This paper estimates a model of two-sided search using micro-level data for a well-defined labour market. It examines the assumption of random matching and contrasts it with the stock-flow (or non-random) matching model of Coles and collaborators. Given a dataset of contacts, matches, and complete labour-market histories for both sides of the market, we estimate hazard functions for both (unemployed) job-seekers and vacancies. For job-seekers, the tests adds the stock of new vacancies to a standard job-seeker hazard which itself depends on the stocks of vacancies and unemployed. Our tentative results find very weak evidence of stock-flow matching.