Skip to content

Methodology

What we changed in the 2026 rubric

Three changes to how we score, effective from the January 2026 round. Two of them moved published scores, and we have listed every one that changed.

Certified Sleep Science Coach (Spencer Institute) · 43 mattresses tested since 2021
Published Updated 2 min read

We buy or independently request every mattress we test. If you buy through a link here we may earn a commission, which never affects our ranking. How this works.

We revise the rubric once a year, in January, and publish what changed before the first review of the new round goes out. Three things moved for 2026.

Durability went from 6% to 10%

We now have five years of annual sag measurements on the mattresses we bought in 2021, which is long enough for the numbers to mean something. When we set the original weighting we were projecting from twelve-month data and hedging accordingly.

The five-year figures separate constructions far more cleanly than the one-year figures did. Latex and coil-on-coil builds are pulling away from foam-over-coil in a way that was invisible at year one, and a scoring system that under-weights that is telling you about the first eighteen months of a decade-long purchase.

Edge support is now measured twice

The old test was a single dial-gauge reading after sixty seconds seated on the perimeter. That turned out to conflate two different behaviours: mattresses that compress and then hold, and mattresses that keep giving way.

The new test reads at one minute and again at ten. The gap between the two readings is the interesting number, and on two mattresses in our set it is larger than the initial compression. A perimeter that reads 1.2 inches at one minute and 2.1 at ten is a worse mattress to sit on than one that reads 1.5 and stays there, and the old method scored them identically.

Cooling covers are no longer scored separately

We used to award a small credit for phase-change and other cooling covers on top of the measured temperature result. That was double-counting: if the cover works, it shows up in the eight-hour surface temperature rise, which is already 15% of the score.

Worse, it rewarded a claim rather than a result. Two mattresses with identically described covers measured 0.9°F apart in the same conditions.

Scores that changed

Four published scores moved as a result. All four reviews carry a dated note at the foot explaining the change.

MattressOldNewReason
DreamCloud Premier8.78.6Edge creep at ten minutes; cover credit removed
Nectar Premier8.38.1Durability reweighting; edge creep
Saatva Latex Hybrid8.78.9Durability reweighting
WinkBed Plus8.58.7Durability reweighting

No score changed by more than 0.2, which is roughly what we would expect from a weighting adjustment rather than a methodology overhaul. If a rubric change moves scores by a point, the rubric was wrong before or is wrong now.

What did not change

The seven categories, the thirty-night minimum, the three tester weights, and the room conditions. Comparability across rounds is worth a great deal, and we would rather live with a slightly imperfect measure than lose the ability to compare a 2022 mattress with a 2026 one.

The full current rubric, including the weighting of every category, is on the how we test page.

More from the journal

Reader questions

Reader questions, July 2026

Answers to the questions we were asked most this quarter: whether toppers fix sag, when a Split King is worth it, and adjustable base compatibility.

3 min read
Analysis

When mattresses are actually cheapest

Eighteen months of monthly price tracking across six brands. Which sale events carry real discounts, and which brands are permanently on sale.

2 min read
All journal entries