Mattress shopping is built on numbers that do not travel between brands. This guide sorts the specs that tell you something from the ones that only sound technical, and shows how to read the trial and warranty terms that decide what happens if you guess wrong.
The Review at NYU is not affiliated with New York University.
Short answer. Compare mattresses on support and pressure relief for your sleep position and body weight, on durability signals such as foam density and coil construction, and above all on the return and warranty terms, because comfort is personal and can only be confirmed at home. The best research evidence, a 2003 randomised trial in The Lancet and later systematic reviews, favours a medium-firm mattress for people with chronic low back pain, but firmness labels such as "6 out of 10" are not standardised and mean different things at different brands. Treat medium-firm as a starting range, not a product spec.
We did not sleep on, measure or test any mattress for this article. It is a research guide built from peer reviewed studies, federal regulations, regulator guidance, certification bodies, a trade association's technical definitions, consumer testing publications and one manufacturer's published warranty. That limits what we can say, and we say so where it matters. Before writing any advice, we set four criteria that every claim in this guide had to pass:
This is the same discipline we apply in our review method for language learning apps: decide what counts before looking at any product, so marketing cannot set the agenda.
Almost every mattress page quotes a firmness rating, usually on a 1 to 10 scale. These numbers are useful only within one brand's lineup or one reviewer's set of tests. We found no regulation or binding industry standard that US mattress sellers must use when they assign them, and in practice each company calibrates its own scale.
The scale direction is not even fixed. In the 2003 Lancet trial discussed below, mattress firmness was described on a European scale where 1.0 was the firmest and 10.0 the softest, according to the PEDro database record of the study. Many US brands use the opposite convention, with 10 as the firmest. A "firm 2.3" in a research paper and a "firm 8" on a product page can describe similar beds.
Firmness is also not a single property. Two concepts get blended into the word:
The Polyurethane Foam Association, the US foam industry group, makes a related point about foam itself: firmness, measured as indentation force deflection, is independent of foam density. A high density foam can feel soft and a low density foam can feel firm. So "firm" tells you about feel at a given moment, not about how long that feel will last.
Independent reviewers who use the same method across many mattresses (outlets such as Consumer Reports and RTINGS run repeatable in house tests) give you a better comparison than brand labels, because at least the ruler stays the same. Even then, a reviewer's firmness number reflects their testers' bodies, which may not resemble yours.
"Medium-firm is best for your back" is one of the few mattress claims with real research behind it. It is worth knowing exactly what that research measured.
Kovacs and colleagues ran a randomised, double-blind, multicentre trial in Spain, published in The Lancet on 15 November 2003 (volume 362, pages 1599 to 1604). They assigned 313 adults with chronic non-specific low back pain to either a firm mattress (2.3 on the European scale above) or a medium-firm one (5.6), and assessed them at baseline and at 90 days. According to the PEDro record, people on the medium-firm mattress were more likely to improve on pain in bed (odds ratio 2.36, 95 percent confidence interval 1.13 to 4.93) and on disability (odds ratio 2.10, 1.24 to 3.56). The result for pain on rising (odds ratio 1.93, 0.97 to 3.86) did not reach statistical significance at 90 days.
What the trial shows: in people who already had chronic low back pain, a medium-firm mattress beat a firm one over three months. What it does not show: that medium-firm is best for people without back pain, for every sleep position, or at every body weight, or that any particular product labelled "medium-firm" matches the mattresses tested.
Radwan and colleagues published a systematic review in Sleep Health in December 2015 (volume 1, issue 4, pages 257 to 267). They included 24 controlled trials published since 2000, classified mattresses as soft, medium-firm, firm or custom inflated, and rated trial quality as moderate to high on the PEDro scale. Their conclusion was that a mattress subjectively identified as medium-firm and custom inflated, meaning self-adjusted, was optimal for sleep comfort, quality and spinal alignment. They also said evidence was not yet sufficient on the right mattress temperature.
Caggiari and colleagues reviewed 39 articles from 2000 to 2019 in the Journal of Orthopaedics and Traumatology in 2021 (volume 22, article 51). They concluded that medium-firm mattresses offered more advantages for people with non-specific low back pain, and flagged real weaknesses: studies varied considerably, participant characteristics were poorly standardised, and the type of back pain was poorly defined.
A 2019 biomechanical review in PeerJ by Wong and colleagues, covering 18 studies from 2008 to 2018, found that the stiffness people need depends on body shape and spinal curvature, and criticised the European firmness guidance (EN 1957) for being based on subjective perception and on lying on the back only. Read together, the evidence points less to one ideal firmness and more to fit: the right mattress is the one that keeps your spine reasonably aligned in the positions you actually sleep in. That is the same conclusion a careful reader reaches with most single headline findings, a habit we describe in our guide to reading claims against the evidence.
Most mattresses sold today fall into five construction families. The table summarises typical tendencies described by manufacturers' technical material and independent testers. These are tendencies, not rules: a well built model in one category can beat a poorly built one in another.
| Type | What it is | Typical strengths | Typical trade offs | Specs worth asking for |
|---|---|---|---|---|
| Memory foam (all foam) | Layers of viscoelastic and standard polyurethane foam, often shipped compressed in a box | Contours closely; usually strong motion isolation | Slower response can feel "stuck"; may feel warmer to some sleepers; edges often softer | Foam density of each layer, layer thicknesses, CertiPUR-US foam |
| Latex | Natural, synthetic or blended rubber foam (Dunlop or Talalay processes) | Springy, responsive feel; natural latex is often marketed for durability | Heavy; usually priced higher; "natural" can mean a blend | Natural versus synthetic share, GOLS or other certificate, firmness per layer |
| Innerspring | Steel coil support core with thinner foam or fibre comfort layers | Bouncy, easy to move on; coil cores allow airflow | Interconnected coils can transfer motion; thin comfort layers may wear first | Coil type (pocketed or connected), coil gauge, edge reinforcement, comfort layer materials |
| Hybrid | Pocketed coil core plus a substantial foam or latex comfort section | Combines contouring with bounce and better edge support than many all foam beds | Heavy; the label "hybrid" covers very different designs | Thickness of the comfort section, foam densities, coil gauge and zoning |
| Airbed (adjustable air) | Air chambers with a pump, usually under foam layers; often dual chambers for two sleepers | Firmness adjustable over time and per side, close to the "self-adjusted" design favoured in the 2015 review | Pumps, hoses and valves can fail; more parts to cover under warranty | Warranty on the pump and electronics separately, chamber layout, noise |
Two definitions come up throughout. Foam density is mass per unit volume, expressed in pounds per cubic foot (pcf) in the US. Indentation force deflection (IFD, older name ILD) is the Polyurethane Foam Association's measure of load bearing: under the ASTM D3574 test method, it is the force in pounds needed to press a 50 square inch circular plate into a foam sample about 4 inches thick and typically 15 inches square or larger, usually reported at 25 percent compression. Higher IFD means firmer foam. Few retail pages publish IFD, but latex and foam specialists sometimes do.
Apply the criteria from the start of this guide and mattress specs sort themselves into three groups.
The guidance below is a starting point drawn from the research above and from how pressure and support usually work, not a prescription. The 2019 PeerJ review is clear that the right stiffness depends on individual body shape, and the trials mostly enrolled people with back pain. If you have a medical condition, a clinician's advice outranks any mattress guide, including this one.
Your shoulder and hip carry most of your weight on a small area. Many side sleepers do better with enough comfort layer depth for the shoulder to sink slightly, so the spine stays roughly level. Pick a softer surface or thicker comfort section if you wake with shoulder or hip soreness on a firm bed. Watch for hip sag on very soft beds.
Back sleeping is the posture closest to the setups in the medium-firm trials. Pick a medium to medium-firm feel if you want to start where the evidence is strongest, and check that your lower back does not feel unsupported.
The concern is the hips sinking and arching the lower back. Pick a firmer support and a thinner soft top layer if you sleep mostly on your front.
If you change positions through the night, responsiveness matters as much as firmness. Pick a bed that is easy to turn on, such as latex, hybrid or a faster responding foam, and aim for the middle of the firmness range.
A firmness label is set by people of some particular weight. Heavier sleepers press deeper into a mattress, so a bed can feel softer to them than its label and wear faster in the comfort layers. Pick higher density foams, sturdier coils and a reinforced edge if you are heavier. Lighter sleepers may barely compress a firm bed and find it feels harder than advertised, so a softer label can be the better comparison point.
Compare motion isolation and edge support in independent tests, and check the width you actually use. If you and your partner disagree on firmness, look for split firmness options or dual chamber airbeds, which let each side be set separately and echo the "self-adjusted" design favoured by the 2015 review.
Evidence here is thin, as noted above. Coil cores usually allow more airflow than solid foam cores, but bedding, room temperature and pajamas also matter. Pick a mattress with a generous trial if heat is your main complaint, since it is hard to predict from specs.
Because comfort can only be confirmed at home, the trial period is the most important spec on the page. A long trial is worth little if returning costs a large share of the price. In a 2015 report on mattress return policies, Consumer Reports said its subscribers considered the return policy paramount, noted that retailers generally assume about 30 days for a body to adjust to a new mattress, and reported that almost 40 percent of subscribers with regrets about a mattress purchase wished they had done more research (from its subscriber survey at the time). The same report suggested lying on a mattress in a store for at least 10 minutes in each position you sleep in, while warning that a short test can differ from a full night.
Read these terms before you pay:
Online purchases add one federal protection. Under the FTC's Mail, Internet, or Telephone Order Merchandise Rule, a seller must ship within the time it states, or within 30 days if it states no time, and must tell you about delays and offer you the choice to cancel for a full refund. That covers delivery, not comfort returns, which depend on the seller's own policy and your state's law. Save a copy of the trial terms the day you order, since policies change.
A warranty covers defects, not comfort. Its length gets the headline, but the conditions decide whether you can ever use it. The FTC's consumer guidance on warranties says that whenever there is a warranty, the law requires it to be available for you to read before you buy, in store or online. It also advises checking what is excluded, whether you must pay shipping or labour, and whether you must register the product first. State implied warranties can apply too, lasting up to four years in some states, according to the FTC.
In a non-prorated period the maker repairs or replaces a defective mattress without charging you a share of its price, though you may still owe transport or inspection fees. In a prorated period you pay a growing portion of the cost as the mattress ages. Sealy's published warranty for mattresses bought from 2021 onward, checked on 7 November 2026, is a clear example: its law tag shows a code such as 5/10, meaning five years without a deductible within a ten year total. After the first period, the deductible equals the current list price multiplied by the years since purchase and divided by the total warranty years. In Sealy's own worked example, a $500 mattress with a 5/10 code carries a $400 deductible in year 8. A "10 year warranty" can therefore be worth much less in its second half.
Most mattress warranties only treat body impressions as a defect beyond a stated depth. Sealy's terms, again as an example checked on 7 November 2026, require impressions deeper than 0.75 of an inch on flat top mattresses and deeper than 1.5 inches on quilted or tufted ones, measured with no weight on the bed. Other makers set different thresholds, so compare them directly: a smaller threshold is more protective.
Every mattress made for sale in the US since 1 July 2007 must meet the Consumer Product Safety Commission's open flame standard, 16 CFR Part 1633. In a 30 minute test, the peak rate of heat release must not exceed 200 kilowatts, and total heat release must not exceed 15 megajoules in the first 10 minutes. When the rule took effect, the CPSC estimated it would prevent as many as 270 deaths and 1,330 injuries a year once fully in effect. A separate rule, 16 CFR Part 1632, covers cigarette ignition. Because compliance is mandatory, "meets federal fire standards" does not distinguish one mattress from another. If the material used as a fire barrier matters to you, ask the seller what it is.
CertiPUR-US certifies polyurethane foam, not the finished mattress. According to the program, certified foam is tested for content, emissions and durability, is made without formaldehyde, ozone depleters, CPSC regulated phthalates, mercury, lead and other heavy metals, and has low VOC emissions of less than 0.5 parts per million. It says nothing about latex, fabrics, glues or coils, and it is not an organic certification.
The Global Organic Textile Standard covers textiles such as cotton or wool covers. A product labelled "organic" under GOTS must contain at least 95 percent certified organic fibres; "made with x percent organic materials" means 70 to 94 percent. The Global Organic Latex Standard covers latex; the BedTimes industry glossary describes it as a third party program for latex from certified organic plantations, with a 95 percent threshold. Check which component is certified: a GOTS certified cover on a conventional foam core is not an organic mattress. Ask for the certificate or scope document, which names the certified company and products.
The lesson for a buyer: ask which component a claim covers and who certified it, and treat a badge you cannot trace to an independent body as decoration.
Use this for every mattress on your shortlist, and keep the answers in one place so you compare like for like.
If you are buying during a sales event, the same checklist applies. A discount does not change the return fees or the sag threshold, and it can come with weaker terms on clearance stock.
It is the best supported starting point, not a guarantee. In a 2003 Lancet trial of 313 adults with chronic low back pain, people given a medium-firm mattress reported less pain in bed and less disability after 90 days than people given a firm one. Later systematic reviews in 2015 and 2021 reached similar conclusions but noted that the underlying studies varied widely in design. Medium-firm is also defined differently by every brand, so treat it as a range to test during a trial.
Only what the company or reviewer using it decides. We found no regulation or binding industry standard that US mattress sellers must follow when they put a number on firmness, so a 6 from one brand can feel different from a 6 from another. Some scales even run the other way: the scale used in the 2003 Lancet trial put the firmest mattress at 1 and the softest at 10.
Not by itself. Coil count depends on mattress size, coil type and whether small microcoils are counted, so two numbers are rarely comparable. Coil gauge, how the coils are wrapped, edge reinforcement and the comfort layers on top tell you more. Compare coil counts only between mattresses of the same size and design.
Check the minimum number of nights before a return is allowed, the last day to start a return, whether you get a full refund or only an exchange or credit, who pays pickup or shipping, any restocking or processing fee, and whether a cover or foundation must stay unopened. Get the policy in writing or save a copy of the page on the day you buy.
No. CertiPUR-US certifies polyurethane foam, not the whole mattress, for content, emissions and durability, including low VOC emissions of less than 0.5 parts per million. Organic claims for textiles and latex are covered by different certifications such as GOTS and GOLS, and the FTC has acted against mattress sellers whose organic or VOC free claims were not supported.
It depends on the warranty, so read the exact threshold before you buy. As one example, the Sealy warranty for mattresses bought from 2021 onward, checked on 7 November 2026, covers impressions deeper than 0.75 of an inch on flat top models and deeper than 1.5 inches on quilted or tufted models, measured with no weight on the bed, and it also sets foundation requirements.