Kruskal‑Wallis

The Kruskal‑Wallis test extends the logic of the Mann‑Whitney test to situations where you need to compare three or more independent groups. It is the non‑parametric counterpart to One‑Way ANOVA and is particularly useful when your data violates the assumptions of normality or equal variances. In many real-world processes—especially those involving cycle times, satisfaction ratings, or defect counts—these assumptions are frequently broken. Kruskal‑Wallis provides a reliable way to evaluate group differences without relying on parametric assumptions. 

Like Mann‑Whitney, Kruskal‑Wallis works by ranking all observations across all groups and examining how those ranks are distributed. If the groups come from the same population, their rank distributions should be similar. If one or more groups differ, the rank distributions will diverge, producing a significant test statistic. 

The test answers a simple but important question: Do at least one of the groups differ from the others? It does not identify which groups differ—that requires post‑hoc tests such as Dunn’s test or pairwise Mann‑Whitney tests with appropriate adjustments for multiple comparisons. 

Kruskal‑Wallis is especially valuable when dealing with ordinal data or when outliers distort the mean. Because the test relies on ranks rather than raw values, it is resistant to extreme values and robust across a wide range of distribution shapes. 

However, the test does assume that the distributions have similar shapes. If one group is more skewed than another, the test may detect differences in shape rather than differences in central tendency. This is why visual tools—boxplots, histograms, and distribution overlays—are essential companions to the test.

In the Analyze phase, Kruskal‑Wallis is ideal for comparing performance across multiple shifts, machines, suppliers, or product types when the data is non‑normal. It provides a structured, statistically sound way to identify whether differences exist, guiding your investigation toward the most influential factors. When used well, it helps you navigate complex, messy data with clarity and confidence.