Distribution

Because there are at least nine definitions of a sample quartile, and charting tools rarely say which one they use. SmartVisuals Distribution defaults to the definition Excel and DAX use in PERCENTILE.INC and QUARTILE.INC, so the box agrees with the measures already in your model. Excel / DAX exclusive (PERCENTILE.EXC), Tukey's hinges and all nine Hyndman-Fan types are one setting away, and the footnote names the definition in use.

Whiskers follow the rule you choose: 1.5 x IQR with an adjustable multiple, percentiles, mean +/- SD, or minimum to maximum. Outliers and far outliers get different markers, drawn as hollow and filled triangles, so they read by shape and never by colour alone. Reference lines for the mean, the median and up to four percentiles use the same quartile method, so a P25 line sits exactly on the lower hinge.

Box plots summarise; density views show shape. A violin or half violin reveals two peaks a box would hide, a split violin puts two levels of a Split By field back to back, a raincloud combines the density, a box and the raw points in one row, and a ridgeline stacks many groups as overlapping curves that read like a mountain range. Densities use Scott, Silverman or the rule-of-thumb bandwidth, with a smoothness control.

The histogram offers Freedman-Diaconis, Sturges, Scott, Rice, square root and Doane rules, a fixed bin count or width, round bin edges, a bin start, or your own edges such as 0, 10, 12, 15, 20. Bars show count, percent or density, with an optional cumulative line, density curve and fitted normal curve. The beeswarm draws every observation, so hovering a point shows its ID and clicking it filters the report to it. All of this is free.

A gap between two medians on a chart may be noise. The premium layer answers the question the chart raises: pairwise Welch t or Mann-Whitney U tests, adjusted for multiple comparisons with Holm or Bonferroni, and an overall Welch's ANOVA, ANOVA or Kruskal-Wallis test. Results are drawn as brackets with stars, or as A / B / C letters when there are too many pairs to read, and the caption names the test and the p-value.

Premium also adds mean confidence intervals from the t distribution and notched boxes for comparing medians, plus a Weight field for survey and sampling weights: weighted quartiles, densities, means and Welch tests with the Kish effective sample size. Every statistic is golden-tested against independent reference implementations, and degenerate cases such as one observation or zero spread show n/a, never a false result.

Process capability is premium. Type one set of spec limits for the chart, or map LSL, USL and Target as measures from a spec table or product master, so every product or line is judged against its own limits and the limits follow slicers. The visual reports Cp, Cpk, Cpl, Cpu, Cpm, Pp and Ppk, the expected and observed share out of spec in ppm, and individuals-chart control limits.

Within-group sigma follows the AIAG moving-range method inside each group's own production sequence, and the footnote names the field that sets that order. An index is shown only when it is defined: too few observations, zero spread, crossed limits or an ambiguous limit give no index at all rather than a misleading number. Small multiples, one panel per value of a field on a shared axis, and advanced density with six kernels and bounded support complete the premium layer.