How we verify the data

1. Every value carries its context

An amplifier rated “165W” on its own is nearly meaningless: measured with how many channels driven at once, into what speaker load, at what distortion? The very same receiver can be advertised honestly at 80W (two channels driven, 8 ohms, low distortion) and at 165W (one channel, an easier load) — the numbers differ because the conditions differ. So the site never stores a measured value without the conditions it was measured under; it treats value + contextas one inseparable pair. Power is recorded together with its channels-driven, impedance, and distortion figures; a speaker’s sensitivity together with whether it was measured in-room or anechoic; a frequency range together with its dB tolerance. Numbers taken under different conditions are kept in different fields, and are never averaged together or ranked against each other.

When a source gives a number but omits the conditions, we don’t invent them and we don’t drop the number — we record the context as “unstated.” That word is doing real work: it’s the honest signal that the figure exists but can’t be compared like-for-like until someone states how it was measured. Because the site stores that silence as data instead of hiding it, it can also count it — brand by brand — in the Spec-Sheet Honesty Report.

2. Every number is traceable to its source

Provenance is tracked per field, not per product. Each specification records where that individual value came from — a source URL — and a confidence tier describing how much weight it earns, because sources disagree field by field. When two sources agree, we keep the citation from the stronger one; when they disagree beyond a set tolerance, the value is routed to human review rather than silently overwritten. The product pages surface this directly: the specifications table is labelled source-tracked, and each field shows the tier it was recorded at.

3. Yes, no, and “not stated” are three different answers

A feature is either present, absent, or simply never mentioned — three states, not two. A retailer that never lists a phono input hasn’t told you the unit lacks one, so we record an unmentioned feature as unknown, never as “no.” Absence of evidence is not evidence of absence. That’s why you’ll occasionally see “unknown” here where another site shows a confident X: the X is often a guess, and the honest answer keeps the data from inventing facts.

4. Prices are real listings, kept as evidence

Prices come from tracked offers at real retailers, pulled through their listings and refreshed on a schedule; each offer keeps the exact listing title it was found under, retained as evidence that it’s the right product. Matching those listings to the right catalogue entry follows one guiding principle: a wrong price is worse than a missing one. So an offer is dropped only when its own title gives a positivesignal that it’s a different thing — it names a different model variant, the opposite sale unit (a single where the product is sold as a pair), or a multi-pack — never merely because a match couldn’t be confirmed. When in doubt the offer is kept and shown with its source, and the site would rather show no price than a confidently wrong one. It never fabricates an offer, a stock status, or a price it doesn’t actually hold from a tracked retailer. Prices are indicative, not a live checkout quote, so always confirm the final number on the retailer’s own page.

5. What we don’t do

There are no user reviews or star ratings on this site, and — just as importantly — none are invented behind the scenes: the pages never emit a fabricated rating or review in their structured data, because none exists. There are no sponsored placements and no paid rankings; the compatibility verdicts are computed from the data, not bought. Some retail links are affiliate links, disclosed wherever they appear and carrying the search-engine markup that flags them as such. We may earn a commission at no cost to you; it never affects rankings or compatibility verdicts — those come from the data.

Catching conflicts, and honest limits

When sources disagree beyond tolerance, that cross-source check routes the value to human review instead of picking a winner. Some disagreements aren’t errors: when one model number genuinely ships with different internals by production lot — the panel-lottery TVs are the classic case — we flag it as a variant, blank the conflicting field, and note the variance rather than thrash between values. None of this makes the catalogue perfect. Coverage grows category by category, not every product carries every field yet, and a clean compatibility result rules out the common mechanical mistakes but can’t judge how a system will sound or look in your room. Where we genuinely don’t know something, we say so.