This paper provides a Bayesian unified framework to analyze mixture models of one parameter distributions from a subclass of the exponential family. Firstly, the problem of choosing a suitable prior distribution on the component distributions is addressed. in a context where the number of components, is known. In order to do this, several experts` opinions are assumed as conjugate prior distributions. Then, a mixture model of these prior distributions is considered to represent a consensus among experts. The weights are fixed and they must be calculated. Next, the label-switching problem, is presented and a general algorithm to solve this difficulty for the considered family is provided. An illustrative example shows that the proposal techniques are easily applied in practice.