Mixed mode surveys, i.e. a combination of various survey modes (e.g. face-to-face, web, mail, fixed telephone, mobile telephone), is increasingly used in probability based sample surveys due to the need of optimising the survey process from the cost and quality point of view. The cheap survey modes (like web surveys) usually face lower coverage and response rates and correspondingly also higher biases. On the other hand, higher coverage and response rates are usually correlated with more expensive survey modes (like personal interviewing). In addition, in mixed mode surveys the solicitation stage can be entirely separated from the data collection, e.g. we can use a mail letter to invite respondents to a web questionnaire. Large number of possible mode combinations may thus appear, especially if three, four or five waves of follow-ups are used. Consequently, the corresponding costs & errors model is a very complex one. In this paper, we first theoretically describe a general approach to the problem of balancing costs and errors in a mixed-mode survey environment. Then, we demonstrate initial results of a pilot study, where mean square error (i.e. the compound bias and sampling error) as a measure of data quality was observed at fixed survey costs for several mixed-mode research designs. Finally, we present the approach used to address this problem in the future research.