I have tested and written about air quality gear independently since 2019. I don’t run sponsored reviews, I’m not paid for scores, and when a company sends a device it has no say in what I publish. This page explains exactly how I arrive at the numbers you see on every review.
Here is the honest part most sites won’t tell you: across all three categories there are hard limits on what anyone can truly verify. I don’t own a reference-grade monitor, only occasional access to one, and even with a reference in hand, humidity, particle type, the individual unit, and the conditions on the day mean no single accuracy figure could ever capture everything. With masks and respirators, fit is personal and can’t be judged on your behalf, and properly verifying filtration takes laboratory equipment that costs many thousands of dollars.
So these scores are not claims that a reading is objectively correct. What runs through every category instead is a pair of commitments: transparency about how I reached a number, and honesty about what that number can and can’t tell you. Where I can measure something I do; where I can’t, I lean on third-party results and studies; and I score against predetermined tiers rather than ranking devices arbitrarily, so a 7 means the same thing in one review as it does in the next.
Tested
Measured first-hand wherever I can, on my own bench: clean-air output, sound, sensor response, and temperature and humidity against a calibrated instrument. The starting point for every score.
Verified
Backed by independent third-party lab results, published studies, and recognised certifications, for the things I can’t measure myself.
Compared
Weighed against the 50+ devices I’ve tested, so the judgement calls, ease of use, comfort, build, stay consistent from one review to the next.
Air quality monitors
The hardest category to score honestly, because what I can verify differs from one reading to the next. I care more about how the data behaves than about any single accuracy figure.
I split this into two questions. Trend reliability: does the device react promptly and consistently when your air changes, and settle back when it clears? I can test that for almost anything, without a reference. Absolute agreement: do the readings line up with something trustworthy? That’s only occasionally possible, and it’s strongest for CO2, temperature and relative humidity. Responsiveness and consistency are important, but I don’t weigh them above raw closeness, because accuracy is something you rely on day to day as well.
PM is about reaction as well as the raw number. Reaction matters less than the absolute reading, but both count: how quickly the reading climbs when I introduce a known source and how fast it recovers, plus the polling rate and reading interval, how often the device truly updates, since a slow or heavily smoothed monitor can look calm while your air isn’t. I judge the sensor module on its documented behaviour, and compare against a reference on the occasions I get access to one.
Temperature and humidity I can hold to a real standard: I compare them against a lab-tested, calibrated hygrometer, so these are among the few readings I can check directly.
CO2 is another one I can get genuinely close on. I don’t use a certified reference, but I do have factory-calibrated devices that serve as a sensible baseline, not a true reference, but a fair yardstick, and readings are anchored to outdoor ambient near 420 ppm, with calibration and ABC drift behaviour weighed heavily.
VOCs are reported as an index, never an absolute concentration, so they’re judged purely on relative behaviour: whether the index responds sensibly to a known source and settles back.
Other pollutants vary a great deal from one device to the next, so for these parameters I generally lean on third-party studies rather than my own readings.
Half quality, half repairability. Quality covers materials, fit and finish, and the display. Repairability covers whether you can open it, service it, and get hold of parts. The two combine evenly into the score.
Split three ways: how long it runs on a charge, how convenient it is to recharge, and how well the cell holds up over its lifespan. I also factor in whether it takes disposable or rechargeable batteries: where a device runs on disposables, I weigh the ongoing cost of keeping it powered in place of recharging convenience. Mains-only monitors carry no battery, so this is marked N/A rather than counted against them.
The most subjective of the monitor scores, and a relative one. Having tested over 50 devices, I have a clear sense of what counts as easy and what doesn’t, so I rate setup, the app and data export, and day-to-day use against that experience rather than against a fixed yardstick.
I start from how much data quality you get for the price, then nudge it up or down for how well the device is built and how pleasant it is to live with. It’s derived from the other scores rather than felt.
Air purifiers
The friendliest category to score, because the things that matter most are things I can measure myself. CADR can be derived from how fast a purifier clears particulate from a sealed room, which I do from the PM2.5 decay curve, and sound can be measured just as directly with a meter.
Scored on an absolute scale, not relative to other models, so a new purifier simply lands where its numbers put it and nothing else has to be re-ranked. I measure clean air delivery from the PM2.5 decay in a sealed room, run in R, and place it on the fixed CADR tiers below. The tier comes purely from the measured figure; where a manufacturer has clearly overstated its claim, I take a point or two off depending on how far off it is.
Roughly 600+ m³/h (about 350+ CFM).
Around 400–600 m³/h (about 235–350 CFM).
Around 250–400 m³/h (about 145–235 CFM), fine for a normal room.
Around 120–250 m³/h.
Under about 120 m³/h.
Measured in dB(A) at each fan setting, at a fixed distance, and placed into set brackets. What matters is the noise at a speed that still does useful work, not the quietest or loudest extreme.
Comfortable to sleep beside at a speed that still cleans (roughly under 35 dB(A) at a useful setting).
Runs in a bedroom without bother, or at full speed in a living room without being a nuisance (about 35–45 dB(A)).
Fine in a living space but clearly audible (about 45–52 dB(A)).
Hard to ignore at the speeds you actually need (about 52–58 dB(A)).
Disruptive at any genuinely useful speed (58 dB(A) and up).
Construction quality and the features that matter day to day: controls, auto mode, onboard sensors, app, and display. Filtration and clean-air output live under Performance, not here, so this score is purely about how well the machine is made and how usable it is.
Clean air per pound is the backbone: measured CADR against the upfront price, then weighed against what it costs to keep running. Running cost is split evenly between filters and electricity (plus any other consumables).
Masks and respirators
Protection comes down to filtration and fit. I can score filtration against clear thresholds, but fit is personal, and I’m emphatic about what that means for you.
Scored against fixed thresholds. A recognised certification (N95, FFP2, KF94, N99) does the heavy lifting; where a mask carries no formal standard, I use independent third-party lab results for particle and bacterial filtration efficiency (PFE and BFE) and name the lab. The top bands need certification or a named lab result, an unverified marketing percentage cannot reach them.
Certified to a recognised standard, or independently tested at 99% efficiency or better.
95% or better by certification or third-party test.
Roughly 90–95% on credible testing.
Roughly 70–90%, or thin evidence.
Under about 70%, or claims with nothing to back them.
I can’t score how a mask fits you, but I can score how hard the design tries to fit everyone, the features that widen the odds of a good seal across different faces. More sizes, adjustable head straps, and a proper mouldable nose wire and foam all push this up; a one-size mask on fixed ear loops with no nose wire pulls it down.
Multiple sizes, adjustable head straps, and a proper mouldable nose wire (usually with nose foam). A realistic shot at a good seal on most faces.
At least two of: more than one size, adjustable straps, a mouldable nose wire. Good odds for a wide range of faces.
One meaningful adjustment (a nose wire, or adjustable loops) but otherwise one-size. Fine for average faces, marginal for others.
Fixed size and fixed ear loops, with a weak nose wire or none. Seal depends heavily on your face shape.
One size, no real adjustment, no nose wire. Many faces will leak.
This is not a fit score, and you cannot inherit mine. Adjustability rates the design’s potential to seal, never whether it seals on your face. A mask that fits me may leak badly on you, and a leaking respirator gives a fraction of its rated protection no matter how good the filter is. If you rely on a respirator for real protection, fit-test it yourself.
Banded against actual inhalation resistance, measured or published at a standard test flow (around 85 L/min). Lower is easier. Where no figure exists, I map consistent wear testing onto the same bands. I keep this strict, the best masks sit around 3–4 mmH2O, so a low number really has to be earned.
About 5 mmH2O or less. You forget you have it on.
Roughly 5–9 mmH2O. Comfortable for all-day wear.
Roughly 9–15 mmH2O. Noticeable on exertion.
Roughly 15–25 mmH2O. Fine for short stints.
Above about 25 mmH2O, approaching the N95 limit of 35. A genuine struggle.
More subjective, and I treat it that way. I weigh pressure points, heat, fogging, strap comfort, and how it wears over a long stretch. Breathability is scored separately, so it isn’t double-counted here.
Protection and comfort relative to real cost, which for masks means cost per use rather than sticker price, so a reusable respirator and a box of disposables can be compared fairly.
I often test only one unit, in real-world conditions, though sometimes I get my hands on more than one. There are also things I can’t fully measure myself, and rather than dress those up in a number I’m open about them: where I lean on tiers, baselines, studies, and third-party results, I say so, and where I can’t stand behind something I leave it out.
Some judgements are also specific to me and can’t simply carry across to you. Fit is the obvious example: it’s personal, so I don’t score it, and if you depend on a respirator you should fit-test it yourself. Treat everything here as a transparent, relative guide from someone who puts the devices through their paces, not as absolute measurement. I version this methodology and re-test flagship devices when it changes, and every score on the site links back to this page. If a score ever looks wrong to you, get in touch and I’ll re-check it.