Effort estimation is a key factor for software project success, defined as delivering software of agreed quality and functionality within schedule and budget. Traditionally, effort estimation has been used for planning and tracking project resources. Effort estimation methods that grew upon those objectives focus on providing exact estimates and usually do not support systematic and reliable analysis of the causal effort dependencies. Moreover, existing estimation methods are typically based either on large data sets or on the extensive involvement of domain experts (human expertise), which, in practice, significantly reduces their applicability in software industry. In order to handle those problems the thesis proposes a WelCoMe method that that integrates data analysis and human judgment to extracting context-specific causal effort dependencies. When applied in the context of two industrial companies WelCoMe contributed to 17% reduction in cost of building an effort model and 50% reduction in complexity of a resulting model, while increasing its predictive performance by 43%-56%. Moreover, proposed estimation method allowed identifying crucial improvement potentials with respect to organizational measurement processes that had not been identified by domain experts.