This paper uncovers the reason why social networks are used in a job market. The data are novel: a nationally representative matched employer-employee data set in Bangladesh with detailed information, including direct measures of the use of social networks. The empirical analysis shows that compared with those who used open channels to find jobs, the employees who used social networks found jobs more easily, have lower observable abilities, and achieved lower employment outcomes conditional on observable and unobservable abilities. These results are robust whether firm-occupation fixed effects are controlled for or not. By comparing these findings with theoretical predictions, the paper concludes that social networks play the role as fast and easy but narrow-spectrum matching. That is, social networks allow job seekers to find jobs quickly and easily and thereby reduce search costs, but the types of jobs available from social networks are narrower than those from open channels. As a consequence, those who choose to use social networks are more likely to end up having mismatched jobs, that is jobs in which they cannot take advantage of their specialties. In the context of developing countries, a considerable number of poor job seekers may use social networks out of necessity even if the returns to finding good-match jobs through open channels are sufficiently high.