In my last entry, I introduced the notion of using conceptual models to estimate the cost and duration of a software project. In this entry, I outline a process for doing so.
- Compose a conceptual model of the system. Typically, most of the nouns mentioned in use cases appear in the conceptual model.
- Identify concept clusters. A concept cluster is a group of related concepts. Groups of concepts that have many associations among them are good candidates. Concept clusters roughly correspond to the components of the system assignable to different teams of developers.
- Count the number of concepts in each concept cluster. The number of concepts in a cluster is roughly proportional to the level of effort required to implement the corresponding component of the system.
- Assign a concept multiplier to each cluster. A concept multiplier represents the number of man-hours it will take to implement the functionality related to a single concept. In some cases, the multiplier for the concepts in one cluster will differ from the muliplier for concepts in other clusters. This step requires close consultation among product management, architects, and developers to determine the most realistic multiplier.
- Compute total man-hours. The total man-hours is equal to the summation of the products of the number of concepts in each cluster and the concept multiplier for that cluster.
After estimating the project using this object-oriented method, you can compare the results to estimates based on a functional estimation method.