Few studies exist on job duration in developing labour markets - an important omission both in our understanding of such markets and for the job duration literature, which is mainly based on developed-country case studies, which differ in structural ways. The main reason for this is likely data constraints in developing countries, since job duration analysis has intensive data requirements. Recently, two data sets meeting these requirements became publicly available in South Africa, covering the apartheid and post-apartheid eras. We use these data to provide a broad baseline about job duration in South Africa using survival analysis techniques with three main aims. First, we investigate to what extent stylized facts from the rest of the literature apply to South Africa; second, we analyse trajectories through the labour market; and third, we home in on early-career trajectories. South Africa broadly adheres to stylized facts about job duration: long-term tenure is a common feature of the labour market and the job hazard is non-monotonic and declines with tenure. Trajectories through the South African labour market, though, deviate from the developed-country case and we link this to the importance of labour market segmentation in South Africa. The influence of apartheid-era labour market policy on restricting the freedom of certain groups is clear in the results from this era, and interpretation of results from this time must take this historical context into account or risk reaching seriously misleading conclusions.