Trust Score methodology

The Trust Score is a transparent number from 0–100 computed by a deterministic engine — not by an AI, and not by raw star averages. It rewards verified ownership and penalises signals of manipulation.

What goes into it

• Verified weighting. Verified-purchase reviews count far more than anonymous community reviews.
• Bayesian shrinkage. Products with few reviews are pulled toward the global mean, so a single 5-star review can't mint a perfect score.
• Reviewer reputation. Reviews from established, helpful reviewers carry more weight than brand-new accounts.
• Fraud penalties. Bursts of similar reviews, duplicate devices, and other manipulation signals reduce the score.
• Recency & volume. Sustained, recent, consistent feedback is more trustworthy than a one-time spike.

The full engine ships in Milestone 4. Scores you see today come from the seeded sample dataset and a Bayesian preview so the pages are meaningful during development.