Quantification of human milk (HM) fat is important for determining the energy intake of infants. The simplest and most rapid method is the creamatocrit method. However, the reliability of the creamatocrit has not been comprehensively investigated. The aims of this study were to test the inter- and-intra-rater reliability of: 1) HM sampling after hand- or-machine mixing methods and 2) HM fat measurement by the creamatocrit method. Inter-and-intra rater HM sampling after hand- or-machine mixing methods had high intraclass correlation coefficient (>0.91). Inter-rater reliability of measurement of HM with low fat (<2%) resulted in high variability (median coefficient of variations (CVs) > 15%). Intra- and inter-rater reliability of measurement of HM with higher fat (>3.5%) had low variability (median CVs < 10%). As the greatest variation in the creamatocrit method occurred during the measurement of HM samples with low fat, duplicate readings are necessary to reduce discrepancies in every HM fat determination.