The HVAC Bench Score

Three kinds of evidence, kept apart

Most equipment ratings online blend two things that should never be mixed: what a manufacturer says a machine does, and what the people who bought it found. The first is documentation and the second is testimony, and treating them as one number is how a specification sheet ends up looking like a reliability record.

1. Manufacturer facts. Model scope, features, and documented behaviour, taken from service and operation manuals. These establish what a system is built to do. They are never used to score how it was found in service, because a manufacturer is not a witness to that.
2. Owner experience. First-hand accounts from identifiable owners, documented field experience, and independent testing. Each observation is stored with where it was read, when it was collected, which model family it concerns, and what it actually said. This is the only class that can move a score.
3. The calculated score. Derived from the observations above and the published weights for that equipment class. There is no field anywhere in the system that lets anyone type a score in by hand.

When a family gets a number

A score publishes only when the evidence behind it clears every one of these. A near miss is still a miss, and the card says so instead of showing a hedged figure.

• At least 12 distinct owner observations about that family, after duplicates are removed.
• Observations from at least 3 independent sources, so one forum thread or one retailer page cannot carry a rating.
• At least 5 observations from 2 sources before any single dimension is scored. Dimensions nobody spoke to stay blank rather than being filled in.
• Enough scored dimensions to cover at least 50% of what matters for that equipment class.

Rules the system enforces on itself

• An observation about one family is not evidence about another, even within a brand. A brand-wide impression never substitutes for a specific family rating.
• Syndicated reviews are collapsed to one observation. The same retailer text appears across several sites, and counting it repeatedly would manufacture a sample.
• An observation votes only on the dimensions it actually addresses. Somebody complaining about noise does not get a reliability vote.
• Sample size and confidence are shown next to the number. Confidence describes how much evidence stands behind a score, not whether the score is good.

Weighting by equipment class

What matters depends on the machine. Serviceability weighs heavily on light commercial equipment because downtime costs money; heating output weighs more on a heat pump than on a cooling-only split, which is not offered a heating score at all. The weights below are configuration rather than code, and revising them changes every card of that class without touching a single page.