Facets or log scales usually replace the broken axis

What it is

The problem broken axes try to solve (an outlier bursting the scale) has cleaner solutions. A log scale compresses magnitude at the root: a tenfold value occupies just one more segment, majority and outlier share the frame under one uniform conversion — every distortion of the broken axis grows from "two scales stitched into one axis," and the log axis has no second scale at all. Facets or paired charts split the comparison into like-for-like panels: one full-range zero-baseline panel (outlier included), one zoomed panel for the majority's detail — each internally uniform, cross-panel comparison handled by shared references and labels. The two cover nearly every use case of the break, with no hidden "where is the gap" knowledge demanded of readers.

Why it happens

The log axis offers one consistent conversion throughout (equal ratios at equal distances): cross-magnitude distance comparisons hold on every segment, at the price of logarithmic literacy on the reader's side (the log group treats that risk separately). Faceting offers splitting the incompatible into two questions: the full-view panel answers "how large is the gap," the detail panel "how does the majority vary" — more honest than a break "pretending one chart answers both." A third road is clipping the outlier with a label (break mark on the bar top, true value annotated outside), which moves the break from the axis onto a single mark, shrinking the failure zone from the whole chart to one element — still a lesser option. For bars, the sturdiest combination is "zero-baseline full chart plus a small outlier-detail panel."

Where it stops holding

The log axis requires positive data where ratios make sense (counts, money, durations qualify; quantities with a meaningful zero like Celsius do not); with zeros or negatives it fails, and faceting takes over. Faceting costs doubled layout and indirect cross-panel comparison (shared-scale constraints live in the facet group). Both demand audience capability — but logarithmic literacy is a one-time investment assistable by labels, whereas the break's hidden knowledge is re-tripped by every new chart; the cost structures differ entirely.

Applying it

• When a break seems needed, try in order: positive ratio data first takes a log axis; impossible (zeros, negatives, interval scales) takes zero-baseline full chart plus a zoomed facet.
• Bar scenes always take the facet route (length encodings and breaks are mutually exclusive); point-and-line scenes prefer the log axis.
• Verification: give break and alternative versions to reader groups answering the same comparison questions; the alternative's accuracy should be no lower — if the break version is worse, the replacement costs nothing.

Related

• Same group: U3.02.1 A broken axis removes a value interval from the middle of the scale · U3.02.2 Across the break, equal distances no longer mean equal values · U3.02.3 Break markers are often too small to notice, equal to unmarked · U3.02.4 Length encodings crossing the break lose all comparability
• Nearby: U3.03.1 Log scales suit data spanning orders of magnitude · U2.15.1 Small multiples replace overlay with repeated like-for-like panels
• Search terms: broken axis alternative · log scale · facet · outlier handling