Cost Estimation

Produce engineering effort estimates: identify scope drivers and effort inductors, design team compositions, calculate risk-adjusted timelines, and build estimation models grounded in evidence.

Guiding Principle

"Estimation is not prediction — it is a range of probable outcomes. Express uncertainty honestly, identify assumptions explicitly, and always estimate in ranges, never single numbers."

Procedure

Step 1 — Scope Analysis & Decomposition

1. Decompose the project into estimable work packages (epics, features, tasks)
2. Classify each work package by type: new development, integration, migration, configuration
3. Identify scope drivers: number of entities, integrations, user roles, environments
4. Map complexity factors: technology novelty, regulatory requirements, performance constraints
5. Produce a Work Breakdown Structure (WBS) with classification per item

Step 2 — Effort Inductors

1. Identify effort inductors per work package: CRUD operations, API endpoints, data models, UI screens
2. Apply sizing model: story points, T-shirt sizes, or function point analysis
3. Calibrate estimates using historical data (velocity, similar past projects)
4. Factor in non-functional effort: testing (30-40%), CI/CD (10-15%), documentation (5-10%)
5. Produce a detailed effort breakdown with confidence ranges (optimistic, likely, pessimistic)

Step 3 — Team Composition

1. Identify required skill profiles: frontend, backend, data, DevOps, QA, architect
2. Design team topology based on project scope and domain complexity
3. Calculate team ramp-up time for new technology or domain
4. Factor in part-time vs. full-time allocation and context switching overhead
5. Produce a staffing plan with roles, FTE counts, and timeline per phase

Step 4 — Risk-Adjusted Timeline

1. Identify estimation risks: unknown requirements, third-party dependencies, technical uncertainty
2. Apply risk multipliers to affected work packages (1.2x-2x based on uncertainty)
3. Calculate three-point estimates: optimistic, most likely, pessimistic (PERT)
4. Build timeline with milestones, dependencies, and buffer allocation
5. Produce a confidence matrix: 50%, 75%, 90% probability timelines

Quality Criteria

• Every estimate has optimistic, likely, and pessimistic ranges (never single numbers)
• Non-functional effort (testing, CI/CD, docs) explicitly included, not hidden
• Assumptions documented and validated with stakeholders
• Historical calibration data referenced when available

Anti-Patterns

• Single-point estimates without ranges or confidence levels
• Padding estimates instead of explicitly modeling risk
• Ignoring non-functional effort (testing, deployment, documentation)
• Estimating without decomposition (gut-feel estimation on large scopes)