Methodology

1. What we do, and what we do not

We have not used any of these products. We do not own them, we have no laboratory, no gaussmeter, no power meter and no test protocol, and we are not going to pretend otherwise. Six of the eight largest roundups in this category pair the word “best” with a hands-on testing claim, and that is the one thing on this site we will not copy.

What we do instead, in order:

1. Read every product listing in full on a stated date - the title, every feature bullet and the whole specification table - and record what the seller publishes and what they leave blank.
2. Quote each figure verbatim with its source, so every number on a product page carries a line saying where it came from.
3. Compute the running cost of every product that publishes a wattage, from a primary-source electricity rate, showing the arithmetic.
4. Score each product on a fixed 100-point rubric whose every weight is published below.
5. Where a figure is not published, print Not published rather than an estimate. That is a finding, not a gap in our notes.

2. Product selection

A product is eligible for a ranked pick with a buy button only if all of these hold:

• It has a live retailer listing we read on a stated date.
• It is actually purchasable through a link we can build. Today that means Amazon: we hold no CJ or Awin programs, so a brand-direct product may be discussed for completeness but never gets a buy CTA.
• Published external dimensions exist, or the product is small enough that they do not decide fit.

Two brand-direct names people search for - Titan and VITAL+ - have pages on this site with no buy button, because we can neither link them nor earn from them. Those pages say so at the top.

3. The scoring rubric, with every weight

One hundred points across six components. No component requires handling the product, and every input is either published by the seller or computed from something they published.

3.1 Small-space fit (20)

Eight points for packing away when not in use, plus up to twelve for a small deployed footprint, computed from published external dimensions: 12 points at or under 6 sq ft, 9 up to 12, 6 up to 20, 3 up to 30, and 1 above that. A product with no published dimensions gets the stow points only.

3.2 Electrical fit (20)

Rated watts against a standard US 120 V, 15 A branch circuit, derated to 1440 W for a continuous load:

3.3 Cost to run (20)

Bands rather than a curve, so you can place a product yourself. Heat gear is banded on cents per session at its published session length; chillers on cents per hour at full draw, because a chiller’s monthly cost depends on a duty cycle we do not measure.

3.4 Key-number disclosure (15)

Every category has one number that decides the purchase and that a seller can either publish or duck. For an infrared heater it is EMF. For cold gear it is the cooling floor and how fast it gets there.

This scores disclosure, never safety. We do not measure EMF and we make no health claim about exposure at any level, in either direction. A manufacturer publishing a figure is a signal about the manufacturer’s transparency. What EMF numbers mean sets out the units and why measurement distance is load-bearing.

3.5 Spec completeness (15)

Fifteen points spread evenly across a fixed eight-item checklist, so the score is 15 x n / 8 rounded. Purely mechanical, and the fastest component to re-derive.

3.6 Warranty (10)

Published term in months: 10 points at 36 months or more (including a published lifetime term), 8 at 24, 6 at 12, 3 below 12, and 0 where no term is published. A promise of “lifetime customer support” is not a warranty term and scores zero, because support is not a commitment about the product.

3.7 The component that is held at zero, and why

4. Running-cost formulas

4.1 The electricity rate

Every figure on this site uses 18.34 cents per kilowatt-hour: the U.S. Total, Residential average for June 2026, from the U.S. Energy Information Administration’s Electric Power Monthly, Table 5.6.A. Re-fetched for this build on 2026-09-07. The same figure for June 2025 was 17.47 cents.

It is a national average and it conceals a spread of roughly three to one between the cheapest and most expensive states. Substitute your own rate from your bill - it is one multiplication in either formula below.

4.2 Resistive heating: blankets, saunas, tents

energy per session (kWh) = watts x (minutes / 60) / 1000
cost per session ($)     = energy per session x rate per kWh
cost per month ($)       = cost per session x sessions per week x 4.33

Worked example, a 1,050 W infrared tent over 30 minutes: 1050 x 0.5 / 1000 = 0.525 kWh. At $0.1834 per kWh that is $0.096, or about ten cents. Four sessions a week: 0.096 x 4 x 4.33 = $1.67 a month.

Mandatory caveat, printed with every such figure on this site: a resistive heater draws its rated power only while the element is energized. Once it reaches setpoint it cycles. So rated watts times time is an upper bound assuming continuous draw, never a measured value, and we label it exactly that way.

4.3 Refrigeration: chillers

cost per hour at full draw ($) = (rated watts / 1000) x rate per kWh
continuous upper bound ($/mo)  = cost per hour x 24 x 30.4

Worked example, a 600 W chiller: 600 / 1000 x 0.1834 = $0.110 an hour, and 0.110 x 24 x 30.4 = $80.31 as an absolute continuous bound.

We publish only those two figures and nothing in between. A chiller’s real consumption depends on duty cycle, which depends on ambient temperature, insulation, whether the lid is on, the setpoint and the water volume. We do not measure duty cycle. Publishing a single monthly figure as though it were measured would mean inventing the most important input.

4.4 Cross-category comparisons

Hold session length and rate constant, and state both. The full cross-category dataset is in what home recovery gear actually costs to run.

5. Sourcing standards

1. Every specification cites its source - the seller’s product title, a feature bullet, the specification table, or a manufacturer page - with the date we read it.
2. Every external figure links to the primary source. We cite the EIA, not a blog quoting the EIA; the CPSC notice, not a news article about it.
3. Health and physiology claims cite peer-reviewed literature, and the evidence page carries a not-medical-advice line and reports the findings that cut against the popular claim as well as the ones that support it.
4. Prices are never static text. They render from the live data layer or the button reads “Check price on Amazon”.
5. Where a fact could not be verified, we print “Not published”. We never fill a gap with a plausible number.

6. What we cannot determine

The honest limits of everything above:

• We have not used these products. Comfort, noise, build feel and durability are outside what we can say anything about.
• Published specifications can be wrong. We report what is published and cite it. Two tubs in our registry publish two different capacities on the same page, and a chiller publishes the same wattage as its own half-size sibling - we flag those rather than resolving them, because resolving them would mean guessing.
• Running costs are computed from rated power, not measured. The caveats in section 4 apply to every figure.
• EMF figures are manufacturer-reported and not verified by us. We do not own a meter.
• We do not carry owner ratings - see section 3.7.
• Our score measures published evidence quality and small-space fit, not real-world performance. A low score means a seller published little, not that a product is bad.

7. Scope of what we read

For each of the 44 products in the registry we read the live retailer listing on the date printed beside every figure on its page: the product title verbatim, every feature bullet, and the specification table where the listing has one. That is a narrower scope than owning the product and it is a wider one than reading a title, and we would rather state it precisely than let a reader assume either.

8. Updates and corrections

Roundups and comparisons are reviewed quarterly, brand pages and cost guides every six months, and this page every six months or whenever the electricity rate materially changes. A recall or safety notice is a same-week priority.

The daily price refresh is not a content change and never touches a page’s modification date - a sitemap that claims every page changed every day is a sitemap Google stops trusting. Corrections appear on the page, dated, rather than as silent edits. The editorial policy has the rest, and corrections are welcome.