In the German unemployment insurance system, Integration Agreements (IA) are mandatory contracts between the employment agency and the unemployed, jointly signed by the latter and the caseworker. IAs stipulate rights and obligations but are generally perceived as instruments to control search behavior. We designed and implemented a Randomized Controlled Trial involving thousands of newly unemployed workers, where we randomize the timing of the IA as well as the extent to which this timing is announced prior to the meeting. Randomization is at the individual level. We use administrative registers to observe outcomes. A theoretical analysis of anticipation of prior announcements provides suggestions to empirically detect this. The results show that IAs early in the spell have on average a small positive effect on entering employment within a year. When classifying individuals using an employability indicator, we find that this result is driven by individuals with adverse prospects. Among them, being assigned to an early IA increases the probability of re-employment within a year from 45% to 53%.