Contemporary survey research encounters increasing problems of declining response rates and growing costs, leading to a serious deterioration of data quality. In this regards, web surveys hold the potential for significant cost reductions, but still suffer from sampling issues, solicitation problems and generally low response rates. Optimization of survey research with the integration of web surveys into probability samples is thus one of the biggest challenges of contemporary social research methodology. The paper addresses this problem by discussing the findings of an experimental study conducted on a probability sample of the general population in Slovenia. The sample was divided into several experimental groups, manipulating the most important factors that impact on readiness of respondents for participation: solicitation mode, survey mode, and provision of incentives. The study employed web, mail, face-to-face and telephone modes. Analysis of cost-effectiveness was performed using the Mean Square Error (MSE) concept with a careful consideration of relations between the quality of obtained data and the research costs. These results, along with the findings on interaction between different factors influencing response rates, constitute the foundation for the formation of a pilot survey optimization model based on Markov chains. The model enables identification of the most optimal selection of solicitation, reminders, time of the next contact and surveying mode for individual respondents. Finally, the paper outlines the future development possibilities of the model, including its integration into an online survey application and automatic suggestions of the optimal survey strategy at different steps of a survey project.