We describe an efficient algorithm for determining exactly the minimum number of sires consistent with the multi-locus genotypes of a mother and her progeny. We consider cases where a simple exhaustive search through all possible sets of sires is impossible in practice (because it would take too long to complete). Our algorithm for solving this combinatorial optimisation problem avoids visiting large parts of search space which would not improve the solution found so far (i.e., result in a solution with fewer number of sires). This is of particular importance when the number of allelic types in the progeny array is large and when the minimum number of sires is expected to be large. Precisely in such cases it is important to know the minimum number of sires: this number gives an exact bound on the most likely number of sires estimated by a random search algorithm in a parameter region where it may be difficult to determine whether it has converged. We apply our algorithm to data from the marine snail, Littorina saxatilis. ; Comment: 16 pages, 5 figures, 3 tables