Sourced fact
Visible source and scope for dimensions, specifications, history, or regulation.
EVIDENCE LAYERS
Keep facts, interpretation, unknowns, and volatile commerce separate.
ICU interpretation
Decision guidance from explicit inputs; never rewritten as a source fact.
Unknown / not reported
No zero, average, substitute, or hidden inference is inserted.
Volatile field
Price, availability, lead time, and offer state must be rechecked at the current merchant/source.
BEDS DECISION INTELLIGENCE / PRACTICAL PATH
Move one explicit question at a time.
Named size, product geometry, compatibility, movement, and cost remain separate authorities. Unknown fields stay unknown, and carrying measurements is always a deliberate action.
EXPLICIT ATTRIBUTES / VISIBLE WEIGHTS
Transparent Product Compare
Compare current host-supplied bed products or offers only when explicit attributes exist, using priorities and weights you can see and control.
CURRENT HOST STATE
No product order is fabricated when the host supplies none
This portable branch cannot name a winner, invent comparison rows, manufacture a price, infer missing specifications, or synthesize review/testing data. ICU must supply current candidates and explicit attributes first.
RANKING METHOD
Every step that can change order is inspectable
- 01
A comparison starts only from current ICU-host-supplied products/offers and explicit comparison attributes.
- 02
The visitor declares every ranking metric, its weight, and whether higher, lower, or an exact target is preferred.
- 03
Numeric maximize/minimize metrics use transparent min-max normalization across only the complete rankable candidate set; target metrics score exact matches only.
- 04
A candidate missing any prioritized metric is unranked for insufficient data rather than assigned zero, average, worst, best, or inferred values.
- 05
If a hard exclusion cannot be evaluated because its attribute is missing, the candidate is unranked with an unresolved-exclusion state rather than assumed eligible.
- 06
Matched hard exclusions are listed with the exact metric and raw value that triggered exclusion.
- 07
Equal weighted scores share a dense rank. Name/product-key ordering inside a tie is deterministic presentation only and carries no merit.
- 08
Scores are relative to the supplied candidate set and declared priorities. They are not universal ratings, quality scores, testing results, popularity scores, or recommendations.
CAN ENTER RANKING
Two authorities only
- Evidence-aware factual product specifications supplied explicitly.
- Current-offer attributes supplied explicitly by ICU, with observation date.
Nothing enters merely because it is present on a sales page, expensive, popular-looking, editorially featured, affiliate-linked, sponsored, or attached to a provider.
NEVER ADDS MERIT
Commercial and adjacent signals stay outside score
- affiliate relationship
- sponsorship or paid placement
- editorial inclusion by itself
- provider presence
- lead destination availability
- high price or high-ticket status
- premium/luxury/custom/bespoke labels
- compatibility results
- Room Fit results
- Move-It results
- True Cost results
MISSING DATA
Unknown means unranked, not zero
If even one visitor-prioritized attribute is missing, that candidate receives no hidden substitute and is listed as insufficient priority data. If a hard exclusion cannot be evaluated because its field is missing, it becomes unresolved exclusion. This prevents missing data from quietly helping or hurting a product.
SEPARATE AUTHORITIES
Useful context can sit beside the order without becoming points
- Compatibility may be shown only as a separate CompatibilityRule result with its own evidence/conditions.
- Room Fit, Move-It, and True Cost may be linked or summarized only under their own planning authority and never converted to comparison score points.
- Affiliate and sponsored relationships may be disclosed beside a candidate but never change score, eligibility, or tie order.
- Provider/lead availability may be shown in high-ticket research but never changes product merit.
What a future populated result must show
For each ranked candidate: exact raw attribute values, source authority, visitor weight, normalization result, weighted contribution, total relative score, dense rank, commercial disclosures, missing fields, exclusions, and tie behavior. A result may describe the order as your declared-priority order; it must not call the first row universally “best.”