tomevault-io MVP & Roadmap Planning
Patterns for defining what to build first and in what order.
Core Principle
"Build the smallest thing that proves value."
An MVP is not a half-built product—it's a complete vertical slice that validates assumptions.
MoSCoW Prioritization
| Priority | Definition | Criteria |
|---|---|---|
| Must Have | System doesn't function without | Core journey incomplete, no workarounds |
| Should Have | Important but not blocking | Workarounds exist, high value |
| Could Have | Nice to have | Enhances experience, low priority |
| Won't Have | Explicitly out of scope | Prevents scope creep, document for later |
MVP Scoping Checklist
- Single complete user journey end-to-end
- Validates core assumption/hypothesis
- Deployable and demonstrable
- Measurable success criteria defined
- No features without corresponding tests
Phased Delivery Pattern
Each phase should:
- Build on previous phase (not parallel development)
- Be independently deployable
- Have clear success criteria
- Include tests for new functionality
Dependency Ordering
Order features by:
| Factor | Question |
|---|---|
| Technical | What must exist first? |
| Value | What provides most value soonest? |
| Risk | What validates riskiest assumptions? |
| Learning | What teaches us most about the domain? |
Roadmap Template
| Phase | Features | Success Criteria | Dependencies |
|---|---|---|---|
| MVP | [Must-haves] | [Measurable outcomes] | None |
| Phase 2 | [Should-haves] | [Measurable outcomes] | MVP complete |
| Phase 3 | [Could-haves] | [Measurable outcomes] | Phase 2 complete |
Extensible Algorithm Design
Design algorithms for evolution using stable interfaces. MVP delivers value immediately while building ground truth for future ML phases.
Evolution path:
- MVP: Deterministic algorithm (keyword matching, rule-based)
- Phase 2: Statistical approach (TF-IDF, collaborative filtering)
- Phase 3: ML/embeddings (neural networks, transformers)
- Phase 4: LLM-powered (if needed)
Interface pattern:
class Categorizer(Protocol):
def categorize(self, text: str) -> tuple[str, float]:
"""Returns (category, confidence_score)"""
...
Key principles:
- MVP uses zero ML infrastructure (no model serving, no embeddings DB)
- Interface stays stable across phases (same input/output signature)
- Each phase is independently measurable (track accuracy improvement)
- Switch implementations via dependency injection, not rewrite
See reference.md for detailed patterns and examples.md for sample roadmaps.
Converted and distributed by TomeVault — claim your Tome and manage your conversions.