Host manipulation by parasites not only captures the imagination but has important epidemiological implications. The conventional view is that parasites face a trade-off between the benefits of host manipulation and their costs to fitness-related traits, such as longevity and fecundity. However, this trade-off hypothesis remains to be tested. Dinocampus coccinellae is a common parasitic wasp of the spotted lady beetle Coleomegilla maculata. Females deposit a single egg in the haemocoel of the host, and during larval development the parasitoid feeds on host tissues. At the prepupal stage, the parasitoid egresses from its host by forcing its way through the coccinellid's abdominal segments and begins spinning a cocoon between the ladybird's legs. Remarkably, D. coccinellae does not kill its host during its development, an atypical feature for parasitoids. We first showed under laboratory conditions that parasitoid cocoons that were attended by a living and manipulated ladybird suffered less predation than did cocoons alone or cocoons under dead ladybirds. We then demonstrated that the length of the manipulation period is negatively correlated with parasitoid fecundity but not with longevity. In addition to documenting an original case of bodyguard manipulation, our study provides the first evidence of a cost required for manipulating host behaviour.