Filbert estimating techniques can be improved by using refined procedures for selecting sample limbs and counting nut clusters. These procedures can reduce survey cost 25 percent and improve sampling and nonsampling errors. Counting nut clusters for two terminal limbs (4 percent of an average tree) by stripping them from limbs reduced counting errors considerably, compared with on-the-limb counts for primary limbs (15 percent of an average tree). The total cross-sectional area of primary limbs is inexpensive to obtain and can be used efficiently in a double sampling survey design.