Practical requirements for the non-invasive measurement of the composition by weight of fat in boneless meat are discussed including taking into account the effects of variation in product thickness, and illustrating the importance of the dual beam approach in making measurements for production-line meat boxes. Insensitivity to the fat distribution and product temperature, and the agreement between the theoretical and experimental values of the statistical spread are shown. A comparison is made of measurement using an 241Am-Be and a 252Cf source.