Software Estimation Techniques

Study Notes

Successful Project Characteristics: Delivering agreed functionality on time, at the agreed cost, with the required quality.
Project Stages: Setting achievable targets, attempting to meet those targets.
Parkinson's Law: Work expands to fill the available time. Over-estimating task time often results in project duration exceeding the ideal timeframe.
Weinberg's Zeroth Law of Reliability: A software project not needing a specific reliability requirement can fulfill other requirements.

Taxonomy of Estimating Methods:
- Bottom-up: activity-based, analytical approach. Breaks down the project into smaller components, estimating each element, then aggregating. Suitable when little similar past project data exists.
- Parametric/Algorithmic Models (e.g., COCOMO, function points): Use historical data and mathematical models to predict effort based on system characteristics. COCOMO uses lines of code for example, while function points consider system complexity aspects.
- Expert Opinion: Subjective estimations based on expertise, potentially unreliable without supporting data.
- Analogy: Comparing new projects to similar past projects.
- Parkinson and "Price to Win": Cost estimates influenced by factors beyond purely technical requirements, such as competitiveness.

Bottom-up: Detailed task breakdown for estimations. Time-consuming, but more precise when detailed project information is available.
Top-down: Overall estimate based on project cost drivers (historical data). Divides the overall estimate across tasks, suitable when similar past project data is accessible.

Calculate the overall project estimate using efficiency metrics or historical data.
Distribute proportions of this overall estimate among project components.

Data Requirements: Historical data is necessary for successful parametric model application for software effort estimation. Examples include lines of code, function points, etc.
System Characteristics: Parameters used for prediction (e.g., lines of code, input/output transaction types)

Origin: Developed by Charles R. Symons. Based on the work of Albrecht.
Use: For software sizing and estimating. Initially developed for use in the UK. Compatible with the system development methodology SSADMs.
Calculation: A formula combining the number of input, output, external interface, and file types for estimation of function point size.

Purpose: Extension of function points for estimation of embedded systems, which consider software components/units working in a layered structure.
Counting Method: Count components in a layered system, including component movement in and out of the layer.

Basis: Provides productivity standards based on industry data. Allows software productivity benchmark comparison between projects.
Effort Calculation: Uses the formula effort = c x sizek for effort calculation.
- Constants (c and k) vary based on project type (Organic, Semi-detached, Embedded, etc.).
- Size is generally measured in thousands of lines of code (kLOC).

Factors Influencing Productivity: Reliability, Database size, complexity, execution time constraints, storage limitations, team capability, and programming experience.
Application: Estimate adjustment by factoring in development complexity issues (multipliers).

Process: Examine past projects and evaluate the resemblance of current project characteristics. Identify an example that closely matches the current project and apply historical data to the current project.

Questions to Consider: Task size drivers, productivity rates utilized in the past, comparable prior project examples, accuracy and reliability of estimated productivity rates.