Open a dashboard you inherited and count the tiles. Six? You can work with that. Twenty-two, half of them gauges, a rainbow pie in the corner? Someone lost the plot, and now every question takes three minutes of hunting. Widgets are the atoms of a dashboard. Pick the wrong ones and no amount of clever layout saves you. This page is about which tiles earn their pixels and which ones you should delete on sight.
What a dashboard widget is
A widget is one self-contained tile that answers one question. That is the whole definition. A KPI card answers “where are we against target.” A trend line answers “which way is this moving.” A table answers “which rows are driving the number.” Each tile carries a single job, and the job should be obvious from two feet back without reading the axis labels.
The trap is treating a widget as decoration you drop in to fill a gap in the grid. Every tile costs the reader attention, and attention is the one budget on a dashboard that never refills. If a widget does not change a decision, it is not neutral. It is a tax on the widgets that do. Before a tile goes on the screen, name the decision it supports. No decision, no tile.
Most tools ship the same short menu of widget types: the card, the line, the bar, the table, and a pile of novelty visuals you can safely ignore. The good news is that a genuinely useful dashboard runs on about four of them. The right chart type for each question is a longer conversation, and most of the answers land on a small, boring set. Boring is the point.
The KPI card: one number, one target
Start here, because the KPI card is the widget people build worst. A card visual shows a single headline number. Revenue this month. Active users today. Win rate this quarter. One number, big enough to read across a room. The mistake is stopping there. A number with no reference is trivia. Is 84 good? You have no idea, and neither does the person you built this for.
A KPI must have a target. So the card needs three things stacked together: the value, the target it is measured against, and a small trend so the reader sees direction without a second tile. Modern tools make this native. The 2025 Power BI card visual embeds a sparkline directly under the number, and the older KPI visual combines value, target, and trend in one frame. Use them. A card that shows 84 next to “target 90” and a line drifting down tells a complete story in one glance. A card that shows 84 alone tells nothing.
Format the number like an adult reads it. Round to significant digits. “$1.2M” beats “$1,238,411.77” on a headline tile every time, because nobody is reconciling the books off a dashboard. Save the exact figure for the detail table below. If the number carries a direction that matters, color it against target, not against zero. Green because you beat plan, red because you missed it. Green because the value happens to be positive is meaningless and trains the reader to ignore your colors.
One card per metric. Do not staple six numbers into a single tile and call it efficient. A row of separate cards along the top reads faster, and it lets you kill a metric later without surgery. If you are building a whole board around headline numbers, the layout and cadence questions live in the KPI dashboard build.
Trend widgets: the line and the sparkline
A number tells you where you are. It says nothing about where you are heading, and heading is usually what the decision hangs on. That is the trend widget’s whole reason to exist.
The line chart is the default, and it earns the spot. A line chart shows a trend over time cleanly: x-axis is the date, y-axis is the value, and the slope does the talking. People read slopes instinctively. Up-and-to-the-right needs no legend. Keep the line to one or two series. The moment you plot five overlapping lines you have built a plate of spaghetti, and the reader gives up. If you need many series compared, that is a different question and probably a different chart.
The sparkline is the line chart’s compact cousin. A sparkline compresses a trend into a small space, no axes, no gridlines, just the shape. It lives inside other widgets. Tuck one under a KPI card, drop one in each row of a table, and suddenly every number carries its own recent history without stealing a grid cell. The sparkline is the highest information-per-pixel widget on the board. Use it wherever a full trend chart would be overkill but a bare number feels blind.
One rule that saves you from a common lie: do not truncate the y-axis to make a trend look dramatic. Starting the axis at 82 instead of 0 turns a rounding wobble into a cliff. For a headline trend that someone acts on, either start at zero or label the axis loudly so the reader knows the scale is zoomed. The data visualization best practices page goes deeper on axis honesty, and it matters more than it sounds.
Table and detail widgets
Analysts underrate the table, and they are wrong to. A table shows precise values, and precision is exactly what a chart throws away. When the reader has already seen the trend and now needs to know which twelve accounts, which three regions, which SKU, they need rows and columns, not another shape. The table is where “why” lives.
Detail belongs at the bottom right. The eye lands top-left and drifts down, so the summary sits up top and the granular table sits where someone goes only after a headline tile made them curious. This is the natural order of a dashboard: status first, explanation on demand. A drill-down reveals detail on demand, which means the top of the board stays calm and the depth is one click away instead of crammed into the first screen.
Build the table to be scanned, not admired. A few practical rules earn their keep:
- Sort by the column that matters, usually the one you want the reader to act on, and sort it descending so the worst or biggest rows sit at the top.
- Right-align numbers so the digits line up and the eye compares magnitudes without doing math.
- Use a single subtle bar or heat shade inside the key column instead of ten decorated columns fighting for attention.
- Cap the visible rows. Ten to fifteen on screen, the rest behind a scroll or a filter. Nobody reads row 40 on a dashboard.
A table with an embedded sparkline per row is the quiet workhorse of good operational dashboards. Account name, current value, target, and a tiny trend, all in one line. That single row often replaces four separate charts. If you want to see it wired up with real measures behind it, the Power BI dashboard examples walk through the DAX that feeds these tiles.
Widgets to avoid: the gauge and the oversized pie
Now the deletions. Some widgets look impressive in a template and fail the moment real data hits them. Two offenders show up on nearly every bad dashboard.
The gauge is the first to go. A gauge shows one value against a range, which sounds useful until you notice it eats a large square tile to encode a single number and a threshold. A KPI card with a target line does the identical job in a quarter of the space and reads faster. The speedometer look flatters the builder and slows the reader. If you are showing one value against a target, use a card. If you are showing several, use a bar. The gauge wins in zero of those cases.
The pie chart is the second, and it fails in a specific, predictable way. A pie chart shows parts of a whole, and it is tolerable at two or three slices where the split is obvious. Past that it breaks. A pie chart breaks down above five slices, because humans judge angle and area badly, and once the wedges get close in size nobody can rank them by eye. You end up reading the legend and the percent labels, at which point the pie added nothing a sorted list would not do better. For real part-to-whole work, a stacked bar or a plain ranked bar wins. The full argument, and the cases where a pie is genuinely fine, sit in the chart selection guide.
Two more habits worth breaking. Skip the 3D anything. A 3D bar or pie distorts the values by tilting the geometry, so the front slice looks bigger than the back for no reason but perspective. And retire the dual-axis chart used to fake a relationship. Two measures on two different scales in one frame can be manipulated into any story you want by choosing the axis ranges, and the reader has no defense. If two measures genuinely relate, prove it with a scatter plot instead.
The short list that actually ships
Here is the whole widget kit for a dashboard that people open twice a day, mapped to the question each tile answers.
| Question the reader has | Widget that answers it | Skip |
|---|---|---|
| Where are we against target? | KPI card with value, target, sparkline | Gauge |
| Which way is this moving? | Line chart, or a sparkline inside a card | Truncated-axis line |
| Which rows are driving it? | Sorted table, detail at bottom right | Wall of unsorted rows |
| How do categories compare? | Horizontal sorted bar | Pie past five slices |
| How is the whole split up? | Stacked bar | 3D pie |
Notice how short that is. Card, line, sparkline, table, bar. Five shapes carry almost every dashboard worth building. The discipline is not learning exotic visuals, it is refusing the ones that flatter you and slow the reader. Pick the boring tile that answers the question in one glance, put the summary up top and the detail at the bottom, and let the fancy widgets stay in the template gallery where they belong. Where each tile sits on the screen is its own craft, covered in the dashboard design guide.