BreakCheck
The Trust Score is the 0–100 composite signal you see on every seller profile, search result, and leaderboard. It is not a star rating. It is a weighted formula that combines reviews, certification, verification, response behaviour, and scam-alert history into one number — recomputed every night.
The Trust Score is on a 0–100 scale. The thresholds are calibrated so that a seller with a clean record, multiple verified reviews, certification, and no scam alerts sits comfortably in the upper brackets — and one or two negative signals visibly pulls a seller down.
Sustained 5★ reviews, recent verification, certified status, no scam alerts.
Strong overall score with mostly verified buyers and a positive recent trend.
Usable with caution — gaps in verification, mixed reviews, or unpaid penalties.
Multiple scam alerts, very few reviews, or no certification. Approach with caution.
Every Trust Score is built from these, in order. Each section links to the exact line in the algorithm code so you can verify the formula by reading the source.
base_raw = weighted_avg_rating × ln(review_count + 1)
Each rating is multiplied by 2× if it is <90 days old and by 1.5× if it carries verified-purchase proof. This means a single verified review from last week counts more than three unverified reviews from two years ago.
ln(review_count + 1)
A first review contributes a lot; the second contributes less; the fiftieth contributes a fraction. Going from 5 to 50 reviews moves the score much more than going from 500 to 550. This prevents one prolific seller from dominating purely on volume, and prevents a new seller from looking unreviewed forever.
+10 if certified
Sellers who passed the BreakCheck vetting process — identity verified, ethics quiz passed, no open scam alerts — get a flat +10.
+2 per handle, cap +8
Each platform handle BreakCheck has spot-verified — Whatnot, TikTok, YouTube, eBay — adds +2, up to +8 total. This catches sellers who are real but new to the BreakCheck directory; cross-platform identity alone is meaningful signal.
+5 if >50% of 90d reviews have a reply
If the seller replied to more than half of their published reviews published in the last 90 days, they earn +5. Sellers who respond publicly demonstrate they take post-purchase accountability seriously.
-10 per verified open alert, cap -30
Each verified, still-open scam alert in the BreakCheck database subtracts 10 points, capped at -30. The scam alert penalty lifts on the next nightly compute once an alert is dismissed or resolved. This is asymmetric because one verified complaint carries more evidence than a single 1-star review.
up to +5, only when ≥25 confirmed spots
For sellers with at least 25 buyer-reported spots in the BreakCheck hit rate database, the bonus is hit_rate_pct × 0.05, capped at 5. Below 25 spots, the bonus is 0 (we require enough data to be statistically meaningful).
final = clamp((base_raw / raw_max) × 100 + bonuses − penalties, 0, 100) with raw_max = 12
Raw base is normalised to 0–100 using a soft anchor of 12 (raw_max) as the "100" target — tuned so a seller with 5★ average rating, 50 reviews, certification, 4 verified handles, and a 60% response rate lands at roughly 90. The normalised base + all flat bonuses − scam penalties are then clamped to [0, 100].
The 30-day trend (displayed on seller profiles) is recomputed each night from the trust_score_history table, which snapshots every active seller's Trust Score plus underlying metrics (rating average, review count, verified review count, response rate, 30-day scam alert count, and certification status) every night. The trend chip on each profile page reads two snapshots — one from "today" and one from ~30 days earlier — and reports the signed delta. The 90-day chart on /seller/:slug/history plots the same table.