Show Me The Numbers Designing Tables And Graphs To Enlighten

The Practical Guide to Visualizing Data Without Losing Your Mind

Data visualization is one of those skills everyone thinks they have until they actually have to do it at 11 PM before a stakeholder meeting. The core problem is rarely about tools. It is about understanding what your data is trying to tell you and removing every unnecessary element between that message and the person reading it. Most dashboards fail because they are cluttered, not because they lack information. Show Me The Numbers Designing Tables And Graphs To Enlighten is a practical book by Donald A. Moore that focuses on exactly this: how to construct tables and graphs that people can actually read and act on. It is not a software tutorial. It does not walk you through Excel or Tableau menus. It is about decision-making at the desk before you open any program. I have spent years building reporting systems for mid-market companies, and the most common mistake I see is people defaulting to whatever chart type their tool suggests first. A bar chart gets auto-generated. Then someone adds three series, colors everything differently, and throws in a legend that takes up half the page. The result is a visual soup that communicates nothing. This book pushes back against that instinct.

Show Me The Numbers Designing Tables And Graphs To Enlighten

The central thesis is straightforward: clarity beats decoration every time. Moore spends considerable time on table design, which most visualization guides skim over, because tables are where most real decisions happen. People read tables more carefully than charts. A well-constructed table can show more data in less space than any chart and still be legible. His guidance on tables includes specific rules that sound obvious until you see the alternatives. Number columns should align on the decimal point, not the left edge. This is a small detail that most people miss. When decimal points are misaligned, readers have to visually track each digit across rows to compare values. That adds cognitive load. It also makes it harder to spot outliers or patterns. Aligning on the decimal fixes that in one step. He also recommends using separators liberally. Thousands separators in number columns, grid lines between rows, and borders around key totals. These are not decoration. They are visual anchors that help the eye move through the data without getting lost. A table without any separation between rows looks like a spreadsheet dump. It forces the reader to do the work of parsing structure that should be automatic.

Another point that comes up repeatedly is the importance of labeling directly. Legends are inefficient. When you have five colored bars and a legend telling you what each color means, the reader has to look at a bar, look away to the legend, look back. That back-and-forth accumulates. Direct labels next to or inside data points eliminate that friction entirely. The book makes this case with specific before-and-after examples that are genuinely useful. On the chart side, Moore breaks down the common types and explains when each one works and when it does not. Bar charts are for comparing categories. Line graphs are for showing trends over time. Scatter plots are for showing relationships between two variables. Pie charts are for showing parts of a whole, but only when there are few categories. He is blunt about pie charts: more than five slices and they become nearly impossible to compare accurately. Most business presentations use pie charts with eight or ten slices. This is why they look bad. The scatter plot section is where he gets into territory that beginners usually miss. He covers how to interpret correlation visually, when to add trend lines, and how to handle overplotting when you have hundreds of data points. Overplotting is a real problem in business data. You might have daily revenue broken down by region for three years. That is over a thousand points. A standard scatter plot becomes a solid dark blob. The workaround he suggests is using semi-transparent markers or aggregating into binned regions. I ran into this exact problem last year with a customer churn dataset. The default scatter plot was unreadable. Switching to binned heat coloring made the pattern visible in under five minutes.

One counter-intuitive point from the book is about axis scales. Most people assume linear scales are the default and therefore the correct choice. Moore argues that the correct choice depends entirely on what question you are trying to answer. If you are showing percentage growth where early values are small and later values are large, a linear scale will compress the early period into an unreadable sliver. A logarithmic scale spreads those early values out. This is not about making the chart look dramatic. It is about showing the data honestly relative to the question being asked. He also addresses a problem I encounter constantly: people use charts to show data that would be clearer in a table. If you have ten categories and three metrics per category, a small multiples approach or a well-formatted table beats a grouped bar chart every time. Grouped bars with three series across ten categories create thirty bars. The reader has to mentally separate them by color and position. A table with the same data takes up less visual space and allows faster comparison because the eye does not have to decode a legend. The section on color is practical rather than theoretical. He covers accessibility, specifically color blindness, and gives concrete palette recommendations. Red-green combinations are problematic for the most common form of color vision deficiency. His suggestion is to use color as a secondary cue alongside shape or label, not as the primary differentiator. This is especially relevant for printed materials where color reproduction varies.

He also warns about the temptation to add three-dimensional effects to charts. 3D charts distort proportions. A 3D pie chart makes the front slices appear larger than the back slices even when the underlying values are identical. The book shows side-by-side comparisons that make this immediately obvious. I have fixed at least a dozen presentations where someone applied a 3D effect "for style." Every one of them introduced a measurable distortion that changed how the data was interpreted.

When This Approach Falls Short

The book is not a universal solution. Its guidance assumes you are working with structured, quantitative data. If your data is qualitative or exploratory, the prescriptive approach to chart selection can feel limiting. The book also skims over interactive dashboard design, which is where most business data lives today. Static chart principles transfer to interactive contexts, but the interaction layer introduces its own problems: drill-down depth, filter cascades, tooltip density, and responsive layout. These are largely unaddressed. Another limitation is that the examples lean heavily toward business and economic data. If you are working with scientific or geospatial data, the principles still apply but the specific chart recommendations may need adaptation. The core idea of reducing cognitive load remains constant across domains, but the implementation details differ. The book also does not cover statistical rigor in depth. It tells you how to make a chart readable. It does not deeply address whether the chart accurately represents statistical significance or uncertainty. A chart can be beautifully designed and still mislead if confidence intervals are omitted or sample size is not considered. This is a gap that matters in regulated industries.

Practical Rules I Actually Use

Here is what I take from this and apply without modification. Every chart I build goes through a simple checklist before it gets shared externally. The title states the finding, not the data type. A title like "Revenue by Quarter" is not useful. "Revenue Declined 12% in Q3 After Two Years of Growth" tells the reader what to look for. I remove grid lines that do not serve a purpose. Horizontal grid lines are sometimes useful for reading values off a bar chart. Vertical grid lines in line charts are almost never needed. I align all number columns on the decimal point and use thousands separators. I label data points directly wherever the chart density allows it. I avoid color palettes with more than six distinct hues. I test every chart in grayscale to catch accessibility issues early. The table formatting rules are the ones I enforce most strictly. Every column gets a header that describes the metric and the unit. Decimal alignment is non-negotiable. I use bold for subtotals and italics for notes. I include a source line under every table that states the date range and data origin. This takes thirty seconds and prevents at least half the follow-up questions I used to get.

What to Actually Do First

If you are starting from scratch, do not open a visualization tool. Write down the question you are trying to answer. Then write down what type of data would answer that question. Then pick the simplest visual format that can show that data. If a table works, use a table. Only escalate to a chart when the question involves spatial relationships, trends over time, or comparisons that benefit from visual encoding. This sounds obvious. It is the step most people skip because they already know how to make a chart and rarely stop to ask whether they should. The book includes a chapter on common mistakes that functions as a checklist of things to avoid. Misleading axis breaks. Stacked bar charts that make individual series unreadable. Charts with more data series than the legend can reasonably support. Overlapping labels. Dual axes that serve no analytical purpose. These are the errors I see most often, and they are all preventable with a bit of upfront planning. There is also coverage of anomaly detection through visualization. Outliers do not always need to be highlighted with special formatting. Sometimes they are simply the tail end of the distribution and the chart should show that naturally. The decision to flag or integrate an outlier depends on whether it is meaningful to the question. This is a distinction that matters. I once built a dashboard where the anomaly highlight feature was turned on by default. It flagged seventeen percent of all data points as anomalies because the threshold was set too aggressively. The constant highlighting trained the audience to ignore it entirely. Fixing it required removing the automatic flagging and adding manual context notes instead.

The book also touches on how to handle missing data visually. This is an area where most guidance is vague. The practical approach is to indicate gaps clearly rather than connecting points across missing periods with a line. A broken line or a dashed segment communicates uncertainty better than a continuous line that implies data exists where it does not. This is a small detail that changes how responsibly the data is interpreted. One section I found useful was on designing for print versus screen. Print requires higher contrast, larger fonts, and consideration for ink density. Screen allows interactivity but introduces glare and scrolling issues. The same chart often needs two different treatments depending on the medium. I learned this the hard way when a stakeholder printed a dashboard and could not read the light gray trend line against the white background. Changing the line weight and switching to a darker blue solved it immediately.

The Core Takeaway

The main point of this book is that good visualization is a discipline, not a talent. It requires decisions at every step: what to include, what to omit, how to align, how to label, how to color. Those decisions are guided by principles, not preferences. The principles are simple. They are also easy to ignore under time pressure. Reading through the examples and the checklists helps make them habitual. If you build tables and charts regularly and want a reference that focuses on the actual mechanics of clarity rather than theory, this is a solid one. It is not exhaustive on every modern visualization technique. It does not cover GIS, network graphs, or animated transitions. But for the core task of making numbers understandable, the guidance is direct and practical. The worst thing you can do with data is make it harder to read than it already is. This book gives you the tools to avoid that mistake.