The 24 widget types in clariBI and how to choose the right one

A chart is a tool for one question. Pick the wrong one and people read the wrong answer, or give up. clariBI offers 24 visualization types for dashboard widgets. Most dashboards need five or six of them. This guide goes through all 24 in groups, says what each one is for and when it is the wrong pick, and ends with a short method for choosing.

Where you choose a type

In the dashboard builder, every metric has a Visualization Type. Suggested metrics come with a recommended type; a custom metric lets you pick any of the 24. In a quick AI-generated dashboard, the AI picks the type for each metric, and in manual mode the template decides. The widget types reference in the help center has the same list in table form.

One practical limit: a dashboard holds a limited number of widgets, and complex types (tables, heatmaps, sankey diagrams) take more of that room than a metric card. When a type no longer fits, the builder marks it as having no room.

Single numbers: metric card, gauge, bullet chart

These answer "where are we right now?"

• Metric card. One headline number: revenue this month, open tickets, active users. It is the fastest thing to read on a dashboard. On its own it lacks context, so put a line chart of the same metric next to it, or phrase the card as a comparison.
• Gauge. One value inside a range, such as capacity used as a percentage. Only use it when there is a real minimum and maximum. A revenue gauge with an invented ceiling tells the reader nothing.
• Bullet chart. One value against a target, with bands for poor, fair and good. It carries more information than a gauge in less space, which makes it the better choice for quota or budget tracking.

Trends over time: line, area, candlestick

These answer "how has it changed?"

• Line chart. The default for anything measured over time: daily orders, weekly sign-ups, monthly churn. One to four lines read well. Past that, split into separate charts on the same scale.
• Area chart. A line with the space below filled in. Use it when the volume matters, for example total sessions. With more than three or four series the filled areas hide each other.
• Candlestick chart. Open, high, low and close per period. It is built for price data, such as stock or commodity prices from a market data connector. Do not force ordinary business metrics into it.

Comparing categories: bar, radar, waterfall, funnel

These answer "how do these compare?" or "where did it go?"

• Bar chart. The most useful chart there is. Revenue by product, tickets by status, spend by channel. It works well for 3 to 15 categories. Sort the bars by value unless the categories have a natural order, and start the value axis at zero.
• Radar chart. Several measures for a few items, such as three suppliers scored on price, quality, speed, support and terms. It gets hard to read with many items or with measures on very different scales.
• Waterfall chart. How a start value becomes an end value. The classic use is a revenue bridge: starting MRR 50,000, plus 6,000 new, plus 2,000 expansion, minus 3,000 churn, ending at 55,000. Only use it when the steps really add up to the total.
• Funnel chart. Drop-off between ordered stages: visits, sign-ups, trials, paid. If the stages are not a sequence, use a bar chart.

Parts of a whole: pie, treemap, sunburst, sankey

These answer "what is it made of?"

• Pie chart. Shares of a total with 2 to 6 slices. People compare angles badly, so with more slices or slices of similar size, a sorted bar chart is clearer.
• Treemap. Shares with many items, each rectangle sized by value. Good for 10 to 50 products or cost lines. Treemaps cannot show negative values.
• Sunburst chart. Shares across two or more levels, such as region and then country. With only one level, use a pie or bar chart.
• Sankey diagram. Flows between groups, such as traffic source to landing page to outcome. It needs data where things actually move from one group to another.

Distributions: histogram, box plot, violin plot

These answer "how are the values spread?"

• Histogram. How often each range of values occurs, for example order sizes in 20-dollar bins. It shows whether most orders are small with a long tail of large ones, which an average hides.
• Box plot. Median, quartiles and outliers, side by side for several groups: delivery times by carrier, deal sizes by rep. Readers who have not seen one before may need a line of explanation.
• Violin plot. A box plot that also shows the shape of the distribution. Useful with larger samples; with a few dozen points it can suggest patterns that are not there.

Relationships: scatter, bubble, heatmap, parallel coordinates, network graph

These answer "do these move together?"

• Scatter plot. Two numbers against each other, one point per item: ad spend and revenue per campaign, discount and margin per order. It shows correlation, not cause.
• Bubble chart. A scatter plot with a third number as bubble size. Overlapping bubbles make it unreadable, so keep the item count modest.
• Heatmap. A value across two dimensions, such as orders by weekday and hour. It is very good at showing when things happen.
• Parallel coordinates. Many measures per item at once, one vertical axis per measure. It suits an analyst exploring data more than a manager reading a summary.
• Network graph. Connections between items, such as which products are bought together. Only use it when your data really contains links.

Maps and tables: choropleth map, table

• Choropleth map. Values shaded by country or region. It needs a location column and is worth the space only when geography matters to the decision.
• Table. Exact values and row-level detail. Tables are for looking things up, so put them at the bottom of a dashboard, under the charts that show the pattern.

Choosing by question

When you are unsure, write the question in plain words and start here:

Two more checks before you commit. First, count the data points: one number is a card, a few categories are a bar or pie, a time series is a line. Second, think about who reads it. An owner checking the business on a phone needs cards and one trend line, not a violin plot.

For layout, a pattern that works: three or four metric cards across the top, one or two trend or comparison charts below, a table at the bottom. The dashboard layout guide explains the 4-column grid and the five widget sizes.

Mistakes that make charts misleading

• Bar charts that do not start at zero. A bar going from 98 to 102 looks like a big jump on a cut axis. Line charts can start above zero; bar charts should not.
• Pie charts with ten slices. Switch to a sorted bar chart.
• Too many series on one chart. If it takes more than a few seconds to read, split it.
• Red and green as the only signal. Roughly 1 in 12 men has some form of color vision deficiency. Add labels or arrows as well.
• Titles that only name the metric. "Monthly revenue, 2026 vs. 2025" says more than "Revenue".

There are more examples in 10 data visualization mistakes to avoid.

How clariBI picks a chart when you ask a question

When you ask a question in plain English, clariBI first computes the answer with regular analytics code, and the shape of the result suggests the chart: a single total becomes a metric card, a value over time a line chart, a ranked list a bar chart. The wording counts too: "how has revenue changed" leans toward a line, "which product sells most" toward a bar. You can always ask for another type ("show that as a table"). How the AI chooses charts covers the logic in more detail, and the widget types page in the docs lists what each type expects from the data.

To try this on your own numbers, connect one source, open the dashboard clariBI generates for it, and swap one chart for the type the table above suggests. Then see whether the question gets easier to answer.