Tableau Course
Dashboard Best Practices
Technical skill builds charts. Design judgement turns charts into dashboards people actually use. This lesson covers the principles, layout standards, and habits that distinguish dashboards that get ignored from dashboards that drive decisions.
Start With a Question, Not a Chart
The most common source of bad dashboards is starting with data and asking "what can I show?" instead of starting with the viewer and asking "what decision do they need to make?" Every chart on a dashboard should answer a specific question. If you cannot write the question a chart is answering, the chart probably should not be there.
The Visual Hierarchy — Where the Eye Goes First
Viewers scan a dashboard the way they scan a page — top left first, then across, then down. Visual weight (size, colour, contrast) signals importance. Deliberately placing the most important information where the eye lands first is the single most powerful layout decision you will make.
Choosing the Right Chart Type
Chart type is not a style choice — it is a communication choice. The right chart makes the relationship in the data immediately obvious. The wrong chart forces the viewer to do interpretive work that the chart designer should have done for them.
| Task | Best Chart Type | Avoid |
|---|---|---|
| Compare categories | Horizontal bar chart — easiest to read category labels | Pie chart for more than 3 segments — impossible to compare slices accurately |
| Show change over time | Line chart — encodes trend visually through slope | Bar chart for long time series — too many bars become unreadable |
| Show part-to-whole | Stacked bar or treemap — proportions visible at a glance | Donut chart — the hole removes the visual centre and reduces accuracy |
| Show correlation | Scatter plot — both variables encoded as position | Line chart or dual axis — implies causation or sequence |
| Show geographic distribution | Filled map or symbol map depending on whether boundaries or magnitudes matter more | Bar chart — loses the geographic context that is the whole point |
| Show a single KPI | Big number tile with a comparison (vs prior period, vs target) | Gauge chart — visually complex, low data-ink ratio, hard to read precisely |
Colour as Signal, Not Decoration
Colour is the most misused element in dashboard design. When everything is colourful, nothing stands out. When colour is used sparingly and purposefully, it directs attention to exactly the information that matters.
Every bar in a different colour "because it looks nicer." A rainbow palette where adjacent colours are hard to distinguish. Colour that represents a dimension already encoded on an axis — double encoding with no benefit. Red and green used together without an accessible alternative — 8% of men are red-green colour blind.
All bars one neutral colour except the one that is the focus of the question — it gets a brand or accent colour. A diverging palette centred on zero for profit/loss. Sequential blue for "how much" comparisons. Red alerts only on genuine threshold violations — if everything is red, nothing is urgent.
| Palette Type | Use For | Example |
|---|---|---|
| Sequential | One variable ranging from low to high — heat maps, geographic density | Light blue → dark blue for sales volume |
| Diverging | A measure with a meaningful midpoint — profit/loss, variance from target | Red → white → blue centred on zero |
| Categorical | Distinct groups with no inherent order | Tableau 10 or ColorBrewer qualitative palettes — max 6–8 colours |
| Single hue + highlight | Bar or line chart where one item needs attention | All bars grey except the selected or anomalous item in orange |
Titles and Labels
The title of a chart is the most-read text on the dashboard. Most chart titles are wasted — they describe the data ("Sales by Category") instead of communicating the insight ("Technology Drives 37% of Total Sales"). A chart with an insight title answers the viewer's question before they even look at the chart.
Tells the viewer what the axes are — which they can already see. Communicates nothing about what to notice or decide.
Tells the viewer what to notice before they read the chart. The chart then confirms and adds detail to the already-stated finding.
Data labels on individual marks should be used selectively. Labels on every bar in a 20-bar chart create visual noise. Labels on the top 3 bars, or only on bars above a threshold, add information without clutter. Use Marks card → Label → Min/Max to label only the extreme values, or a calculated field to conditionally show labels only on notable marks.
Filter Design — Helping Viewers Without Overwhelming Them
Filters are powerful but every filter you expose to a viewer is cognitive load. The goal is to expose the filters the viewer genuinely needs for their workflow, and hide or remove the rest.
| Filter Style | Best Use | Avoid When |
|---|---|---|
| Single value (dropdown) | High-cardinality fields like Country or Product — dropdown prevents a huge list from dominating the dashboard | Low-cardinality fields where the viewer benefits from seeing all options at once |
| Multiple values (list) | Low-cardinality fields like Region (4–6 values) where multi-select is expected | Fields with more than 10–12 values — the list becomes unwieldy |
| Relative date filter | Time-based dashboards where "last 30 days" or "this quarter" is the natural frame of reference | Historical analysis dashboards where specific date ranges matter |
| Slider | Continuous ranges like price or quantity where the viewer wants to set a threshold | Discrete categories — a slider on a categorical field is confusing |
| Parameter control | When the filter needs to drive calculated fields — "show top N", "switch metric", "change date grain" | Simple include/exclude use cases — parameters are overkill and harder to maintain |
The Dashboard Design Checklist
Run through this checklist before sharing any dashboard. Each item represents a category of mistake that appears in real production dashboards.
The best dashboard design advice has nothing to do with Tableau: show the draft to one real user before you think it is finished. Watch them try to answer a question without your help, and do not explain anything. Where they hesitate or click and nothing happens — those are your redesign moments. No checklist catches everything a five-minute user test reveals. On the technical side: before adding a chart, name the specific question it answers and the specific person who will use it to make a decision. If you cannot name both, do not add the chart.
Practice Questions
1. A bar chart is titled "Revenue by Category". The real finding is that Technology accounts for 37% of total revenue and is growing. How should the title be rewritten to follow best practices?
2. A bar chart shows profit and loss by product — some bars are positive and some are negative. Which colour palette type is most appropriate and why?
3. A dashboard needs to compare sales across 8 product categories. A colleague suggests using a pie chart. Which chart type is more appropriate and why?
Quiz
1. On a standard dashboard layout, which zone should receive the highest visual weight and the most important information?
2. A chart needs to show how monthly revenue has changed over 24 months. Which chart type is most appropriate?
3. Before publishing a dashboard to 200 salespeople, a designer wants to validate that the layout communicates clearly. What is the single most effective way to test this before release?
Next up — Lesson 49: Full Dashboard Project — building a complete end-to-end sales performance dashboard from raw data to a published, interactive, mobile-ready output applying every technique from this course.