Despite widespread recognition that the aggregate labour market is composed of a number of heterogeneous submarkets, there is little guidance on how to appropriately delineate such submarkets when conducting economic research. This paper contributes to a small but growing body of work addressing this issue by exploring the potential for community detection algorithms to delineate labour submarkets using observed patterns of labour market mobility. Two alternative approaches to community detection - modularity maximisation and stochastic block model estimation - are compared from a theoretical perspective and implemented on network data formed by worker transitions observed in the UK between 2011 and 2019. The theoretical comparison shows the two approaches implement very different definitions of labour submarkets, while the empirical application finds they also produce different submarket partitions in practice. This highlights that future research using community detection methods to delineate labour submarkets should ideally implement both approaches and examine whether any subsequent results are robust to the choice between them. Additional analysis looks at how occupational skill requirements change following worker transitions and how they vary within labour submarkets. This provides preliminary evidence that differences in manual skill requirements are a greater impediment to occupational changes that are made involuntarily than differences in non-manual skill requirements.