How we calculate

VERSION 1.1 · OCTOBER 2026

100 points, four categories

Lab reports make up the largest part. Documentation and how recent the reports are add to it.

Lab results · 50

How well the samples perform. Each sample is scored in four parts:

• Purity (HPLC), 35 %: full points from 99 %, 90 % from 98 %, 75 % from 97 %, half from 95 %, a fifth from 90 %, nothing below.
• Content, 25 %: up to 20 % more than stated or up to 5 % less gets full points. Up to 10 % less still gets 85 %. After that it drops fast: up to 15 % less half, up to 20 % less a fifth. For overfill beyond 20 %: up to 30 % gets 60 %, up to 50 % gets 30 %, beyond that nothing.
• Identity (LC-MS), 25 %: points only if mass spectrometry confirms it is the right substance.
• Endotoxins, 15 %: points only if tested with a result below the lab's limit. If the lab names none, the limit is 5 EU per mg of substance.

What was not measured earns no points. Without LC-MS and an endotoxin test, 100 points are not possible. If a sample contains a different substance than stated, this category gets 0 points.

Lab reports · 25

How many of the shop's products are tested. Each product adds points, from 8 products you get the full 25.

Documentation · 15

5 points each if the reports can be verified with the lab, name a batch number and are publicly available. We count the share of reports that meet this. A batch number that appears for different products does not count and costs extra points (see deductions).

Freshness · 10

Newest report at most 90 days old: 10 points, up to 180 days: 6, older: 3.

When there is a rating

• We only count reports from Janoshik and Analiza. Reports from other labs earn no points.
• Tests the shop commissioned itself count too. Every report must be verifiable with the lab, though.
• A report counts for at most twelve months. For each product the newest one counts.
• A rating needs at least one valid report.

The legal notice does not cost points. If it only lists a virtual office, looks like a shell company or is made up, we show a note on the shop's page.

If a shop presents a forged report or someone else's report as its own, we set the rating to 0 and show a warning. Such a shop always ranks last, no matter how good the rest is.

Deductions

• Errors in a lab report, for example a sample photo or sample name from another order or an impossible date: 2 points per report, at most 10. This also applies if the lab made the mistake. Once the shop has the report corrected, the deduction goes away.
• Wrong batch number: the same number for different products, or a batch whose results do not match. Up to 10 points, by the share of products affected. A wrong batch number is worse than none.

How reliable is a rating?

High: at least 8 tested products. Medium: 3 to 7. Low: fewer. For every rating we store the date, the methodology version and which reports went into it.

If we change the weighting, the methodology gets a new version number. Old ratings stay traceable. Nobody changes a rating by hand.