Stop guessing how the bed will feel until check-in.

Community right now

· Last updated: 09/18/2026, 02:04 PM JST

Why only softness?

Existing hotel reviews are flooded with "clean", "friendly staff", "great location" — but mention the bed surprisingly little. Yet for business travellers, back-pain sufferers, and sleep-conscious tourists, that single night shapes the whole trip.

Softness is subjective, but the weight × perception correlation is clear. By prioritising same-weight reviews, Plush turns subjectivity into data.

How scores work

• 1–10 scale · 1 = board-firm / 10 = cloud-soft
• Aggregated by weight · Median per weight band (under 50 / 50–70 / 70–90 / over 90 kg)
• Verdict label · Cloud / Plush / Balanced / Firm / Board — derived from the score
• One vote per user · Re-posts to the same hotel are de-duped — only the latest counts
• Instant aggregation · A Postgres trigger recomputes stats the moment a review lands

Data sources

Plush combines multiple sources for hotels and scores. We make provenance explicit so you know which numbers to trust.

About AI-inferred scores

To bootstrap coverage in early stages, Plush uses Claude (Anthropic's AI model) to extract hardness mentions from existing Google reviews and generate a supplementary score.

• AI scores are aggregated at half weight (0.5×) of user reviews
• The review surface labels them "Google review" so you can tell them apart from real users
• As real reviews accumulate, the relative impact of AI scores naturally drops

Reviewer trust

Even anonymous, each review can be voted "helpful / not helpful". Reviewer trust is computed from those votes and feeds back into aggregation weight and review ordering.

• High-trust reviewer · Up to 4× weight in median calculations, ranked higher in hotel detail
• Low-trust reviewer · Weight clipped down to 1× (same as AI-inferred), reducing impact
• No self-voting · Row-Level Security blocks votes on your own reviews
• Names are never shown — weight depends on "are past contributions useful", not on "who"

Why anonymous?

Bed softness is deeply personal — body, condition, taste all overlap. Posting under your real name is essentially "announcing your weight". Plush aggregates per-weight medians so the "who" disappears, leaving only "how it feels at this weight".

• Display name optional, no real name needed · Post with a nickname like "Traveller A"; change or delete anytime
• Weight as range only · Just one of 4 bands — Plush never receives a specific number
• Reviews shown as "Guest" · No traceable IDs or emails are exposed
• Hotels can't identify you either — trust badges only reflect prior helpful votes

Privacy & anonymity

Weight is stored as ranges only (4 bands), never specific numbers. Display name is your choice. Sign-in goes through Google or other OAuth, so passwords never reach Plush. Analytics is cookie-free Plausible.

See the privacy policy.

About the alpha

Plush is currently in alpha. Coverage, hotels and features are limited, and there's plenty of room to improve. Bug reports and feature requests welcome at support@plushrate.com.

What's next

Direction post-alpha. Order and granularity depend on user voice and operator capacity. "I want this first" requests have the most leverage.