Snowball sampling is frequently advocated and employed by qualitative social researchers. Under certain circumstances, however, it is prone to faltering and even failure. Drawing on two research projects where the snowball failed to roll, the paper identifies reasons for this stasis. It goes on to argue that there are alternative forms of networking that can be developed by the qualitative social researcher in lieu of snowballing. Specifically, when research momentum fails to build, rather than drilling down 'vertically' through social networks, we argue that the researcher can move 'horizontally' across social networks and cast the sampling and recruitment net wide and shallow rather than deep. This change in emphasis can, we argue, make the difference between a project failing and a project succeeding, and points to the importance of a variegated understanding of the social networks on which our social research depends.