Original research
The Purifier Lab
Our own testing, with the method published before the result and the raw data published alongside it. Right now that means 5 open protocols and no completed runs — and this page says so rather than implying otherwise.
Important
Almost every purifier comparison published in India is a restatement of manufacturer specifications with a ranking imposed on top. That is not testing, and no amount of confident writing makes it testing.
This section is the alternative, built the slow way. A protocol is written and published first — including its controls, its instrument, and the specific things it cannot establish. Then it is run. Then every reading is published, whether or not the pattern is tidy.
The awkward consequence of doing it in that order is a period during which this page has methods and no results, which is where it currently sits. We would rather show that plainly than launch the section on the day the first favourable number arrives.
Open protocols
Published, not yet run. Read them, and tell us if the method is wrong — that is what publishing early is for.
- Protocol published — not yet runAir
How fast does an air purifier actually clear a 200 sq ft room?
Starting from a known elevated PM2.5 concentration, how much does a given air purifier reduce it in 5, 15, 30 and 60 minutes in a sealed 200 sq ft room — and how does that compare with the reduction its CADR rating implies?
- Protocol published — not yet runAir
What happens to an air purifier's performance as its filter loads?
How much does an air purifier's measured PM2.5 decay performance change between a new filter and the same filter after 3 and 6 months of continuous use in Indian conditions — and does the unit's own filter-life indicator track that change?
- Protocol published — not yet runWater
How much does RO rejection actually fall as a membrane ages?
Over twelve months on a real Indian domestic supply, how does a standard RO membrane's rejection rate change — and at what point does treated-water TDS cross the level the household bought the system to avoid?
- Protocol published — not yet runWater
How much water does a domestic RO system actually waste?
For a given domestic RO unit on a given feed water, how many litres of reject are produced per litre of product — and how does that ratio change with feed TDS, feed pressure and membrane age?
- Protocol published — not yet runAir
At what fan speed can you actually sleep — and what does that cost you?
For a given air purifier, what is the measured noise level and the measured PM2.5 decay rate at every fan speed — and how much clean air delivery is given up to reach a level someone can sleep beside?
How we test
Protocol before result
Every method is published in full before the first run. That removes the possibility of quietly amending it once a result is known, and it invites correction while correction is still cheap.
Raw data, not headlines
Published tests show every reading taken, including the ones that complicate the conclusion. A percentage with no series behind it is an assertion.
Controls, especially the boring one
A sealed room clears particulate by itself. Every air test records a no-purifier control run and subtracts it, which most published purifier testing does not, and which usually reduces the headline figure.
Limitations up front
Every protocol states what it cannot establish, published with the method rather than buried after the result.
Failures stay published
A run that went wrong is recorded with a note. Deleting inconvenient runs is the difference between testing and marketing.
No manufacturer involvement
Nobody pays for a test, chooses what is tested, or sees a result before publication.
Frequently asked questions
Why are some protocols published with no results?
Are you an accredited testing laboratory?
Do manufacturers pay for these tests?
How is this different from a CADR rating?
Can I see the raw data?
Nothing in this section is an accredited laboratory result, and none of it is represented as one. Where a manufacturer’s published figure and our measurement disagree, both will be shown with their respective methods, because the methods are usually why they differ.