The use of professional connections in job search is commonplace, however, endogeneous sorting into networks makes it difficult to assess the effect of connections on labour market outcomes.  Using Swedish administrative data, I treat the unexpected death of past coworkers as an exogeneous shock to individuals’ job search networks.  Following such an event, an individual’s annual earnings decline by around 2000 crowns on average relative to a control group, or around 1 per cent of pre-treatment earnings.  Employment rates also decline and job finding declines for individuals who were unemployed pre-treatment. I then use generalised random forests to explore which covariates are associated with larger treatment effects and find substantial heterogeneity in the value of a connection.  The interquartile range of conditional treatment effects is 3700 crowns and treatment effects are positive for around a quarter of the sample.