How factor analysis reveals the hidden dimensions inside a questionnaire—and the sample sizes and pitfalls behind honest results.
Factor analysis asks how many independent threads a bundle of items really measures. Exploratory (EFA) discovers structure; confirmatory (CFA) tests a claimed structure against data. Loadings show which items belong to which dimension.
Correlation matrix → extracted factors (PCA/EFA); fit indices (CFI, RMSEA) for CFA.
An 18-item engagement survey loads cleanly on three factors—voice, growth, energy—so the report scores three dimensions, not one average.
Before naming dimensions in any productized assessment, and whenever alpha is suspiciously low on a supposedly single scale.
EFA needs ~100+ respondents and 5 items per factor minimum; rotate and interpret theory, not eigenvalues alone.
EFA to discover, CFA to confirm—ideally on separate samples. Skipping EFA is guessing; skipping CFA is unproven.
No—PCA decomposes variance, EFA models latent structure. For questionnaire dimensions, EFA is the honest default.
Turn the concept into a live assessment：Factor Analysis
Browse templates →