What is the problem or goal you're trying to solve or accomplish?
The user wants bar colors to reflect the value each bar represents, so viewers can spot high, low, or out-of-range values at a glance. Currently, bar colors come from the color scheme and are assigned per series or dimension, not by the bar's value. The customer wants to:
define value thresholds (ranges) and assign a color to each, for example below 100 = red, 100–500 = yellow, above 500 = green, and
optionally choose a number of thresholds and have them calculated automatically from the data, for example evenly spaced ranges or quantiles.
How are you solving it currently?
The customer didn't describe a current workaround. A possible one, not yet tested, is to create a "bucket" column with a CASE WHEN expression that labels each value range, use it as the chart's dimension or breakdown, and map each bucket label to a fixed color with custom label colors in the dashboard's JSON metadata (label_colors). This has real limitations:
Row-level vs aggregated values. A calculated column is evaluated per row, before aggregation. If the threshold should apply to the aggregated metric (for example, SUM(revenue) per bar), the bucketing has to happen in a virtual dataset that pre-aggregates the data. The bucket labels then only stay accurate for that exact grouping.
Static thresholds. Thresholds are hardcoded in SQL, so automatic thresholds based on the data aren't practical. Quantile-style functions such as NTILE in a virtual dataset are possible but brittle once filters change the data.
Chart side effects. Using the bucket as a breakdown turns it into separate series. That changes the legend and tooltips and may need the stacked bar style to avoid offset bars.
Dashboard-only colors. label_colors lives in dashboard metadata, so the mapping doesn't follow the chart outside that dashboard.
What is your recommended solution?
Add a conditional formatting option to the bar chart's Customize panel where users can:
define value thresholds and assign a color to each range, applied to the bar's metric value, and
alternatively, enter a number of thresholds and have the ranges calculated automatically from the data.