Our commitment to transparency
The GuardScore is an informational trust signal based on GuardOpinion review activity, recency, review volume, verification signals, response activity and moderation patterns. It is not a guarantee, recommendation or substitute for reading individual reviews.
The 3 pillars of the GuardScore
Three factors combine to produce a score that reflects reality — not marketing budgets.
Recency
Recent reviews carry more weight than older ones. A company that improved two years ago but declined recently will have a score that reflects the current reality.
Verification signals
Reviews with stronger experience signals may carry more weight. Reviews can be published without a verified label when they meet community rules, and suspicious activity can reduce confidence.
Volume and stability
We use a Bayesian average that prevents a single review from distorting the score. The more reviews a business has, the more stable and reliable its score becomes.
Why we use Bayesian average
A simple average can be deceived with 1 or 2 fake reviews. The Bayesian average prevents this.
| Business | Reviews | Simple avg. | GuardScore |
|---|---|---|---|
| New business | 2 reviews | 5.0 | 3.8 |
| Established business | 240 reviews | 4.1 | 4.1 |
The Bayesian average pulls new businesses toward the platform average until they have enough reviews to establish a reliable score.
What does NOT affect the GuardScore
These factors have zero influence on the score, regardless of the business's plan or relationship with GuardOpinion.
- The business's subscription plan (Free, Pro, Enterprise)
- The source of the review invitation (email, widget, QR code)
- Official replies from the business to reviews
- The number of times a profile is visited
- Whether the business has claimed or verified its profile
Combating fake reviews
Three independent layers help protect the integrity of reviews and reduce the risk of fake, manipulated or abusive activity.
Technology
AI that analyzes IP address, device, behavioral patterns, and writing style in real time on every review submitted.
Community
Any user can report a suspicious review. Reports are acted on within 24 hours by the integrity team.
People
Human moderators review complex cases. No algorithm makes the final decision alone — people do.