Current country-based group crop insurance, i.e., Group Risk Plan (GRP), is not an effective risk-reducing tool in counties where natural conditions are different across the area. Using only the historical yield information, a statistical approach is developed to group farmers by their yield similarity rather than linking them based on their association with a particular county. The cases of Washington State wheat farms and Iowa corn farms are the focus of this investigation. Sub-county or cross-county zones (clusters) are identified, and each farm is classified into a cluster where individual farm identification remains unknown. To improve risk-management and cost effectiveness of the crop insurance instrument, we propose implementation of zone-based GRP as a substitute for county-based GRP, where homogeneous zones rather than county boundaries are used for indemnifying yield.