Software cost and effort estimation models mostly focus on predicting the overall project cost rather than the cost of phases separately. In addition to that, no previous research has a domain specific view on phase by phase effort estimation. In this research, domain specific data analysis is conducted to discover the difference in phase distribution profiles in different application domains. Furthermore, a regression model is built in order to see whether a domain specific approach provides improvement in effort estimation accuracy. Our experimental results showed when domain specific data and phase by phase approach is used the prediction accuracy increases up to 74%.