Charts worth stealing: the design calls BI should ship

At a dbt Summit 2026 breakout session titled "Charts worth stealing: The design calls BI should ship," RJ Andrews, chartographer on the dbt Charts team at dbt Labs, walked through the design decisions behind dbt Charts, the open source charting language dbt Labs launched earlier in the week at dbtcharts.com.

Andrews opened by noting that most named chart types, bar charts, area charts, line graphs, scatterplots, are really just variations on the same underlying grammar, each one a small step away from the next. He demonstrated the point live, morphing a single default bar chart into a stepped area chart, a line graph, and a scatterplot by changing only a couple of lines of YAML.

dbt Charts' default bar chart puts vertical series labels on the right side instead of the left. Andrews said that's deliberate, since the most recent data is usually the most important, and right-side labels keep a viewer's eye where the newest values live.

The tool also switches automatically between monthly and quarterly axis labels depending on how dense the data is, and shows the year only at the start of a series or when the year changes, rather than repeating it on every label.

dbt Charts ships with three built-in themes: Clarity, the default light theme with serif prose type; Neon, a high-contrast dark mode; and Paper, a cream-background theme Andrews said was inspired by the Financial Times and the Economist. Moving to dark mode required more than swapping colors, since the color palette and typography both had to change to stay readable against a black background.

On color, Andrews said dbt Charts' two main categorical palettes both start with blue in the first slot. He noted that when he surveyed color palettes across digital BI tools, nearly all of them defaulted to blue first and then jumped to a high-contrast orange or red, the pattern behind Tableau's familiar blue-and-orange dashboards.

Andrews said that kind of high-contrast pairing draws a viewer's eye to the contrast itself rather than to the data, so he designed dbt Charts' palettes to stay more harmonious when a chart only has a few series. The tool also keeps the same color assigned to the same category across every chart in a dashboard, and uses a darker shade for single-series charts than for multi-series charts to preserve legibility.

Every color palette in dbt Charts was checked against roughly half a dozen types of color vision deficiency, weighted by how much of the population has each type, according to Andrews. Pattern fills, a way to distinguish categories without relying on color alone, are built but not yet publicly available.

Andrews showed an internal testing gallery of dozens of chart examples he uses to track known issues and confirm that edge cases, like data too dense to render legibly, keep working as the tool evolves. When a chart's data is too dense to display clearly, dbt Charts surfaces a warning during the review process rather than letting a broken chart reach production.

Tables built with dbt Charts keep decimal points aligned without adding false precision, so a value with no decimal in the underlying data doesn't get a fabricated ".0" just to make columns line up. Charts also support interactive tooltips that highlight an entire stacked segment and its total, and dashboard filters update the URL so a specific view can be shared or bookmarked directly.

In one worked example, a dual-axis line chart paired nominal gigawatt values with a percentage axis calculated to share the same gridlines, an approach Andrews said was inspired by an 1870s balloon aeronautics chart that plotted altitude in both feet and meters.

Andrews said users aren't limited to chart types defined explicitly in the dbt Charts YAML syntax, since an agent can generate less conventional forms, like lollipop charts or bump charts, from a plain-language request. He added that the bump chart's name comes from nineteenth-century Cambridge and Oxford boat races, where crews try to bump the boat ahead of them out of position.

dbt Charts and its dashboards became open source with this week's launch, and Andrews closed the session by pointing attendees to dbtcharts.com to read the language documentation and review example dashboards.

Watch every keynote and breakout from dbt Summit 2026 on demand. Watch dbt Summit 2026 on demand →