Dual frames of fixed and mobile phone numbers are an attractive alternative to traditional fixed phone surveys which suffer from under-coverage due to the growing share of mobile-only households. However, adding the mobile phone frame can bring some operational problems and raise costs. To make best advantage of this mixed mode design we try to determine the optimal mixture parameter of the two frames: what share of the mobile subsample is sufficient to reduce error without substantially raising costs? First, we divide the target population into five strata according to phone use domains determined by possession of availability rates of both phone services. Then, we develop an optimization procedure that minimizes the product of error and costs giving four analytical solutions. This is applied to a 2008 Flash Eurobarometer survey where dual frame sampling was used for eight countries. The minimized function is very gradual and the optimal parameter lies in a flat era from 30-70%. The procedure was performed for various variables and with different conditions. Contrary to expectations, the result does not depend on variables but only on strata weights and fixed to mobile cost ratio. Simple regression using these as depending variables gives excellent prediction of the equation solution.