Precise third-variable comparison belongs in facets or a separate chart

What it is

When the third variable must support exact differences, strict ranks, or threshold decisions, re-encode it on position or length with a common scale. A sorted dot plot, bar chart, or table can sit beside the bubble chart and coordinate selections; facets may represent genuine categories or analytical conditions. “Facets” in the title does not mean arbitrarily binning a continuous third variable. Binning removes within-interval differences and usually does not turn area comparison into precise magnitude reading.

Why it happens

Position and length supply a common reference and reduce area-ratio estimation. Separate coordinated views let one chart preserve the x-y relation and multivariate context while another supports exact third-variable comparison. Facets for nominal conditions still require comparable panel scales. If a continuous value is binned into panels, panel identity represents an interval rather than the exact value. Brushing and linking preserve entity identity only when mappings, filters, and time state remain synchronized.

Studying it

Compare a bubble chart alone with a bubble plus coordinated dot plot or bar chart, a table, and defensible facets. Test exact read-off, difference, rank, threshold, x-y relation, and cross-view entity matching; measure error, time, switching cost, and lost selections. When binning a continuous variable, separately test boundary cases and hidden within-bin differences. Coordinated views also require filter, update, and export-state tests rather than static screenshots alone.

Where it stops holding

Another view costs space and cross-view integration effort; if the third variable is low-risk context, that cost may exceed the benefit. Facets suit categories, conditions, or a small number of meaningful intervals, not continuous-value precision in general. Shared scales aid comparison, but extreme ranges may compress most values and require transformation, local detail, or a table rather than arbitrary independent scales. Aggregation must not hide minority or high-consequence records, and readers need a route back to entities.

Applying it

• Express the decision task as read-off, difference, rank, or threshold, then choose common-scale position, length, or a table column for the third variable.
• Link comparison and bubble views through brushing, focus, and stable entity identifiers; synchronize filters, units, time scope, and data version.
• Facet only by meaningful categories or explained intervals, keep panel scales comparable, and expose original values within each interval.
• Test both precision and x-y relation tasks. If the new view lowers magnitude error but breaks entity matching, improve linking and labels instead of returning to precise area reading.

Related

• Same group: U2.06.1 Bubbles carry a third variable on the least precise common channel · U2.06.2 Map values to bubble area, not radius · U2.06.3 Big bubbles occlude small ones; draw in size order and lower opacity · U2.06.4 Bubble size needs a reference-bubble legend
• Adjacent: U2.15.1 Small multiples replace overlay with repeated like-for-like panels · U2.17.1 A mismatched chart forces readers to compute the answer mentally
• Search terms: coordinated views · re-encoding · precise comparison · faceting continuous variables