Although activation services such as monitoring, training, or job subsidies have been shown to increase exits from unemployment, there is little comprehensive evidence about the effects of activation during recessions. Here we evaluate a large activation programme introduced in Portugal in 2012, a time of very high unemployment. This programme required specific unemployment benefit recipients to meet jobcentre caseworkers and then participate in active labour market policies. Our analysis draws on rich longitudinal data, the programme's focus on those unemployed for at least six months, and fuzzy regression discontinuity methods. We find that, despite the weak labour market, the programme is very successful as it doubles the monthly reemployment probability. The results are robust to a number of checks, including a falsification exercise based on pre-programme data and an analysis of non-employment and income effects. Moreover, in a novel IV approach using information on all unemployed, we find no evidence of substitution effects such as decreased transitions to employment amongst non-eligible individuals.