What are the assumptions of K-Means clustering?

Updated May 16, 2026

Short answer

K-Means assumes spherical clusters, equal variance, and similar cluster sizes.

Deep explanation

It assumes clusters are convex and isotropic in feature space. It also assumes features contribute equally after scaling. Violations lead to poor clustering results.

Real-world example

Works well for well-separated customer segments.

Common mistakes

  • Applying K-Means on non-spherical clusters like moons or spirals.

Follow-up questions

  • What happens if clusters are non-spherical?
  • Why scaling is important?

More K-Means Clustering interview questions

View all →