How we compare

Run and checked by Dan, a registered nurse who trains and uses these supplements. Every figure names its source, and links to it where the source is public. About this site. Not medical advice.

Cost per dose

Bottle price is close to meaningless in this market. A $12 magnesium bottle that needs four capsules to reach a real dose costs more per dose than a $20 bottle that needs one. So we convert everything to a single comparison unit per category and divide.

The comparison is on the active substance as the panel prints it, not on the weight of the compound carrying it. A magnesium panel states elemental magnesium and that is the figure used; where a label prints only a compound weight, the product is listed and not priced per dose, because the amount cannot be read from it. A fish oil panel that prints total oil and no EPA or DHA is treated the same way.

“If you just want one answer”

Most people do not want to read a table, so where one product clears every bar below, the category names it. Where a family is compared within labeled strength bands, it names one per band and none across them. And where nothing clears every bar, it says so and names what stopped the closest products, which is more useful than a winner chosen by a lower standard. It is a rule, not a score, and the rule is published so you can disagree with it. There are no weightings inside it to argue about.

A product has to clear four factual bars.

1. An independent certification we could confirm applies to what is in the package. From a certifier that names packages, that means a listing naming this one. From a certifier that tests batch by batch instead, it means a current listing for the product, which is all such a program publishes. USP Verified names a product and a strength and puts its mark on the label, so for USP it means both at once: the product in USP's directory, and the mark on this package's own label.
2. No recall, warning letter or failed test in the sources we checked.
3. A price its retailer let us read recently, inside that retailer’s own cache limit, which each product page names. A price older than that is shown as a last known price and cannot win.
4. No more than a year’s supply, counted as 365 of the label’s own servings, which is a year at one serving a day. A pick is a first purchase, and a 5 kg bag is not one.

Among everything that clears all four and has a delivery cost we actually know, the cheapest per delivered dose wins. Many brands publish only a free-delivery threshold and no rate below it, so in some categories no certified product has a knowable delivery cost. There the pick is the cheapest certified option priced before delivery. It is labeled “before delivery”, and its card names the figure at which that retailer’s delivery becomes free. That is the whole calculation. A clean record here means we found nothing against the product, not that nothing exists. See below.

What “a listing naming this one” looks like in practice: every package NSF names, with its certificate number, beside the packages whose listing names a different size. Where we recorded the listing’s own package text it is quoted there too, and that page says on how many rows.

Underneath it we list the cheaper products that fail exactly one bar, and name which one. That shows a different trade-off each time rather than the same one twice. A certified product that fails only because its retailer will not publish a delivery rate is a very different proposition from an uncertified one, and you may well judge those differently to us. That is the point of showing them.

One wrinkle in the delivery test: a product a penny short of a free-delivery threshold is not a worse buy, it is a cart you have not finished. So when a listing is within 5% of its own price below the threshold, we treat its delivered cost as the threshold, because you can add any small item and get free delivery, and we show the gap on the card. Beyond 5%, the shortfall is a real second purchase and the delivered cost stays unknown. Five percent is a published number you can disagree with, not a weighting buried in a score.

Further out, we still flag cheaper products that fail only because delivery is not free, and we name the gap, rather than letting the arithmetic quietly rule out a product you might rightly prefer.

Delivery

A price without delivery is not a price. A cheaper tub from a retailer whose free-delivery threshold is $59 can cost more than a more expensive one from a retailer whose threshold is $25, and a comparison that ignores this quietly gets the ranking backwards.

Where a retailer publishes its rate, or the order qualifies for free delivery outright, or by adding a small item when the shortfall is within 5% of the item’s own price, we include it. Several instead calculate delivery at checkout from weight and destination. There we say the cost applies, say we do not know it, and refuse to invent a figure. Such a product can still appear in the table, but it cannot hold a “cheapest” badge on a price that is missing a charge you will actually pay. We also show how much more you would need to spend to cross the free-delivery threshold, because that is usually the useful thing to know.

When a product is excluded from a “cheapest” badge for that reason and might still be the better buy, we name it and say so, rather than letting it disappear. The cheapest certified one gets its own card at the top, labeled “cheapest before delivery”, next to the pick, with the gap to free delivery in dollars, so a cheap verified product is a choice you can see and not a row buried mid-table. When no certified product in the category has a knowable delivery cost, that product is the pick itself, and it says “before delivery” wherever it appears.

Pill burden

We also show how many capsules, softgels or scoops it takes to reach the comparison unit, and flag products where that is not a whole number. We do not fold this into the price. Rounding up to a whole capsule changes the amount as well as the price, so charging a product for it would penalize concentrated products for being concentrated. It is a caveat you can see, not an adjustment we make on your behalf.

The comparison unit is not a recommendation

Each category has one comparison unit, chosen so that products of different strengths land on the same scale. It is a yardstick for price. It is not advice about what you should take, and we do not tailor it to you. Every comparison unit is published with the reference we took it from:

Vitamin D3, Vitamin B12 and Zinc are sold in strengths that differ by tenfold and more, so there the comparison unit prices a product only against others at the strength its label prints, never against a stronger or weaker one.

Verification levels

We do not publish a quality score, because we cannot defend the weightings such a score would require. What we can defend is a statement about which documents exist for a product and who produced them. Each level below is checkable by clicking through to the source.

Recalls and warning letters affecting the product you are looking at are shown as a warning. Actions affecting a brand's other products are listed separately under brand history, without a warning badge. Many large companies have had something recalled eventually, and badging those would penalize size rather than sloppiness.

A Tier D product is not a bad product, and it does not mean nobody looked. We looked: of the 385 products we track, 235 appear in no certifier database we searched. These programs certify what a company submits to them, so an absence records what was submitted rather than anything found in the product. Recalls, warning letters and failed tests are listed separately and never folded into the level. Where we list none, that means none has reached us, not that we have run an exhaustive search and cleared the product.

Money

Some links on this site earn a commission. Every link that does says so at the link itself. No commission, of any size, changes what is listed or how it ranks.

No commission, of any size, affects which products we list, how they rank, what verification level they receive, or what we write about them. Rankings are computed from the data by the same code for every product. If we ever accept paid placement it will be visually separate from the comparison and labeled as advertising.

Four separate decisions are worth pulling apart, because they are not equally protected. Which product ranks first, and which retailer link we send you to, are never influenced by what we earn, the Buy button points at the lowest price we have recorded even when that retailer pays us nothing, and every link is labeled so you can check. Which retailers we track, and which categories we build first, are influenced by commercial reality, because we have finite time. Those affect what the site covers, not what it tells you.

The failure we are most alert to is bias by omission: quietly tracking only the retailers that pay, so that every page is technically true while the site systematically shows worse prices. The defense is that we record every retailer price we can find, including retailers with no affiliate relationship at all, and a check runs on every build confirming that each Buy button points at the cheapest listing on record.

What we will not do

• Infer a fact we have not read from a source. Unknown stays Unknown.
• Treat absence of contaminant data as evidence a product is clean.
• Tell you what to take for a medical condition.
• Reprint a manufacturer's marketing claim as though it were a test result.