name: npd-methodology description: > Reference documentation for the NPD validation methodology including scoring rubrics, dimension definitions, verdict thresholds, and report templates. Loaded by evaluator subagents and the orchestrator. Do not invoke directly — use /npd-validator:validate-product instead. disable-model-invocation: true
NPD Validation Methodology Reference
Scoring Framework — 15 Dimensions
| Evaluator | Dim 1 | Dim 2 | Dim 3 |
|---|---|---|---|
| Market Demand | Category Size & Growth | Demand Signals | Timing |
| Competitive Intel | White Space | Competitor Weakness | Barriers to Entry |
| Brand Fit | Brand Alignment | Portfolio Coherence | Audience Fit |
| Commercial Viability | Pricing Power | Unit Economics | Channel Fit |
| Consumer & Trends | Social Sentiment | Trend Momentum | Cultural Fit |
Each dimension scored 1–10 with confidence rating (HIGH/MEDIUM/LOW). Minimum 150 words reasoning per dimension with cited sources.
Composite Score Weights (default)
| Perspective | Weight |
|---|---|
| Market Demand | 25% |
| Competitive Intelligence | 20% |
| Brand Fit | 15% |
| Commercial Viability | 25% |
| Consumer & Trends | 15% |
Custom weights: If the user specifies different priorities (e.g., "brand fit matters more for us"), adjust weights accordingly. When custom weights are used: (1) document them in the report header alongside the composite score, (2) note the default weights for comparison, (3) recalculating with new weights does NOT require re-running evaluators — only the Consensus Director needs to re-synthesize.
Verdict Thresholds
| Composite | Verdict | Action |
|---|---|---|
| 7.5–10.0 | GO | Proceed to development |
| 5.5–7.4 | CONDITIONAL GO | Proceed if conditions met |
| 3.5–5.4 | REVISIT | Significant rework needed |
| 1.0–3.4 | NO-GO | Redirect resources |
Override Rules
- Fatal Flaw Override: Any dimension ≤2 after deliberation AND devil's kill shot ≥7 → downgrade one level
- Confidence Override: >50% dimensions LOW confidence → append "(LOW CONFIDENCE)" to verdict
- Devil's Advocate Override: Kill shot ≥8 AND composite in CONDITIONAL range → downgrade to REVISIT
Stacking: Overrides do NOT compound. Apply all against the original mechanical verdict, take the most conservative result. Max one level downgrade from overrides.
Research Depth Standards
- Research Coordinator: 8–12 web searches
- Each Evaluator: 3–5 additional web searches minimum
- Devil's Advocate: 3–5 adversarial searches
- Total per validation: 25–40+ searches
- Cross-reference claims across multiple sources
- Cite everything with URLs
- Mark LOW confidence when evidence is thin
Concept Brief Template
PRODUCT CONCEPT BRIEF
═════════════════════
Product Name / Working Title: [...]
Category: [...]
One-Line Description: [≤15 words]
Target Consumer: [...]
Price Point Range: [...]
Key Differentiator: [...]
Launch Horizon: [now / 6 months / 12 months / 24+ months — specify target month if seasonal]
Brand Context (optional): [...]
Launch Horizon is critical for the Timing dimension. A trend that looks perfect to enter "now" may have peaked 18 months before a late 2027 launch. Always capture this — if the user doesn't specify, ask before running the pipeline.
B2B vs B2C Products
The methodology applies to both, but B2B products require evaluator adaptation. Detect B2B when the Target Consumer is another business (manufacturers, retailers, distributors, professional users like salons/dentists/labs) rather than end consumers.
When validating B2B products:
- Market Demand — size the B2B buyer market (how many potential customers exist), not the end-consumer market
- Competitive Intel — analyze competing suppliers/vendors, not consumer brands. Pricing is typically per-kg/per-MT/per-seat, not retail
- Brand Fit — B2B credibility comes from industry certifications, trade show presence, partnerships, case studies, and white papers — NOT consumer brand alignment
- Commercial Viability — B2B has long sales cycles, volume-based pricing, contract terms (MOQ, payment terms, exclusivity), channel partner margins
- Consumer & Trends — "social sentiment" for B2B lives in trade publications, LinkedIn, industry conferences (Cosmoprof, in-cosmetics for beauty B2B), formulator forums — NOT TikTok or Reddit consumer threads
The scoring rubrics and dimensions are the same; the RESEARCH TARGETS shift. If the brief doesn't make B2B vs B2C clear, ask before launching the pipeline.
Evaluation Output Template
Each evaluator writes their output in this format:
# [Perspective] Evaluation
Product: [Name]
Date: [Date]
## Scores Summary
| Dimension | Score | Confidence |
|---|---|---|
| [Dim 1] | [X]/10 | [H/M/L] |
| [Dim 2] | [X]/10 | [H/M/L] |
| [Dim 3] | [X]/10 | [H/M/L] |
| **Subtotal** | **[avg]/10** | |
## Detailed Assessment
### [Dim 1]: [X]/10 [Confidence]
[150+ words with evidence and sources]
### [Dim 2]: [X]/10 [Confidence]
[150+ words with evidence and sources]
### [Dim 3]: [X]/10 [Confidence]
[150+ words with evidence and sources]
## Key Findings
- [Top 3 findings]
## Risks & Uncertainties
- [What could change this assessment]
## Sources
- [Source](URL)
For detailed evaluator instructions, see the individual agent prompt files in
skills/npd-methodology/agents/.