Our methodology
The long version, for people who want it
How we choose what to buy, how many units we test, how a score is built, and what happens when a result is challenged. Written so that another lab could check our work.
Choosing what to buy
We prioritise the products people are actually taking. That means the brand's flagship item first, the format most commonly sold, and the dose most commonly bought. Where a brand sells across both Canada and the United States we try to buy in both, because supply chains and manufacturing sites often differ.
We do not accept product submissions in place of buying. A brand can ask us to test them, and that moves them up the queue, but we still buy the product ourselves at retail.
Sampling
- Units per lot
- Multiple units are drawn from the same purchased lot so that unit to unit variation is measurable rather than averaged into a single figure.
- Chain of custody
- Every sample is logged from arrival through to dispatch, with the lot identifier recorded where the brand provides one.
- Blinding
- Samples are submitted to the laboratory without the brand relationship attached to the sample identifier.
- Retention
- Where quantity allows, part of the purchase is retained unopened so a disputed result can be checked against material from the same lot.
How a score is built
Scores run out of 100 and are built from four components. We publish the component figures alongside the total, because a single number hides more than it explains.
Dose accuracy
How close the measured content is to the labelled content. Both directions count. Overdosing a customer is not better than underdosing them.
Consistency
How tightly units from the same lot cluster together. Wide spread costs a brand heavily even when the average looks correct.
Purity
Contaminant screening. Anything above a safety threshold is disqualifying regardless of how the rest of the product performs.
Transparency
Whether the brand publishes lot numbers, ingredient detail, and its own testing, and whether those claims match what we measured.
Scores are not live yet. No brand on this site currently carries a published score. Samples are with the labs and results go up as they come back, in full, whatever they say.
Thresholds
A contaminant result above the relevant safety limit fails the product outright and no total score is issued. We would rather publish a clear failure than a number that lets a contaminated product look mid table.
Retesting
A brand can request a retest after a failure. We buy again, at retail, from a later lot, and we publish both results side by side with their dates. The old result stays up. A brand that fixed a problem deserves to show that it fixed a problem, and the record of the fix is part of the story.
Corrections
If we publish something wrong we correct it on the same page, say what changed, and date the change. We do not remove a result because a brand asked us to, and we do not soften wording in exchange for anything.
Conflicts of interest
Nobody involved in producing a result holds a financial stake in the brand being tested, and nobody is compensated based on what a result says. Where any connection exists between a person working with us and a brand under review, that person is removed from the work on that brand and the connection is disclosed on the result.