Section 3.3. Other Estimation Techniques

3.3. Other Estimation Techniques

Wideband Delphi is not the only technique that can be effective in estimating software tasks. Here are a few popular and effective alternatives.

3.3.1. PROBE

Proxy Based Estimating (PROBE), is the estimation method introduced by Watts Humphrey (of the Software Engineering Institute at Carnegie Mellon University) as part of the Personal Software Process (a discipline that helps individual software engineers monitor, test, and improve their own work). PROBE is based on the idea that if an engineer is building a component similar to one he built previously, then it will take about the same effort as it did in the past.

In the PROBE method, individual engineers use a database to keep track of the size and effort of all of the work that they do, developing a history of the effort they have put into their past projects, broken into individual components. Each component in the database is assigned a type ("calculation," "data," "logic," etc.) and a size (from "very small" to "very large"). When a new project must be estimated, it is broken down into tasks that correspond to these types and sizes. A formula based on linear regression is used to calculate the estimate for each task. Additional information on PROBE can be found in A Discipline for Software Engineering by Watts Humphrey (Addison Wesley, 1994).

3.3.2. COCOMO II

The Constructive Cost Model (COCOMO) is a software cost and schedule estimating method developed by Barry Boehm in the early 1980s. Boehm developed COCOMO empirically by running a study of 63 software development projects and statistically analyzing their results. COCOMO II was developed in the 1990s as an updated version for modern development life cycles, and it is based on a broader set of data. The COCOMO calculation incorporates 15 cost drivers, variables that must be provided as input for a model that is based on the results of those studied projects. These variables cover software, computer, personnel, and project attributes. The output of the model is a set of size and effort estimates that can be developed into a project schedule. Additional information on COCOMO can be found in Software Cost Estimation with Cocomo II by Barry Boehm et al. (Prentice Hall PTR, 2000).

3.3.3. The Planning Game

The Planning Game is the software project planning method from Extreme Programming (XP), a lightweight development methodology developed by Kent Beck in the 1990s at Chrysler. It is a method used to manage the negotiation between the engineering team ("Development") and the stakeholders ("Business"). It gains some emotional distance from the planning process by treating it as a game, where the playing pieces are "user stories" written on index cards and the goal is to assign value to stories and put them into production over time.

Unlike Delphi, PROBE, and COCOMO, the Planning Game does not require a documented description of the scope of the project to be estimated. Rather, it is a full planning process that combines estimation with identifying the scope of the project and the tasks required to complete the software.

Like much of XP, the planning process is highly iterative. The scope is established by having Development and Business work together to interactively write the stories. Then, each story is given an estimate of 1, 2, or 3 weeks. Stories that are larger than that are split up into multiple iterations. (Often, an organization will standardize on a regular story and iteration duration: for example, day-length stories and one- or two-week iterations.) Business is given an opportunity to steer the project between iterations. The estimates themselves are created by the programmers, based on the stories that are created. Finally, commitments are agreed upon. This is repeated until the next iteration of the project is planned.

Additional information on the Planning Game can be found in Extreme Programming Explained by Kent Beck (Addison Wesley, 2000).