Trust Center
Our principles
- Evidence first. Every score and highlighted statement rests on real reviews; you can see the review count and sources.
- No ranking for sale. Businesses cannot pay to change their score or position.
- We do not hide uncertainty. When data is thin we show "low confidence".
- AI is labelled. AI-assisted summaries and tags are clearly marked.
- The method is open. How the score is computed is published below; every change ships with a version and a note.
- We respect your privacy. Analytics runs only with your consent, and you can withdraw it at any time.
Display policy
Results show the score, a confidence level and reason chips. Ads and sponsored content never affect ranking. Venues without enough data show "Not enough data" instead of a score.
How to report content
For an inappropriate, misleading or wrong review, use the "Report" button next to it. We review reports and remove content where needed. For wrong venue details, use the feedback link on the venue page.
Score methodology
What is the score?
The Nesnel Score is a single number from 0 to 10 for a venue. It is not a star average. Stars behave very differently from one platform to another: five stars is routine on one and rare on another. A plain average also gives a venue with 8 reviews the same standing as one with 2,000. The Nesnel Score is designed to fix both problems.
What goes in
- Ratings normalized per platform. Each rating is judged against the distribution of ratings on the same platform in the same city. Five stars on a generous platform can count for less than five stars on a stricter one.
- Aspects from the review text. Besides the stars, what the review says about food, service, atmosphere and so on is taken into account. A review without text contributes its rating alone.
- Recency. Newer reviews count more than older ones; the influence of old reviews fades over time but never reaches zero.
- Review quality. Detailed, informative reviews carry somewhat more weight than short, empty ones.
- Source. Structured reviews written on Nesnel by an attributable author carry somewhat more weight than third-party reviews.
Shrinkage toward similar venues
A venue with few reviews has its score pulled toward the average of similar venues (same primary category) in the same city. As reviews accumulate, the pull fades and the score follows the venue's own reviews. This is why 8 perfect reviews cannot outrank 2,000 reviews averaging 8.5. If a category does not have enough venues, the city as a whole is used as the reference.
Confidence band
Every score is shown with a Low, Medium or High confidence band. The band says how much effective evidence stands behind the score (review count, recency and quality together). Low means "treat this number with caution".
Dine-in and delivery sub-scores
Separate sub-scores are computed for dining in and for delivery or takeaway, each from reviews of that kind only. A venue may not have enough reviews of one kind. In that case the sub-score is left empty; it is never guessed and never filled in from the overall score.
Aspect scores
Separate scores can be shown for aspects such as food taste or service. An aspect is shown only when enough reviews mention it.
Trend
The 3, 6 and 12-month trend values start from the venue's own overall score and are nudged by how much recent reviews differ from the venue's all-time average. The nudge is damped by the number of recent reviews: with few recent reviews the value stays close to the overall score. As a result, a steady venue reads flat.
This deliberately differs from simply re-scoring the window on its own. A narrow window scored alone would be pulled toward similar venues, which would falsely mark steady venues that are above average as "declining" (and below-average ones as "improving"). The direction is shown as up, down or flat, and only up or down when the difference is clearly noticeable.
What never goes in
Payments, ads, plans, owner requests and claim status never affect the score. The score is computed from reviews only.
Numbers
The exact weights and thresholds are not published, to make the score harder to game. The version number and the record of every change are published; changing any setting produces a new version.
Score changelog
Every change to services/scorer/params_v1.go needs a new score version, an entry here and a Trust Center note.
v1.0 — 2026-11-03
Initial version.