Home › How We Rank

How We Rank Vendors

RefurbVerify uses a Bayesian-adjusted, 3-component scoring model — the same statistical method used by Amazon, IMDb and Netflix to produce fair, sample-size-aware trust scores.

1. Overview

Every vendor receives a TrustScore from 0 to 100, computed from three components plus a volume bonus:

60
Review Aggregate
Bayesian-adjusted weighted average of Trustpilot (2×) and ProductReview.com.au (1×) star ratings, scaled to 50 points.
20
Refurb Transparency
Grade descriptions, battery health disclosure, warranty clarity, returns policy quality. Max 20 points.
20
Operations & Legitimacy
Australian ABN, physical AU operations, contact information, years in business. Max 20 points.
+ Volume Bonus (up to +10 pts) Vendors with higher verified review counts earn bonus points — rewarding sustained, long-term customer engagement. Formula: (vendor reviews ÷ total reviews) × 10.

The final formula is:

TrustScore = Review Aggregate (0–60)
               + Refurb Transparency (0–20)
               + Operations & Legitimacy (0–20)
               + Volume Bonus (0–10)
               = 0 to 100

Score Composition · Bayesian-adjusted · v1 · 9 June 2026

Review aggregate — Bayesian-adjusted TP (2×) + PR (1×), weighted avg → scaled to 50 pts50 pts
Refurb transparency — grading clarity, battery disclosure, warranty, returns20 pts
Operations & legitimacy — AU ABN, physical ops, contact, years active20 pts
Volume bonus — (vendor reviews ÷ 194K total) × 10 ptsup to +10
Bayesian adjustment (industry standard): each star rating is shrunk toward the market mean (3.5★) in proportion to its sample size. Small samples are pulled strongly toward the mean; large samples (e.g. a vendor with 97K+ reviews) are trusted at close to face value. Formula: (n × rating + 5,000 × 3.5) ÷ (n + 5,000). This is why a 4.7★ rating backed by 97K reviews is statistically more reliable than a higher raw rating from far fewer reviews — and why review volume drives ranking.

2. Bayesian Rating Adjustment

Raw star ratings are not used directly. Instead, each rating is Bayesian-adjusted to correct for sample size bias — the same technique used by Amazon product rankings, IMDb movie scores, and Netflix recommendations.

The problem with raw ratings

A vendor with 4.9★ from 12 reviews and a vendor with 4.7★ from 12,554 reviews cannot be fairly compared using raw stars alone. The small-sample rating is statistically unreliable — it could easily be 4.2★ or 5.0★ with more data. The large-sample rating is close to ground truth.

The Bayesian formula

Adjusted Rating = (n × raw_rating + K × prior) ÷ (n + K)

Where:
  n = number of reviews for this vendor
  raw_rating = observed star rating (e.g. 4.8)
  K = 5,000 (prior weight — equivalent "virtual" reviews)
  prior = 3.5★ (market mean — neutral starting point)
How to read this formula When a vendor has very few reviews (n ≪ K), the adjusted rating is pulled strongly toward 3.5★ — reflecting our uncertainty. When a vendor has many reviews (n ≫ K), the adjusted rating stays close to the observed rating — reflecting high confidence. K = 5,000 means we need roughly 5,000 reviews before we give full weight to the observed rating.

Worked example: OzMobiles vs Reebelo

This shows how Bayesian adjustment and the volume bonus interact — and why counting reviews from the right country matters:

Note on Reebelo's review count. Reebelo operates a single global Trustpilot profile (trustpilot.com/review/reebelo.com) covering Australia, New Zealand, Canada, the United States, the United Kingdom, Singapore, Malaysia and Hong Kong — there is no separate Australian profile. Its global total is roughly 94,000 reviews. We score Reebelo AU on the 12,554 Australian reviews verifiable through ProductReview.com.au, not the global figure, so it is measured on the same basis as every Australia-only vendor on this leaderboard.
Step OzMobiles Reebelo AU
Raw Trustpilot ★4.84.7
Review volume (Australian)~20,000~12,554
Bayesian TP adjustment
(n × rating + 5000 × 3.5) ÷ (n + 5000)		 (20,000×4.8 + 5,000×3.5) ÷ 25,000
= 4.54★		 (12,554×4.7 + 5,000×3.5) ÷ 17,554
= 4.36★
Raw ProductReview ★4.54.6
Bayesian PR adjustment4.30★4.29★
Weighted avg
(TP×2 + PR×1) ÷ 3		 (4.54×2 + 4.30) ÷ 3
= 4.46★		 (4.36×2 + 4.29) ÷ 3
= 4.33★
Review score (÷5 × 60)53.5 / 6052.0 / 60
Transparency + Ops18 + 18 = 3617 + 17 = 34
Volume bonus+1.0+0.6
Final TrustScore90.5 → 91 / 100 86.7 → 87 / 100

OzMobiles leads on every component: a higher raw Trustpilot rating, more Australian reviews, and stronger transparency and operations marks. Reebelo's larger global presence doesn't earn it points here, because reviews written by overseas customers say nothing about the Australian buying experience.

This example also shows the volume bonus doing very little work — 1.0 point against 0.6. That is by design: review volume already influences the score through Bayesian shrinkage, so the bonus is a modest tiebreaker rather than the deciding factor.

3. Component Detail

Review Aggregate (max 50 pts)

We combine Trustpilot and ProductReview.com.au star ratings using a weighted average, with Trustpilot carrying 2× the weight per review (reflecting its stronger fraud-detection and verification processes). The minimum sample threshold is 20 reviews per platform — smaller samples receive a warning flag.

Weighted avg = (Bayes_Trustpilot × 2 + Bayes_ProductReview × 1) ÷ 3
Review Score = (Weighted avg ÷ 5) × 60

If only one platform has data, that platform's Bayesian-adjusted score is used directly for all 50 points.

Minimum sample size: 20 reviews per platform. A platform rating backed by fewer than 20 reviews is excluded from the review aggregate entirely, and the remaining platform carries all 50 points. Bayesian shrinkage alone is not enough here — with a very small sample the rating tells us almost nothing, and including it adds noise rather than signal. Example: Back Market AU is rated 2.1★ on ProductReview.com.au, but from only 8 reviews, so its TrustScore is calculated from its Trustpilot rating (4.3★ from 556 Australian reviews) alone.

Refurb Transparency (max 20 pts)

We assess how clearly the vendor communicates refurbishment quality to buyers:

Grade descriptions (A/B/C or equivalent): 0–6 pts
Battery health disclosure: 0–4 pts
Warranty terms (length + coverage): 0–4 pts
Returns policy (timeframe + conditions): 0–4 pts
Refurb process transparency: 0–2 pts

Operations & Legitimacy (max 20 pts)

We verify that the vendor is a genuine, operating Australian business:

Australian ABN registered & active: 0–6 pts
Australian physical operations: 0–5 pts
Published contact (phone/email/address): 0–4 pts
Years in continuous operation: 0–3 pts
Active social & support channels: 0–2 pts

Volume Bonus (up to +10 pts)

Review volume is a signal of sustained customer engagement. Vendors with larger verified review counts receive a proportional bonus:

Volume Bonus = (vendor_review_count ÷ 194,000 total) × 10

Example: OzMobiles has 20,000 Australian reviews, so 20,000 ÷ 194,000 × 10 = +1.0 pt. Only reviews attributable to the Australian operation count — a vendor's global review total does not. The total review pool updates as new vendors are added.

4. Sentiment Bands

Final TrustScores are grouped into four tiers:

≥ 78
Top Rated
Exceptional review scores, strong transparency and verified AU operations. Highest confidence for buyers.
64–77
Rated
Good track record. Some limitations in review volume, transparency or operations relative to Top Rated tier.
< 64
Fair
Below-average review sentiment, limited transparency, or marketplace-model vendors where quality varies by seller.
—
Unrated
Insufficient review data on either platform to compute a reliable score. Listed in the directory for awareness.

5. Data Sources

Trustpilot
Primary review platform. Carries 2× per-review weighting due to stronger identity verification and anti-fraud systems. Minimum 20 reviews required for inclusion without warning.
ProductReview.com.au
Australian-native review platform. Carries 1× weighting. Provides local consumer sentiment specific to the AU market.
Australian Business Register
Used to verify Australian Business Numbers (ABN) for all assessed vendors. ABN status is a mandatory legitimacy check.
Vendor websites
Assessed directly for transparency components — grading descriptions, battery health disclosure, warranty and returns policies.

6. Important Notes

Whole-platform scores Some ratings reflect whole-platform scores — not just refurbished products. For example, Kogan, Big W, eBay and Amazon reviews cover all products sold on those platforms, not exclusively refurbished items. Treat scores for marketplace-model vendors with extra caution.
Point-in-time snapshot Review data is captured periodically and may not reflect the most recent reviews. Vendor policies (warranty, returns, grading) are checked at time of scoring and may have changed. Always verify current terms directly with the vendor before purchasing.
ℹ Information only — not advice RefurbVerify TrustScores are compiled for information purposes only and do not constitute purchasing advice, financial advice or endorsement of any vendor. See our full disclaimer & terms.
Methodology version: v6 (Bayesian-adjusted) · Introduced: May 2026
Key change from v5: Star ratings now Bayesian-adjusted using K=5,000 virtual reviews and a 3.5★ market prior, replacing the raw-rating approach. This corrects for sample-size bias and aligns with industry-standard methods used by Amazon, IMDb and Netflix.
Score recomputed: