Quality signal (60%)
Rating is normalized and Bayesian-weighted so low-review listings cannot outrank stronger products on stars alone.
Public Methodology
This page documents how we collect signals, score products, review outcomes, and maintain editorial quality over time.
Current model weights
Quality signal0.60
Review confidence0.30
Value score0.10
Scoring Framework
Review confidence (30%)
Review depth is normalized against category peers to reflect confidence in marketplace signal strength.
Value score (10%)
Price is evaluated relative to category range and balanced with quality to avoid over-rewarding the cheapest option.
Testing Protocol
- Comparative testing across multiple hair routines and use cases
- Emphasis on repeatability, heat control consistency, and user-fit
- Structured pros/cons synthesis for practical purchase decisions
Data Inputs
- Public listing-level signals (rating, review count, and pricing context)
- Category-level normalization for fair cross-product comparison
- Product metadata enrichment for clearer on-page comparison output
Update Cadence
- Ranking dataset refreshed weekly where feasible
- Product pages reviewed every 6-12 months or earlier when signals shift
- Metadata and schema refreshed when material content changes occur
Editorial Safeguards
- No paid placement in ranking order
- Affiliate economics are separated from scoring logic
- Tradeoffs are documented even for top-ranked products
Corrections Policy
If we identify material errors in claims, pricing context, or product attributes, we update affected pages and adjust rankings when required.
Correction Requests
To request a correction, include page URL, the specific statement, and supporting evidence.