Scoring methodology
Evidence before narrative
Product Momentum describes near-term commercial evidence over roughly 30 to 90 days. Confidence and Competition are separate scores. None is a probability of business success.
Nominal weights
- Retail: 30
- Search and purchase intent: 25
- Distribution: 15
- Social or discovery: 10
- Supply chain: 10
- Commercialization: 10
Missing data is unknown
The weighted mean includes available components only. Coverage is the available nominal weight divided by 100.
adjusted = 50 + (raw - 50) × (0.65 + 0.35 × coverage)
No available input means no score. Source outages never become zero observations.
Publication and stages
Publication needs two source families, at least 55% coverage, fresh data, and search or retail. Accelerating also requires both search and retail, three families, confidence of at least 60, momentum of at least 72, and persistence of at least 0.60.
Normalization and seasonality
Features are normalized against comparable category, geography, and time-window observations using median absolute deviation. Robust z scores are clipped to -4 through 4 and mapped with a hyperbolic tangent. Seasonality requires 104 weekly observations; shorter histories are provisional.
Review and reproducibility
Only exact provider IDs and approved, compatible normalized aliases auto-match. All fuzzy and semantic suggestions require human review. Scores retain input provenance, a version, and a calculation run.
Validation
Historical replay uses only observations available at the evaluation date. Weekly prediction snapshots are immutable. No live historical performance claim is made for this sample site.