I analyze the effects of household sociodemography, interviewer performance in the current survey, and fieldwork characteristics on cooperation in a central telephone survey, where households with no publicly listed landline number receive face-to-face visits. Using the 2013 refreshment sample of the Swiss Household Panel, I employ household-interviewer cross-classified multilevel models and analyze first and later contacts separately. Some sociodemographic groups are less cooperative in the first contact only, others in both the first and later contacts, and still others in later contacts only. I offer recommendations about which households should be finalized at the first contact, which should be transferred to the face-to-face sample instead of being worked by telephone, and which interviewers should work which household groups.