Method / 01

Exact mattress first. Product-line context second.

The Research Library begins with the full mattress name. A visitor can open one exact research page without decoding our taxonomy. Product-line and brand pages are optional context for related models, shared policies, and naming history.

Every separately marketed mattress name gets its own page. Firmness, height, and size choices sold under that same name stay together on the page, where any documented construction differences are explained. A source-supported former name appears in the version history instead of being counted as a second mattress. The research covers the layer stack, dimensions, firmness, coil and foam specifications, policies, and version-matched findings.

Why the separation matters

A 365-night company trial is not a construction detail. A Classic coil count cannot be applied to every Saatva mattress. Mixing those levels creates false facts.

Research architectureThe exact mattress stays at the center
Mattress research pageModel and version details

The mattress, construction, and relevant options being evaluated.

FamilyRelated context only; never transferred automatically
BrandCompany information and direct-sale policies
SourcesOfficial, independent, owner-reported, archive

Method / 02

Four evidence statuses

Verified document

Current primary source located and linked: product page, legal policy, government record, or certifier record.

Brand claim

Attributed company statement that has not been independently replicated by this publication.

Owner report

Self-selected experience used to surface questions, never to manufacture a prevalence rate.

Open question

Missing, conflicting, outdated, or version-ambiguous information that remains unresolved.

Method / 03

The workflow

Research workflowFive checks between a mattress name and a conclusion
  1. 01IdentifyFreeze the exact model, configuration, and date.
  2. 02CollectGather product, policy, archive, and test sources.
  3. 03ReconcileSeparate confirmed facts from conflicts and claims.
  4. 04SynthesizeCompare like with like and preserve source scope.
  5. 05PublishShow the conclusion, citations, limits, and open gaps.

The rule at every stage: facts, attributed claims, analysis, and unanswered questions never share the same label.

  1. Freeze a market, product name, configuration, version window, and retrieval date.
  2. Collect the current product, specifications, law-label, shipping, trial, return, warranty, care, certification, and corporate records.
  3. Normalize layer order, dimensions, material language, foam density, coil count and gauge, price, fees, policy clocks, remedy stages, and support requirements.
  4. Search for conflicts across pages and compare current documents with current or archived versions when a build changed.
  5. Review owner reports by model and purchase period to identify recurring questions. Record their sampling limits.
  6. Publish supported findings, unanswered questions, and a linked source list. Never fill a blank with an industry assumption.

Method / 04

How published review scores work

The number is not a Nappedia hands-on score. It summarizes published expert overall ratings that can be matched to the mattress, configuration, and generation.

1Search, then match

We record the review pages checked, then separate exact-current scores from broader model-level evidence.

2Normalize to 100

A 9.5/10 and a 4.5/5 become 95 and 90. Brand customer-star ratings are not included.

3Require two sources

One overall rating is not enough. With fewer than two matched ratings, the score remains unavailable.

4Use the median

The middle normalized value reduces the influence of one unusually high or low overall rating.

Worked exampleHow three scope-matched publisher ratings become one median
Source A8.8 / 1088 / 100
Source B4.6 / 592 / 100
Source C9.5 / 1095 / 100
Ordered values

88 → 92 → 95

Published median: 92 / 100

Illustrative numbers only. Publisher methods are not equivalent. Each research page lists its included and excluded sources and preserves the method used.

The score does not prove comfort, durability, safety, or medical benefit. When exact and broader current review sets both exist, the page shows both medians without blending them. It also lists the pages reviewed but not averaged and explains every exclusion, including a different height or generation, duplicate testing, owner aggregates, missing overall scores, or an unverifiable method. The search count is a dated research snapshot, not a claim that no other page exists anywhere online.

Method / 05

How owner reports are used

Voluntary reviews and complaints are discovery sources. They disproportionately represent people motivated to post, can be unverifiable, and often omit foundation, body weight, version, purchase date, or measured impression depth.

Repetition therefore creates an investigation queue, not a defect rate. A stronger future dataset would require verified purchase, structured questions, version matching, follow-up intervals, denominators, and publication of positive and negative outcomes.

Method / 06

What document research cannot prove

It cannot independently measure pressure relief, surface or core temperature, motion transfer, edge deflection, noise, odor, recovery, or long-term indentation. It also cannot predict comfort for an individual sleeper.

Nappedia does not publish its own physical-performance scores. A company performance claim remains a company claim even when plausible.

Method / 07

The commercial firewall

  • Commission size is not a research field.
  • Affiliate availability cannot change a factual finding.
  • A non-affiliate mattress can lead a comparison.
  • Negative supported findings are not deleted for a merchant.
  • Paid links are labeled close to the link and use appropriate sponsored relationship attributes.
Read the editorial policy ↗