For each manufacturer, exact cost estimation is both a major priority and a challenge. This routine task is far from being optimized and depends on a very large number of parameters both strategic and technical. To help estimate a cost, applications have been developed but estimation is a task that would greatly benefit from the re-use of knowledge and data found in alphanumeric and geometric documents. Every manufacturing company has switched to numerical data with CAD, CAM, or ERP systems but one of the main drawbacks is the low usage of all these information that contain valuable knowledge and expertise. This paper describes the current state of cost estimation and proposes a new hybrid approach whose purpose is to maximize the re-use of information for machined parts. Our approach is based on a parameterized and customized cost model, an extractor of semantic descriptors in geometric documents (Model Based Definition files) and related textual documents and finally correlations to adjust the cost of machined parts.