, highlight it. If the conclusion is a gap, emphasize the gap. The audience should not have to search for the point. **Figure:** Figure 65: Chart with a single important point — Attention goes to the strongest visual cue first—for example the intersection of two lines—before the audience reads labels and axes sequentially. *(Graphic: see the published guide.)* --- ## 3. Respect visual perception limits People can only process a limited number of visual elements at once. As a rule of thumb: ``` Few elements → individual comparison is possible. Many elements → the audience sees an overall pattern instead. ``` Use this intentionally. If the goal is to compare individual categories, limit the number of elements. If the goal is to show a general trend or distribution, many elements can work, as long as the overall pattern is clear. Example: ``` Show 5 product categories if the audience must compare them. Show 100 customer trajectories if the goal is to reveal a broad behavioral pattern. ``` **Figure:** Figure 66: Global trend versus individual trends in a line chart — Many lines can stop being read one-by-one and instead read as a single pattern; highlighting selected series while muting the rest keeps both views useful. *(Graphic: see the published guide.)* --- ## 4. Avoid misleading patterns The audience naturally looks for meaning, patterns, trends, and relationships. This is powerful, but dangerous. Be careful with: ``` Color choices Category order Axis scales Time windows Missing context Truncated axes Dual axes Inconsistent intervals ``` The same data can tell very different stories depending on the time period, scale, grouping, or reference point. For example, a trend may look stable over ten years but alarming over three months. Both may be technically true, but the chart must make the chosen context clear. **Figure:** Figure 67: Different meaning conveyed by choosing a different time frame — The same series can imply a different story depending on the time window—always make the framing explicit. *(Graphic: see the published guide.)* --- ## 5. Respect conventions Charts are easier to understand when they follow familiar conventions. Common conventions: ``` Time goes on the horizontal axis. Time moves from left to right. Higher values go upward. Red often means bad or warning. Green often means good or positive. Blue often suggests cold or neutral. Similar colors imply similar categories. Darker or more saturated colors imply higher importance or value. Categories should be ordered logically or by value. ``` Breaking conventions increases cognitive effort and can mislead the audience. Only break conventions when there is a strong reason, and make the design very clear. **Figure:** Figure 90: Order and color modified to respect conventions — Logical category order and meaningful color speed interpretation and reduce accidental misreading. *(Graphic: see the published guide.)* --- ## 6. Choose the best available chart There is rarely a perfect chart. The goal is to choose the best available option for the message, audience, and medium. A practical process: ``` 1. Write the message in one sentence. 2. Identify the key action words: compare, show trend, rank, relate, distribute, compose. 3. Sketch possible charts before building them. 4. Prototype with real or realistic data. 5. Remove, simplify, and refine. ``` Example: ``` “I want to compare the number of job postings to hires across job types.” ``` Key ideas: ``` Compare Ratio Different categories ``` This suggests a chart that compares ratios across categories, such as a bar chart or dot plot. **Figure:** Figure 71: How to modify your chart to reach the best available solution — Iterate across alternatives—density, simplified bars, color, line comparisons—because the “best” chart depends on context, not a single default. *(Graphic: see the published guide.)* --- ## 7. Prefer simple charts Simple charts are usually more effective than sophisticated ones. Most business communication can be handled with: ``` Simple text Tables Heatmaps Bar charts Line charts Scatterplots Slopegraphs Waterfall charts 100% stacked bars ``` Use complex charts only when a simpler chart cannot communicate the message. Complexity is not bad by itself, but the audience must be able to understand the chart quickly enough for the situation. A chart for a live executive presentation should usually be simpler than a chart in a detailed analytical appendix. **Figure:** Figure 88: Simplifying by reducing colors, elements, and decimals — Fewer colors, fewer ornaments, and less decimal noise usually improve clarity without changing the underlying message. *(Graphic: see the published guide.)* --- ## 8. Structure the chart clearly A good chart has a clear visual hierarchy. Useful starting structure: ``` Title: around 10–15% of the space Subtitle: around 5–10% Chart field: around 70–80% Source / footnote: around 5% ``` These proportions are not strict rules, but they help create balance. The title should communicate the message, not merely describe the chart. Weak title: ``` Sales by Month ``` Better title: ``` Sales declined sharply after March ``` The title should answer: **What should the audience notice?** --- ## 9. Use alignment and whitespace Good alignment makes a chart feel professional and easier to read. Best practices: ``` Align titles, subtitles, axes, and source notes. Avoid diagonal text. Use left alignment for most text. Use consistent spacing between elements. Leave enough margin around the chart. Do not fill every available space. ``` Whitespace is not wasted space. It helps the audience focus. A single important number on a mostly blank slide can be more powerful than a crowded dashboard page. **Figure:** Figure 73: How to visually improve a chart — Stronger alignment, clearer structure, and deliberate whitespace make the same data faster to read. *(Graphic: see the published guide.)* --- ## 10. Lead the audience to the main idea Use visual emphasis deliberately. The main tools are: ``` Contrast Color Saturation Thickness Position Labels Annotations Reference lines Frames Arrows Ordering ``` Good highlighting usually follows this pattern: ``` Important element → stronger color, label, or thickness. Secondary elements → light gray, thinner lines, fewer labels. ``` Avoid using many colors at once. A blue item stands out among gray elements. It does not stand out among ten other bright colors. A strong default palette is: ``` Gray for context One saturated color for emphasis Optional second color for contrast ``` **Figure:** Figure 74: Highlighting the main line — Emphasize the series or segment that carries the insight—for example the line where debt splits fastest—so viewers see the conclusion before the footnotes. *(Graphic: see the published guide.)* --- ## 11. Use color with discipline Color should communicate meaning, not decoration. Good uses of color: ``` Highlight the main category. Separate positive and negative. Show intensity. Group related categories. Distinguish actuals from forecast. Represent status or risk. ``` Poor uses of color: ``` One random color per bar. Many saturated colors with no meaning. Changing category colors across pages. Using red and green without considering accessibility. Using color when direct labels would work better. ``` Best practices: ``` Use ideally two main hues. Use saturation to show emphasis. Keep secondary elements in light gray. Keep category colors consistent across the full report or presentation. Avoid bright colors unless you need strong warning or emphasis. ``` **Figure:** Figure 77: Use of saturation and text to highlight information — Use saturation tiers plus concise text so “normal,” “important,” and “most important” read instantly. *(Graphic: see the published guide.)* **Figure:** Figure 89: Creating categories to simplify the chart — Grouping categories cuts competing hues and makes comparisons easier without hiding the story. *(Graphic: see the published guide.)* --- ## 12. Remove unnecessary elements Every chart element should justify its existence. For each element, ask: ``` Does this help communicate the message? Would removing it make the chart less clear? Is there a simpler way to achieve the same purpose? ``` Usually remove or reduce: ``` Heavy borders Dark gridlines Unnecessary data markers Redundant legends 3D effects Excessive decimals Unnecessary axis titles Repeated labels Decorative backgrounds Too many tick marks ``` Use lighter versions when the element is useful but not central. For example: ``` Use light gray gridlines instead of dark gridlines. Use direct labels instead of a legend. Use only the key data labels instead of labeling every point. ``` **Figure:** Figure 87: Better chart by removing unnecessary elements — Strip grid clutter, prefer direct labels over noisy legends, and keep only marks that earn their ink. *(Graphic: see the published guide.)* --- ## 13. Reduce eye travel Good charts keep related information close together. Instead of placing a legend far away from the data, label lines directly. Poor pattern: ``` Line chart on the left Legend below the chart Audience must look back and forth repeatedly ``` Better pattern: ``` Line labels placed near the end of each line ``` This reduces cognitive effort and makes the chart faster to read. --- ## 14. Choose the right level of detail for the medium The same chart should not necessarily look the same in a dashboard, document, and presentation. ### Live presentation Use fewer elements. ``` Highlight only the key message. Use fewer labels. Explain verbally when needed. Use animation to reveal complexity step by step. Pause before speaking so the audience can read the chart. ``` **Figure:** Figure 85: Chart suitable for animations in a presentation — Slides can reveal steps sequentially—animations help steer attention when detail must unfold live. *(Graphic: see the published guide.)* ### Document or PDF Include more detail. ``` Add source notes. Add definitions. Add more labels if useful. Include footnotes. Allow the reader to inspect the chart independently. ``` ### Dashboard Support exploration. ``` Use consistent layout. Include filters carefully. Provide tooltips. Avoid too much visual density. Make definitions available. Keep interactions intuitive. ``` --- ## 15. Be careful with axes and scales Axes strongly influence interpretation. General rules: ``` Bar charts should start at zero. Line charts should usually start at zero, but exceptions are acceptable. Use constant intervals. Make truncated axes obvious. Clearly label logarithmic scales. Avoid dual axes when possible. ``` Bar charts rely on length comparison, so a non-zero baseline exaggerates differences. For line charts, a truncated axis can be acceptable when small changes matter and values are far from zero. But the truncation must be clear. Better alternatives to truncating the axis: ``` Show absolute change. Show percentage change. Show an index with base 100. Add a small reference number. ``` Avoid dual axes because they can create false relationships between unrelated variables. Prefer two aligned charts with the same x-axis. **Figure:** Figure 82: How to deal with scale problems — Truncated baselines, logs, absolute versus percent change, and split panels each solve scale tension differently—pick honestly. *(Graphic: see the published guide.)* --- ## 16. Use reference points to improve understanding Sometimes raw numbers are hard to interpret. Translate them into a familiar reference. Instead of: ``` Employees waste 18,000 hours per year on manual reporting. ``` Use: ``` That equals around 10 full-time work years lost annually. ``` Useful reference points: ``` Per employee Per customer Per day Per month % of total Index versus baseline Days of work Cost per transaction Equivalent headcount ``` The right reference point makes the message easier to grasp. --- ## 17. Use simple text when the number is the message If you only need to communicate one or two numbers, do not force a chart. Use large text. Example: ``` 20% of customers completed onboarding ``` Add a short explanation if needed. Simple text is often better than a pie chart, gauge, or single-bar chart when the message is just one figure. --- ## 18. Use tables when users need to look up values Tables are useful when the audience needs to read individual values. Use tables for: ``` Detailed comparison Operational reporting Reference values Financial statements Ranked lists with several metrics ``` Improve tables by: ``` Reducing borders Using whitespace Aligning numbers to the right Aligning decimals Using clear row and column headers Highlighting only the important values ``` If the table is large and pattern recognition matters, use a heatmap. Heatmaps help users quickly identify high and low values without reading every number. Always include a clear high-low legend when using heatmaps. --- ## 19. Use scatterplots for relationships Scatterplots are useful when you want to show the relationship between two variables. Use scatterplots for: ``` Correlation Clusters Outliers Distribution Relationship between price and demand Relationship between effort and impact ``` You can add: ``` Reference lines Average lines Trend lines Color highlights Labels for selected points ``` Be careful with bubble charts. Size and color can add information, but too many encodings make the chart difficult to understand. If the audience is not familiar with scatterplots, keep the design simple and use annotations. **Figure:** Figure 94: Example of scatterplot — Show the relationship between two measures and lean on guides or highlights so outliers and clusters read quickly. *(Graphic: see the published guide.)* --- ## 20. Use line charts for trends over continuous data Line charts are best for continuous data, especially time. Use line charts for: ``` Monthly sales Daily users Weekly tickets Annual revenue Forecast versus actuals ``` Avoid line charts for nominal categories because the line implies continuity or sequence. Good practices: ``` Keep time intervals consistent. Use no more than around four main series. Directly label lines where possible. Use light gray for context lines. Highlight the most important line. Use shaded bands for confidence intervals when relevant. Differentiate forecasts with dotted lines or visual separation. ``` If the focus is on individual period values rather than trend, use bars instead. **Figure:** Figure 98: Line graph with additional elements — Line labels, deliberate color and weight, and careful axes keep multi-series trends legible without extra chrome. *(Graphic: see the published guide.)* --- ## 21. Use slopegraphs for before/after comparisons Slopegraphs are useful when comparing two points in time or two conditions. Use them for: ``` This year vs last year Before vs after Current vs target Pre-campaign vs post-campaign ``` They show: ``` Direction of change Magnitude of change Relative ranking changes ``` Avoid slopegraphs when too many lines overlap. Highlight the most important lines and mute the rest. **Figure:** Figure 100: Example of slopegraph — Compare two moments in time and spotlight the category whose slope carries the narrative. *(Graphic: see the published guide.)* --- ## 22. Use bar charts for category comparison Bar charts are one of the clearest chart types because people compare length well. Use bar charts for: ``` Sales by product Customers by segment Tickets by category Reasons for churn Survey responses Ranking priorities ``` Best practices: ``` Start the axis at zero. Sort categories by value or logical order. Use horizontal bars for long labels. Avoid too many categories. Use direct labels when helpful. Keep spacing between bars consistent. ``` Horizontal bars are often better when category names are long or there are many categories. --- ## 23. Use stacked bars carefully Stacked bars are useful for showing total plus composition. Use stacked bars when: ``` The total matters. The composition matters. The number of subcategories is limited. ``` Be careful: only the bottom segment is easy to compare across bars. Other segments are harder because they do not share the same baseline. Use 100% stacked bars when the composition matters more than the absolute total. When using 100% stacked bars, consider adding absolute values separately so the audience understands the size behind the percentages. For surveys, horizontal 100% stacked bars often work well, especially when responses follow an ordered scale. **Figure:** Figure 105: 100% stacked column chart — 100% stacks compare composition across periods—pair them with labels or callouts so viewers track the segments that matter. *(Graphic: see the published guide.)* --- ## 24. Use waterfall charts for bridges Waterfall charts are useful when explaining how a starting value becomes an ending value through positive and negative changes. Use waterfall charts for: ``` Budget to actual variance Revenue bridge Profit bridge Headcount movement Cost increase explanation Sales objective deviation ``` They work well when each intermediate driver has a clear additive impact. **Figure:** Figure 109: Example of waterfall chart — Walk readers from a starting balance through gains and losses to the ending value so each bridge bar tells part of the story. *(Graphic: see the published guide.)* --- ## 25. Avoid pie charts in most cases Pie charts are usually harder to read than bar charts because people are not good at comparing angles and areas. Avoid pie charts when: ``` There are many categories. Differences are small. You need precise comparison. You want to compare multiple pies. ``` A pie chart is only acceptable when: ``` There are very few categories. The message is about a simple part-of-whole relationship. The difference is obvious, such as half or quarter. ``` If you need to compare composition across groups, use a 100% stacked bar chart instead. --- ## 26. Use unit charts when individuality matters Unit charts are useful when each unit represents a person, customer, employee, product, or case. They work well when the message is about: ``` Probability Rarity Population share Individual impact Human outcomes ``` Example: ``` Only a small number of high school basketball players reach the NBA. ``` A unit chart can make that idea more concrete than a pie chart because each dot feels like an individual. **Figure:** Figure 81: Unit chart as a better alternative to pie chart — Unit marks preserve individuality—often more persuasive than abstract wedge angles for rare outcomes. *(Graphic: see the published guide.)* --- ## 27. Use advanced charts only when justified Advanced charts can be useful, but they require more audience effort. Examples: ``` Alluvial diagrams Network graphs Boxplots Marimekko charts Maps Sankey diagrams Unit charts ``` Use them when: ``` The audience can understand them. The data structure requires them. A simple chart would hide the important pattern. You have enough space to explain the chart. ``` Do not use advanced charts just to look sophisticated. --- ## 28. Practical chart selection guide | Goal | Recommended chart | | --- | --- | | Show one number | Simple text | | Look up detailed values | Table | | Compare categories | Bar chart | | Compare long category labels | Horizontal bar chart | | Show trend over time | Line chart | | Compare before vs after | Slopegraph | | Show relationship between two variables | Scatterplot | | Show distribution / outliers | Scatterplot, boxplot | | Show part of whole | Bar chart, 100% stacked bar | | Show change from start to end through drivers | Waterfall | | Show many individual trajectories | Line chart with muted individual lines | | Show geographic patterns | Map, only if location matters | | Show networks or relationships | Network graph | | Show flow between categories | Sankey / alluvial, only if needed | --- ## 29. Final checklist Before publishing a chart, ask: ``` Is the main message clear? Is the chart type appropriate? Can the audience understand it quickly? Have I removed unnecessary elements? Is the title meaningful? Are colors used consistently? Is the key point highlighted? Are axes and scales honest? Are labels readable? Is the order meaningful? Is there enough whitespace? Is the chart adapted to the medium? Would a simpler chart work better? ``` --- ## 30. Core rules ``` Start with the message. Choose the simplest chart that communicates the idea. Guide attention deliberately. Use color sparingly and consistently. Respect chart conventions. Reduce clutter. Label directly where possible. Avoid misleading scales. Avoid dual axes. Avoid pie charts unless the case is very simple. Use more detail in documents and less detail in presentations. Design for the audience, not for the tool. ```