Builder Score
How we rate niches — Opportunity × Buildability × honest Risk.
How it all connects
Score, §1-12 analysis and sources form one closed system: every score axis rests on the same real data the analysis explains — you can jump from an axis straight to the matching § reasoning.
Example: demand axis 92 ← §2 (user perception) ← review volume + HN/Reddit signals. The score is never a black box.
How we rate
The score is a hybrid of 12 components — 8 additive axes and 2 multiplicative gates for fatal-flaw factors. We don't use iTunes stars and review counts as direct quality assessment, but as data input for other components (e.g. demand as Sweet-Spot, not linear).
8 additive axes
- 1 · Demand (Sweet-Spot)
- 2 · Momentum (Timing)
- 3 · Attackability (Incumbent Love)
- 4 · Wedge × Pain-Intensity
- 5 · Discoverability + Distribution
- 6 · Monetizability + Retention
- 7 · Switching-Costs (inverted)
- 8 · Trend Durability
2 multiplicative gates
- 9 · Buildability (AI-Builder × Required Features)
- 10 · Risk (Compliance/Platform)
Floor: no gate can push the score below 20 % of the additive core.
What you see
Every score carries a confidence tier badge. Instead of percentages, we show four tiers to make clear how robust the rating is:
- Strong Signal — ≥75 % confidence, data complete.
- Signal — 50–75 %, score plausible, please double-check.
- Trend — 25–50 %, limited reliability.
- Data-Sparse — <25 %, orientational only.
Axes with low coverage are shown dashed in the radar — missing data is visible, not hidden.
How we learn
The score is not an ML model. It is a versioned, rule-based hybrid model that evolves through four mechanisms:
- Data Growth: daily snapshots sharpen the Momentum component; larger corpus → better Sweet-Spot calibration.
- Version Activation: Score versions are manually activated. Old ratings show 'Update available'.
- Outcome Feedback (v3.1): after Pro-clicking a niche, optional 1-click feedback.
- Cross-Validation (v3.2): Score prediction vs. actual build outcome.
Intentionally not in v3.0: pure ML learning, auto-tuning of weights, LLM-in-score-loop. We want to protect deterministic explainability.
Which scores do we show where?
Two orthogonal scores: the Builder Score rates build substance (opportunity + buildability + honest risk); Market Snapshot is just a market snapshot — never a build assessment.
| Page | Primary | Secondary |
|---|---|---|
| Dashboard / niche overview | 🛠️ Builder Score | 📊 Market Snapshot |
| Niche detail | 🛠️ Builder Score | 📊 Market Snapshot |
| App detail | 🛠️ Builder Score (App) | ⭐ iTunes stars |
| App lists / Explorer | 🛠️ Builder Score | — |
Detail pages were already consistent; with this update the dashboard & lists follow the same standard.