Methodology

Where the product information comes from

We read the public Amazon listing — the same page you're looking at when you shop.

The product title, photos, price, star rating, brand, and category all come straight from the item's public Amazon page. That's the same information anyone can see; we gather it up and add a sustainability layer on top.

We don't touch your account, your cart, your orders, or anything private. Nothing shown here is based on your personal shopping history.

How we figure out which company actually makes a product

The name on a listing isn't always the company responsible — so we trace it back to the one that is.

Sustainability practices are set at the company level, not the individual-product level. But a listing might show a random storefront name, a marketplace seller, or a small label owned by a much larger parent company. To score the right organization, three things happen:

• Ignore generic marketplace names. Sellers and storefronts that aren't the actual maker (including plain "Amazon" listings) are set aside — they don't tell us who's responsible for how something is made.
• Identify the real brand. The genuine brand behind the product is worked out from the listing itself.
• Trace sub-brands to their parent. Many familiar brands are owned by a larger company that sets the real environmental and labor standards. When that's the case, the smaller brand inherits its parent's record.

A quick example

Say a smaller label is wholly owned by a large parent corporation. On its own the label may publish very little, but the practices that matter — factories, sourcing, labor standards — are its parent's. So we attach the parent's verified record to it, rather than pretending the sub-brand exists in a vacuum. If we can't confidently match a product to a real company, we'd rather leave it unrated than guess.

How a company's practices get researched

A purpose-built research agent goes and reads the public record for us — and it's only allowed to read from a short list of places, with a link attached to everything it brings back.

A company's environmental and labor record is scattered: some of it on the company's own site, some in certifier directories, some in reporting and regulatory filings. Gathering that by hand for every brand doesn't scale, so we built a research agent to do the reading. It answers the same fixed set of questions about every brand, in the same order — so a small label and a multinational get asked exactly the same things.

It's pointed at three kinds of sources, and it's built not to wander outside them:

• The company's own public pages — sustainability, impact, materials, factories, supplier codes of conduct.
• Certifier directories — B Lab, GOTS, Fair Trade, Climate Neutral, 1% for the Planet and the rest, where the certifier itself publishes who holds what.
• Credible public reporting — news outlets, regulatory bodies, court filings, and advocacy reporting with a named reporter behind it. Anonymous forum and social posts are ruled out explicitly; they don't count as a source here.

The rule that does the most work: every claim has to arrive with a direct quote and a working link. A finding with no quote, or with no source we can open, is dropped before it ever reaches a page — it isn't shown as "probably true," it simply doesn't count. The same goes for vague marketing language: "We care deeply about our planet" is recorded as vague and earns nothing, while a published, specific commitment with a citable source does.

Does the agent decide the rating?

No. It researches; it doesn't judge. Its job ends at a structured set of answers with evidence attached. The rating is then calculated from those answers by fixed rules applied identically to every brand — the same certification is worth the same amount everywhere, and the same severity of controversy costs the same everywhere. That separation is deliberate: it means a rating can be re-derived and checked, rather than resting on a model's opinion of a company.

What we check by hand before it counts

The two things that move a rating most — certifications and controversies — are the two things a person reads the source for, not just a machine.

Any system that reads the web can get things wrong. It can misread a page, credit a parent company's record to a small subsidiary, or report a certification a brand genuinely held five years ago and has since let lapse. So the claims carrying the most weight aren't taken on trust.

First, an automatic cross-check. Where a certifier publishes its own searchable registry — B Corp, USDA Organic, ENERGY STAR, EPA Safer Choice, Leaping Bunny, 1% for the Planet — every certification claim is checked against it. A claim the registry actively contradicts, such as a certification that has lapsed or been revoked, is taken off the page rather than displayed.

Then, human review of what's left. Two queues, worked in order of how much the answer would change:

• Certification claims no registry could confirm. Each one gets looked up in the certifier's own database by hand. Confirmed there, it counts in full. Backed only by the brand's own website, it stays visible and clearly labelled as the brand's claim, but carries far less weight. Not found at all, the claim is stripped out.
• Controversies serious enough to affect a rating. The source gets opened and read. From there a finding is confirmed as reported; downgraded in severity if it was overstated (it stays on the page, it just stops driving the rating); marked resolved if the company has since addressed it; re-labelled as a parent company's or supplier's conduct rather than this brand's (it stays on the page, labelled, and scores nothing — you're still buying from them); or removed entirely if the source doesn't actually support it.

To be straight with you about scope: this is ongoing work, not a finished sweep. The queues are deliberately ordered by consequence, so the claims that would change a brand's rating get read first, and lower-stakes ones wait their turn. Minor findings that can't affect a rating are displayed as-is, as reported by the source.

Product details come straight from the listing

"BPA-free," "organic," "plastic-free" and the other tags on a product card are read off the listing itself. They're the seller's claims about the item, and unlike brand ratings they aren't reviewed one by one.

A brand rating and a product tag are two different kinds of information, held to two different standards, and it's worth knowing which you're looking at.

Product tags come from the listing's own metadata — the title, the bullet points, the specification table, and structured fields supplied with the listing data, such as Amazon's own Climate Pledge Friendly designation. Those get matched and shown on the card, and they're what the category filters run on when you narrow by an attribute.

Nobody hand-verifies them. If a listing says "BPA-free" and it isn't, the card will repeat it. And they're kept separate from the rating in both directions: a product tag never moves a company's rating up or down, and a company's rating never adds a tag to a product.

How fresh the information is

Prices are refreshed every day, straight from Amazon — and when we're not sure a price is current, the page says so instead of guessing.

Updated prices are pulled from Amazon daily, so what you see stays close to the live number. Sustainability information updates as new public reports, certifications, and records become available.

If a daily price refresh is ever missed, you don't get shown a stale figure. It's replaced with a simple "See price on Amazon" link so you can check the real, current price before you decide.

How we make money — and why it never touches the scores

Money and scores are kept completely separate. A brand can't pay for a better score, and we're never paid to rank one product above another.

Leif List is free to use. We earn in two disclosed ways:

• Amazon Associates. When you buy through an Amazon link on this site, we may earn a small commission — at no extra cost to you. (You'll see this noted at the bottom of every page.)
• Partner brands. A handful of brands are affiliate partners, so some products link to the brand's own store instead of Amazon. If you buy there, we may earn a small commission — same as Amazon, at no extra cost to you.

Here's the firewall that matters: none of that money influences a score or a ranking. Scores come only from the sustainability data. The order products appear in comes from that score plus the values you choose to prioritize — never from who pays us. A partner brand is scored and ranked exactly like every other brand; being a partner only changes where the buy link points.