Confused about what to put on a dashboard, which chart to pick, or why nobody acts on your reports? This page walks you from the basics to hands-on practice: six lessons, nine bad dashboards to fix, a what-if lab and seven exercises on choosing the right data.
Your tools do not know your story. The analyst does. Follow the six lessons in order, because each one builds on the last.
Who decides? What should they do? Which data proves it?
A dozen visuals cover nearly everything. Retire pies, 3D and dual axes.
Every element costs attention. Delete what carries no information.
Grey for context, one accent colour for the point.
Make it obvious how to read it. Use words, align, leave white space.
Setup, conflict, resolution, then a clear call to action.
Exploratory work is how you find out what is interesting. Explanatory work is how you help someone else understand one thing you already know. The book's image: open 100 oysters to find two pearls, then show your audience the pearls, not the oysters.
| Exploratory | Explanatory | |
|---|---|---|
| Goal | Find out what might be interesting | Explain one specific point |
| Audience | You and fellow analysts | A named decision maker |
| Output | Many quick, rough views | One or two polished views with a message |
| Typical mistake | Sending the rough views to the boss | Hiding the conclusion and making them re-discover it |
| In logistics | A pivot table with 40 filters | A weekly page that opens with "what changed, what to do" |
Be specific about the audience. "Internal and external stakeholders" is not an audience; "the Operations Director who signs the lease" is. Name the decision maker, and think about whether they already trust you as an expert.
State what they should know or do: approve, fund, stop, escalate, change. If you cannot name an action, ask whether you need to communicate at all. If you present only data, people say "interesting" and move on.
Only now ask which data supports the point. The data is evidence for the story, not the story. Do not hide evidence that cuts the other way; a sharp reader will find it and doubt everything else.
You control pace and attention, and you are there to answer questions. Keep slides sparse and do not use them as a teleprompter.
One file serving both a talk and a read-alone document. It satisfies neither fully; later lessons show ways around it.
The reader controls pace and nobody is there to explain, so titles, labels and notes must do the talking.
Also decide the tone: celebrating a win, lighting a fire, or reporting calmly. Tone drives later choices such as colour and layout.
A self-serve dashboard sits at the "written" end. Nobody is in the room, so each page needs a headline that says what to conclude, not just a topic label.
The 3-minute story is what you would say if you only had three minutes. The Big Idea shrinks it to one sentence. Being short is harder than being long, and it frees you from depending on slides.
A Big Idea must (1) state your unique point of view, (2) say what is at stake, and (3) be a complete sentence.
"Storage demand will outgrow our current warehouse by year 3. We recommend securing additional space now, because waiting puts projects at risk; please approve the lease study."
Point of view: lease now. Stake: projects at risk. Complete sentence with an ask.
Do not start in presentation software; it pulls you into slide-making mode and makes you attached to what you built. Use sticky notes or paper so ideas are cheap to move or bin. Here is a storyboard for the fictional warehouse case.
Busy audience that already trusts you? Move note 5 to the front and lead with the ask.
Fill in what you know. The canvas drafts a Big Idea and a 3-minute story skeleton and tells you what is still missing. Nothing leaves your browser.
Tables interact with our verbal system: we read them. Graphs interact with our visual system, which is faster. That one idea explains most of the advice below.
Choose what you want to say. The picker recommends a visual and tells you what to avoid.
Eyes cannot judge angles and areas well. When slices are close, you cannot rank them at all.
Swap: sorted horizontal bars. Try it in Lab B.
It asks readers to compare arc lengths, which is harder still than a pie.
Swap: a bar, or just the number.
Use 3D only if you truly plot a third dimension. Otherwise it skews values; in the book a bar of 1 reads as about 0.8.
Swap: flat 2D.
Readers must work out which axis each series uses, and the pairing implies a relationship that may not exist.
Swap: direct labels or two panels sharing the x-axis. Try Lab C.
Pause and ask why. If you can answer, label the values and keep it flat. It should never be the first chart you reach for.
Great for mixed audiences who each look up their own row, or when units differ. Let the data lead: light or no borders, right-aligned numbers, consistent decimals. In live talks, a table pulls attention away from you.
If the big-picture trend matters, keep a light grey axis. If exact values matter, label the data points and drop the axis, because showing both is redundant.
Manipulating a scale because it supports your point is not acceptable, and it is risky: one discerning reader spots it and your credibility goes with it.
The mental effort needed to take in new information. Extraneous load is effort that does not help understanding.
What matters most is how hard your audience believes it will be. Decided in a glance, often unconsciously. If it looks hard, many will not try.
Signal is the message; noise is anything that does not add to it, or detracts. Maximize the share of ink spent on data.
These six principles tell you what viewers will group together, and therefore which extra elements you can safely remove.
Things close together belong together. Use: group related KPIs; put labels next to the data they describe.
Same colour, shape or size means related. Use: colour a label like its line; no borders needed to show rows.
Things inside a boundary form a group, even a faint one. Use: light shading to separate forecast from actuals.
We fill gaps to see a whole shape. Use: delete chart borders and background fills; the chart still looks complete.
Eyes follow the smoothest path. Use: drop the vertical axis line; aligned labels and bars imply it.
Connected things are grouped, even across colours. Use: lines in a line chart link points into a series.
Create clean lines. Left-align text; centre-aligned blocks leave ragged edges and look sloppy. Put the title, axis title and legend top-left so readers learn how to read before they see data. Avoid diagonal text: one study the book cites found 45° text about 52% slower to read and 90° text about 205% slower.
People fear empty space, but it works like a pause in a talk. Keep margins clear, do not stretch charts to fill the page, and never add data just because there is room. If one thing matters most, make it the only thing on the page.
It is easy to spot a hawk in a sky of pigeons, and hard when there are many kinds of birds. The more things you make different, the less any one stands out. Use contrast only for the point.
Closure: the chart still reads as a whole.
Or make them thin and light grey so they never compete.
The line already shows the data; use markers on purpose, not by default.
Trim trailing zeros, abbreviate months, keep text horizontal.
Proximity: put the name next to the line and drop the legend.
Similarity: colour the label like its data.
Keep $, % and thousands separators on the numbers themselves. A chart titled "$ millions" with bare axis numbers forces the reader to remember the unit.
Lasts a fraction of a second and is tuned to differences. This is where preattentive attributes act.
Holds about four chunks at once. Ten coloured series plus a legend overload it. Label directly and group information into larger chunks.
Pictures help recall words, and repetition helps things stick. Pair visuals with words.
These are features the brain notices without effort. Some can encode quantity; others only separate categories.
| Attribute | Quantity or category? | Dashboard use |
|---|---|---|
| Position, line length | Quantity can encode value | Why bars and lines are easy to read. |
| Line width, size, colour intensity | Quantity to a limited extent | Size signals importance; saturation makes a heatmap. |
| Colour hue, shape, orientation, curvature, enclosure | Category | Separate series or highlight one; "which is greater, red or blue?" is meaningless. |
| Added marks (labels, markers, underline) | Category | "Look here" signals, for example a marker and label on the last point. |
| Bold, italics, underline (text) | Category | Bold is the cleanest; italics are mild; underline adds noise. |
| Motion, blinking | Avoid | Flashing is more annoying than helpful. In a live talk, a simple appear/disappear build can guide attention; avoid fly-ins and fades. |
Finish the visual, look away, look back, and note where your eyes land first. Ask a colleague to talk you through it. If they land elsewhere, change the emphasis.
Highlighting one point makes the others harder to see, so avoid it during exploration. Also keep highlighted elements to a small fraction of the design; the book cites a guideline of at most about 10%.
A rainbow means nothing stands out. Keep one scheme across the whole communication so readers learn that, say, copper means "look here". Change colour only to signal a real change. If regions have colours, keep them the same everywhere.
Roughly 8% of men and 0.5% of women have trouble with red and green. Avoid putting them side by side, or add a second cue such as bold, saturation, or a + / − sign. Blue and orange are a safe positive/negative pair.
Colour sets the mood. A clinical report and a light feature need different palettes. Check cultural meanings of colours for international audiences.
Pick one or two brand colours as your "look here" cues and keep the rest grey. If the brand colour is too pale to stand out, use bold black or another colour that does not clash.
Give equal-importance items equal size. The book recounts a dashboard where the data that happened to be available took about 60% of the space and nearly steered the decision. Make design a decision, not an accident.
Without cues, readers start top-left and zigzag in a "Z". Put the most important thing top-left, and do not fight habits such as negative on the left, positive on the right.
Feel preattentive attributes yourself in the Count-the-3s lab →
A door handle shows you to pull. A good visual shows you where to look. Three moves create that.
Layer size, colour and bold on the one thing that matters. Keep the highlighted share small.
Ask of every item: would removing this change anything? If not, remove it. Summarize detail the audience does not need; move the rest to an appendix.
Order things by importance with size, colour and position. Super-categories (for example Age / Region / Income) give readers a frame.
Research the book cites found that instructions in a fussy font were judged harder, and people were less likely to try them. The more complicated it looks, the less likely people are to spend time on it.
Every chart needs a title and every axis needs a title. State the conclusion in words, because two people can read the same chart and conclude differently. Annotate to explain a spike, a policy change or an outage.
Rare exception: when you want readers to focus on peaks and valleys rather than the exact metric, you may drop the vertical axis labels deliberately.
The title is prime real estate. Use it for the conclusion or the ask.
| Descriptive (topic label) | Action title (conclusion) |
|---|---|
| Inbound OTIF by week | Inbound OTIF fell below the 90% target in 3 of the last 4 weeks |
| Warehouse utilization | Zones A and C exceed the 85% planning ceiling; B and D have room |
| Vessel waiting time, Q3 | Vessel waiting time doubled in Q3, mostly because the berth was occupied |
| Spend by vendor | Three vendors take 60% of spend: consolidate before contracts renew |
| Incident trend | Incidents in Area 3 rose for four straight months; other areas are flat |
| Receiving team capacity | Receiving has trailed inbound volume since June; two backfills would close the gap |
These checks are my own rules of thumb for practice. They cannot judge whether your title is true or useful.
People rate attractive designs as easier to use, accept them more readily and tolerate their flaws. Three habits create it without being a designer: use colour deliberately, align elements on clean lines, and use white space on purpose. When you meet a chart you like, ask why, and keep a library to borrow from.
Familiarity breeds comfort, so even better designs meet resistance. Four tactics help.
Say why it will look different and what readers gain.
A before and after makes the improvement visible.
Create two or three versions and ask for input.
Win over an influential reader one to one.
If resistance persists, test with someone neutral: the problem may be your design, not their habits.
Set the stage and answer "why should I care?" Five questions help:
Build the case for "what could be":
Say exactly what you want them to do. A classic close ties back to the opening problem and restates the urgency.
A deck of bulleted facts persuades only intellectually, and the audience argues with you in their heads. A story joins an idea to an emotion and gives people a reason to act. If you think "I have no problem to frame", look again: conflict is what holds attention, and a story where all is well rarely moves anyone.
| Order | Works best when… | How |
|---|---|---|
| Chronological | You must build credibility, or the audience cares about your method | Problem → data → analysis → finding → recommendation |
| Lead with the ending | The audience already trusts you, is busy, or wants the "so what" | Start with the ask, then back up with the strongest evidence |
Tell a live audience which structure you are using ("I will start with what I need from you"). In a written report, the words on each page must make the point on their own.
Read only the slide or section titles. Together they should tell the whole story. This needs action titles. An executive summary up front, mirrored by the titles that follow, makes the path visible.
Everything on one page reinforces everything else: title supports visual, words support chart, nothing unrelated. Cutting or moving to an appendix is as important as adding.
Flip through the finished deck and write the main point of each page. The list should look like your original storyboard. If not, move, add or remove pages.
Give it to someone with no context and ask what they notice, what matters and what is unclear. You are too close to see it as your audience does.
Tell them what you are about to tell them: an executive summary.
Tell them: the detailed content.
Tell them what you told them: recap, with the action you need.
Short, repeatable sound bites help your message travel without you.
These six headlines belong to one deck, but they are shuffled. Click them in the order that makes the best story. Reading only the headlines is the horizontal-logic test.
A good dashboard page works like a short story you can read at three depths: a headline that states the conclusion, a few supporting visuals that show why, and detail on demand for those who want to check. The Toolkit shows a page layout built this way.
Study the "before". Tick every flaw you can see. Be careful: some listed flaws are decoys.
Correct ticks go green, wrong ticks red, and missed flaws amber, each with a reason.
Switch to "After" and read what changed. Each flaw is tagged with the lesson that fixes it.
All data in the cases is fictional. Titles in the "after" views are computed from the data shown, so the claims match the chart.
Five hubs hit between 93.8% and 96.1% on-time delivery. Slide the baseline of the bar chart and compare what the eye sees with what the numbers say.
Enter the share of delays by cause (any numbers; they are converted to percentages). Try to rank the pie slices, then look at the bars.
Monthly volume (bars, left axis) and on-time departures (line, right axis). Move the right axis range and the relationship appears to change. The numbers never do.
Inbound pallets received vs. processed per day, by month. Toggle clutter on and off, then add focus aids. Notice when the story becomes visible.
Two rounds, same task: count how many times the digit 3 appears. In round 2 one visual cue is added. Your own time is the data. Results vary by person and screen; this is a demonstration, not a study.
Late deliveries by hub over the last four weeks. Pick the audience: the data stays put, but the headline, the highlight and the ask change with the decision maker.
Each exercise scores 100 points: 50 for the data you pick, 25 for the visual and 25 for the title. For data, each core item you include earns full credit, supporting items earn half credit, and noise or misleading items cost points. Missing a core item costs you the points that item would have earned.
Teaching heuristic This is my own scoring scheme for practice. In real projects, the right answer depends on your audience and decision, so use the explanations, not the number.
Application note This layout is my application of the book's lessons to dashboards, not a template from the book.
0 of 0 checks done
| Term | Meaning |
|---|---|
| Explanatory analysis | Communicating one specific message to a specific audience. |
| Big Idea | One complete sentence with your point of view and what is at stake. |
| 3-minute story | What you would say if you had three minutes. |
| Storyboard | A visual outline of the communication, made on paper first. |
| Slideument | One file that tries to be both a presentation and a document. |
| Cognitive load | Mental effort needed to take in information. |
| Signal-to-noise | Share of what you show that carries the message. |
| Gestalt principles | How people group things: proximity, similarity, enclosure, closure, continuity, connection. |
| Preattentive attributes | Features noticed before conscious thought, such as colour, size and position. |
| Visual hierarchy | Ordering elements by importance so readers know where to look first. |
| Affordance | A cue that shows how something is meant to be used. |
| Action title | A title that states the conclusion or ask rather than the topic. |
| Slopegraph | A line chart of two time points that shows direction and rate of change across categories. |
| Small multiples | Several small charts with identical axes, placed side by side. |
| Horizontal logic | Reading only the titles tells the story. |
| Vertical logic | Everything on one page supports its title. |
| Reverse storyboarding | Summarizing each page of the finished piece to compare with the plan. |
Pick one dashboard you own. Write its Big Idea, run the checklist, rebuild the top page with a headline, one accent colour and no legend, then show it to someone who has never seen it. Where their eyes go first tells you whether it works.