Charts

Chart types, when each one is right, and how to configure axes, series, and formatting.

A chart binds a dataset to a visual form. It stays bound: when the dataset syncs, the chart updates.

Create a chart

Pick a dataset, then a chart type, then assign fields to the chart’s roles — usually an x (dimension), a y (measure), and optionally a series (split).

Text

Dataset:  Orders
x:        created_at   → truncated to month
y:        total        → sum
series:   status
type:     stacked bar

Choosing a type

Type	Use for	Avoid when
Line	A measure over time	Categories with no order
Area	Cumulative totals over time	Comparing many series
Bar	Comparing categories	More than ~25 categories
Stacked bar	Composition within a total	Comparing individual segments precisely
Grouped bar	Comparing series side by side	Many series — it gets noisy
Scatter	Relationship between two measures	Categorical axes
Table	Exact values, many columns	Spotting shape or trend
Metric	One headline number	Anything needing context
Map	Values by geography	Data without a geographic field

When in doubt, bar for categories, line for time

Most chart-choice mistakes are a pie chart that should have been a bar chart, or a bar chart of dates that should have been a line.

Aggregation

Every measure needs an aggregation, because a chart shows fewer rows than the dataset has.

Aggregation	Notes
`sum`	Default for amounts
`avg`	Watch for nulls — they are excluded, not zero
`count` / `count distinct`	`count distinct` is heavier on large datasets
`min` / `max`	Useful for ranges
`median` / `p95`	Better than `avg` for latency and price

avg of an avg is not an avg

Averaging a column that is already an average weights every group equally regardless of size. Aggregate from the raw measure instead.

Time handling

Setting	Options
Truncate	hour, day, week, month, quarter, year
Timezone	Team default, or per chart
Range	Fixed dates, or relative (`last 90 days`)
Compare	Previous period, or same period last year

Partial periods look like a crash

The current month is incomplete, so it always dips. Enable Exclude current period on trend charts, or label it clearly.

Formatting

Number format — decimals, thousands separator, €/% prefix or suffix, compact (1.2M).
Colour — a team palette by default; assign fixed colours per series so paid is the same green everywhere.
Axis — override the label, set an explicit min/max, or use a log scale.
Null handling — show as gaps, as zero, or interpolate. Gaps are the honest default.

A truncated y-axis exaggerates change

Starting a bar chart’s axis at a non-zero value makes a 2% change look like a collapse. For bars, keep zero in range; for lines, truncation is acceptable when labelled.

Transforms

Applied after the query, before rendering:

Transform	Effect
Top N	Keep the largest N, group the rest as “Other”
Cumulative	Running total across the x axis
Moving average	Smooth over N periods
Percent of total	Convert to share
Derived field	An expression over existing fields

Text

margin_pct = (revenue - cost) / revenue * 100

Cross-dataset charts

Coming soon

A chart reads one dataset today. Combining two requires a shared mapped field, and field mapping is still in development — see Mappings.

Export

Format	Contents
PNG / SVG	The rendered chart
CSV	The chart’s result rows, post-transform
Query	The generated SQL, to run yourself

Exports respect permissions — a Viewer can export what they can see.