This report is concerned with designs for field intervention studies utilized in operations research. Beginning with an overview of the objectives of operations research, the authors go on to review the difficulties associated with using true experimental designs. Proposed guidelines for designing quasi-experimental studies are then discussed. The guidelines focus on using multiple sources of data, multiple measurements over time, and multiple replications of the study intervention. Finally, an example is presented of one ongoing operations research study in Sri Lanka. Since 1981, 14 research studies have been implemented in Thailand, the Philippines, Sri Lanka, and Nepal as part of an ongoing family operations research (OR) program in Asia. 2 features of true experimental designs often prove problematic for health and family planning field studies: random selection of study units (individuals, households, villages, districts) from some larger grouping of these units, and then subsequent random assignment of these units to experimental and control groups; and maintaining full control over timing, intensity and duration of the experimental variables. Random assignment of study units to experimental and control groups relates to internal validity and is important if the objective of a study is to make causal inferences about the impact of a program intervention. Managers are concerned with such questions as: what type and length of training should fieldworkers receive? What is the most cost-effective means of providing accurate information to large numbers of rural people about the availability of services? Do incentives increase contraceptive prevalence, and if so, what type of incentives? A guideline for selecting field research design is the principle of the 3 multiples: seeking multiple data sources to obtain information on the same variables; seeking multiple measurements over time of the same variables; and seeking multiple replications of the study intervention in different field settings. Experience suggests that a good quasi-experimental design includes: background information from 2ndry sources; selecting multiple sites for replication of the study intervention; collecting preliminary qualitative data; conducting a baseline survey; implementing a study monitoring system; conducting a follow-up survey; collecting postintervention qualitative data; and collecting longterm followup data. The Sri Lankan IUD study which followed this basic design presented many difficulties, but at least allowed the investigators to isolate statistically the data from areas known to have been affected by nonstudy intervention variables.