Principal component analysis replaces correlated inputs with orthogonal directions of variation. The largest direction comes first. You will project observations, measure retained variance and reconstruct the inputs to inspect what the reduced representation lost. A component can change sign without changing the geometry, so a correct check must allow that harmless difference. PCA also depends on measurement units: a feature measured in thousands can dominate one measured in fractions. Scaling is a modeling decision, not a formatting step. Explained variance measures this representation of the input table; it doesn't establish that the components will predict a business outcome.