Speak Low is in Shanghai, China, with a restaurant identity at the center of the experience.

Our model reads Speak Low as a tentative 1 Star call, not because of one headline metric but because the guide-level and tier-level signals point in the same direction. The model does not see enough guide-level separation for a confident starred call. The question becomes ceiling: the model sees enough strength for a star, but not enough separation to publish a higher-star call. Because this is an upgrade call, the model needs more than general quality; it needs the higher tier to pull away from the alternatives. Because the starred signal sits close to a boundary, we mark the prediction as low confidence.