How a score is built.

1. 01

   Real places and sources stay real

   Names, public place details, press mentions, and other cited context are shown separately with links. A citation supports only the claim beside it and does not automatically affect a preview score.

2. 02

   Preview scores use synthetic scenarios

   The current score engine runs on deterministic sample scenarios, not consumer reviews. Sample tiers demonstrate how anonymous, registered, local, and expert inputs could carry different weights after a real verification system launches.

3. 03

   Evidence flags are demonstrations

   Receipt, check-in, photo, and reservation flags currently attached to sample scenarios are synthetic too. They demonstrate intended math; they do not claim a real person visited a venue.

4. 04

   Recency is part of the model

   The engine can reduce the influence of older inputs. In preview, those dates belong to synthetic scenarios and should be read as a product demonstration, not a current consumer signal.

5. 05

   Thin samples stay humble

   Small samples are pulled toward the category average, and each score shows a confidence label and effective sample size. Those labels describe model depth, not certainty about real-world quality or safety.

6. 06

   Then it is read through your taste

   The quiz, your ratings, and your tweaks re-weight the dimensions you care about. Personalization changes how the preview model is sorted; it does not make the underlying information objectively true.

7. 07

   Commercial influence stays separate

   Payment, free meals, gifts, favors, and business-submitted details do not change placement. Business information and owner picks are labeled as coming from the business, and material relationships must be disclosed.

The same preview scoring engine can run on any category — coffee, cocktails, neighborhoods, or whatever's next. Only the dimensions being weighed change; the weighting rules stay consistent and visible.