Multiple imputation is increasingly becoming a standard method for analyzing data with missing values. One of the main reason is the development of software and powerful computing environment. Given that a real time imputation is possible, the multiple imputation framework can also be used to leverage the existing data sources, recently collected data on respondents to identify targets for the next stage of the data collection. The proposal is to use the replicated or interpenetrating sampling framework, a large benchmark data source and multiple imputations of a key set of variables for the nonsampled subjects to develop the sampling strategies for the subsequent replicates. The goal is to obtain valid inferences from completed data sets consisting of the observed data from final set of respondents and multiple imputations of the nonsampled subjects. The benchmark data is used to assess the validity of the inferences. The approach is illustrated and evaluated using simulated and actual data sets. All three missing data mechanisms, MCAR, MAR and MNAR are considered.