A frozen deterministic classifier
Our methodology does not re-compute segments with each study. The classifier is calibrated once on a large reference base and then remains fixed. Each new respondent is assigned a profile based on pre-defined reference centroids (a nearest-centroid procedure), without recalibration. The result is consistent: the same respondent receives the same profile in every study, every market and every subsequent wave. Comparison across waves, markets and years is methodologically clean.