Given a set of keyphrases, we analyze how Web queries with these phrases can be formed that, taken altogether, return a specified number of hits. The use case of this problem is a plagiarism detection system that searches the Web for potentially plagiarized passages in a given suspicious document. For the query formulation problem we develop a heuristic search strategy based on cooccurrence probabilities. Compared to the maximal termset strategy [3], which can be considered as the most sensible non-heuristic baseline, our expected savings are on average 50% when queries for 9 or 10 phrases are to be constructed.