Most Power BI dashboards fail not because the DAX is wrong, but because the design is. Users open the report, see 15 charts, can’t find what they need, and go back to Excel. This guide covers 12 design principles that separate reports people use from reports people ignore.
Principle 1: One page, one question
Every page should answer one business question. Before building, write the question at the top of the page (literally — in a text box). If you can’t state it in one sentence, split the page.
Bad: “Sales dashboard” (too vague, ends up with 20 charts) Good: “Which regions are underperforming vs. target this quarter?”
Every visual on the page should contribute to answering that question. If a chart doesn’t, move it to another page or delete it.
Principle 2: Inverted pyramid layout
Put the most important information at the top, details below. Users scan top-to-bottom, left-to-right (in Western reading order).
[KPI Cards: Revenue | Target | Variance]
[Main chart: trend over time]
[Secondary charts: by category, by region]
[Detail table: drill-down]
KPIs at the top tell users if they need to dig deeper. Charts in the middle show context. Details at the bottom support drill-down.
Principle 3: Limit to 8-10 visuals per page
Each visual runs a DAX query on every interaction. More visuals = slower pages. But the bigger issue is cognitive — users can’t process 15 charts at once.
If you need more, use:
- Bookmarks to toggle between views
- Drill-through to move details to another page
- Tab navigation to split content logically
Principle 4: Choose the right chart type
The wrong chart hides patterns. Match the chart to the question:
| Question | Chart type | Why |
|---|---|---|
| How does X change over time? | Line chart | Shows trends clearly |
| How do categories compare? | Bar chart (horizontal) | Easy to read labels |
| What’s the composition? | Stacked bar | Shows parts of whole |
| What’s the distribution? | Histogram or scatter | Reveals outliers |
| What’s the correlation? | Scatter plot | Shows relationship |
| What’s the single value? | Card | No chart needed |
Avoid: pie charts with more than 5 slices, 3D charts, donut charts for precise comparison. See our visual types guide for the full reference.
Principle 5: Use color with intent
Color should encode meaning, not decoration.
Do:
- Use one accent color for the “story” (e.g., blue for the current period)
- Use gray for context (e.g., previous period for comparison)
- Use red/green only for bad/good (and always pair with a label — 8% of men are colorblind)
Don’t:
- Use a rainbow palette for categorical data
- Color every bar a different color
- Use red for emphasis (it triggers alarm)
Color palette guidelines
- Background: white or very light gray
- Text: dark gray (#333), not pure black (easier on eyes)
- Primary accent: one color (blue is safe and professional)
- Secondary: gray for comparison
- Semantic: green (#28a745) for positive, red (#dc3545) for negative
Principle 6: Label everything clearly
Users shouldn’t have to guess what a number means.
- KPI cards: include the metric name, value, and unit ($, %, etc.)
- Charts: include axis titles and units
- Slicers: use clear labels (“Date range” not “Date”)
- Tooltips: customize to show relevant context (see our tooltips guide)
Avoid abbreviations unless they’re universally understood. “Rev” could mean revenue, revisions, or reverse.
Principle 7: Design for the user, not for yourself
Different audiences need different dashboards:
Executives (view weekly, 2 minutes per session):
- High-level KPIs and trends
- Exceptions and alerts
- Minimal interaction — they want answers, not tools
Managers (view daily, 5-10 minutes):
- Drill-down capability
- Comparisons (vs. target, vs. last period)
- Ability to filter by their team or region
Analysts (view ad-hoc, 30+ minutes):
- Detailed data
- Export options
- Flexible filtering and slicing
Before building, ask: who will use this, how often, and what decisions will they make?
Principle 8: Make filters obvious
If users can’t find the slicers, they think the data is wrong.
- Place slicers at the top or left of the page (consistent across pages)
- Use clear labels
- Show the current selection state
- Consider using the “Filter” pane for advanced filtering and slicers for common filters
Principle 9: Provide context for numbers
A number without context is meaningless. “$1.2M revenue” — is that good or bad?
- Compare to target: $1.2M vs. $1.5M target (80% achieved)
- Compare to previous period: $1.2M, up 12% vs. last month
- Compare to benchmark: $1.2M, top quartile in the region
Use KPI cards with target values, or add a reference line to charts showing the target.
Principle 10: Use consistent formatting
Inconsistent formatting makes dashboards feel amateurish.
- Number formats: $#,##0 for currency, 0.0% for percentages, #,##0 for counts
- Date formats: choose one and use it everywhere (e.g., “Jan 2026”)
- Font: use the same font family throughout (Segoe UI is Power BI’s default)
- Font size: titles 14pt, headers 12pt, body 10pt
Principle 11: Test with real users
The biggest design mistakes are invisible until a user tries the dashboard.
- Watch someone use it for the first time (don’t explain it first)
- Note where they hesitate, click the wrong thing, or ask questions
- Ask them to complete a specific task (“find which product had the lowest sales last month”)
- Iterate based on what you observe
You’ll be surprised how often your “obvious” design isn’t.
Principle 12: Keep it maintainable
A dashboard that only you can maintain is a liability. Design for handoff:
- Name measures clearly (
Revenue YTDnotM1) - Group measures into display folders
- Document the data model (use the Model View diagram)
- Comment complex DAX with
//or/* */ - Keep a changelog of major updates
Common design anti-patterns
The “kitchen sink” — 20 visuals on one page, showing everything. Users can’t find anything.
The “rainbow” — every bar a different color. Color encodes nothing.
The “3D pie” — looks fancy, distorts proportions, impossible to read precisely.
The “no context” dashboard — numbers without comparison. Is $1.2M good?
The “static” report — no slicers, no drill-down, no interactivity. Just a PDF with extra steps.
FAQ
How many pages should my report have? As many as needed to answer distinct business questions — typically 3-8. Each page should have a clear purpose stated in its title.
Should I use custom visuals? Only when a native visual can’t do the job. Custom visuals add dependencies, may not be supported in all environments, and can break with Power BI updates. Native visuals are always preferred.
What’s the ideal load time? Under 3 seconds for the first visual, under 8 seconds for a full page. If you’re slower, see our performance optimization guide.
What’s next
- Learn visual types to choose the right chart
- Master bookmarks and navigation for multi-page reports
- Use conditional formatting to highlight exceptions
- See our performance guide to keep dashboards fast