Bayesian Averaging

Bayesian Averaging is a scoring method used by review platforms that blends a product's actual review ratings with a statistical baseline, preventing a small number of reviews from creating an extreme or misleading score.

Instead of calculating a simple average of all reviews, the platform adds a set of imaginary "baseline" reviews to every profile. As real reviews accumulate, the baseline's influence shrinks and the score increasingly reflects actual customer sentiment. The result: a new profile with one five-star review doesn't display a perfect 5.0, and a profile with one angry one-star review doesn't display a 1.0.

Trustpilot is the most transparent about its implementation — they document that every new business profile starts with seven phantom reviews at 3.5 stars, which they call the TrustScore's "starting point." But the method is widespread across rating systems.

IMDB uses a weighted Bayesian formula for its Top 250 rankings. Amazon factors review recency, helpfulness votes, and purchase verification into a weighted score rather than a straight average. Google is believed to use a Bayesian approach for local business ratings (a business with all five-star reviews may still show 4.8). Yelp, Algolia, and Netflix have all used variations of the concept.

The practical implication for SaaS companies: on any platform using Bayesian averaging, early reviews matter disproportionately. Each new review must outweigh the phantom baseline before your score starts reflecting your actual customer experience.

On Trustpilot specifically, your first seven real reviews are essentially competing with seven imaginary 3.5-star reviews — which is why starting early and building review volume consistently matters more than waiting for a "perfect moment" to launch your profile.

Example: A SaaS company claims its Trustpilot profile after receiving three organic five-star reviews. Instead of displaying 5.0, their TrustScore shows approximately 4.0 — because those three real reviews are averaged alongside seven phantom reviews at 3.5 stars. After 50 genuine reviews averaging 4.5 stars, the phantom reviews barely register and the TrustScore converges toward 4.5.

Understanding Bayesian averaging helps set realistic expectations for new profiles and informs review campaign planning. The "dig out" period every new profile goes through isn't a bug — it's the scoring system working as designed.