The study of low diluted powder mixtures poses problems due to the difficulty of accurately observing the evolution of the mixing process thus making it essential to have structural information of the mixture. A new systemic model is presented here, based on the definition of homogeneous regions, where the deviation from a pure random mixture is expressed in terms of the relation between standard deviation and sample size. A one-dimensional example is provided and applied to experimental work for the case of a 1% binary mixture. The numerical results together with the determination of the autocorrelation function leads to a complete identification of the mixture structure.