Methodology

What we do, how often we do it, and what we refuse to do. Written so you can hold us to it.

Where every fact comes from

Load-bearing facts live in typed data files, not in prose. A price, a specification or a policy is written once and read by every page that shows it. Hand-maintained copies of the same fact drift, and on a health site drift is not an inconvenience, it is a false claim.

Every price renders the date it was verified, right next to the number. If you see a price here without a date, that is a bug and we want to know.

How we grade evidence

Every source we cite carries a tier, shown next to the citation. This is the part most sites leave out, and in this field it changes the meaning of almost everything.

• Clinical guideline. A professional body’s formal recommendation. Note that a guideline can still rate its own evidence as very low quality, and the SIBO guideline does exactly that for five of six recommendations.
• Peer-reviewed study. Published in a journal after review. Not a guarantee of correctness, and sample sizes in this field are often small.
• Preprint, not peer reviewed. Posted before peer review. May never be published. We say so every time.
• Conference abstract. A conference summary, usually a few hundred words with no full methods. A surprising share of what circulates in this niche rests on these.
• Patient registry. A collection of cases contributed by practitioners. Useful for describing what clinicians see, not for establishing how common anything is.
• Community report. Something patients reported to each other. We use it to know what to write about. It is not evidence about the body.

Sources at the weaker tiers must carry a stated caveat before they can be cited at all. That rule is enforced by the build, not by memory: a preprint without its limitation fails compilation.

How we choose what to write about

Not by keyword tools. We count how often patients discuss two conditions together across 7.2 million comments, and write about the pairs above 2,000 co-mentions. We publish all 36 measured pairs including the ones below the line, and we set the threshold after seeing the distribution rather than before, because a threshold chosen in advance tends to admit exactly the pages you already wanted to write.

How we handle things that are not settled

We state uncertainty as a fact rather than hedging around it. Where specialists actively disagree, we say they disagree and name the positions. Where a test has no validated threshold, we say so on the page that discusses the test. We would rather be less useful than falsely confident, because confident wrong information in this field costs people money and years.

What this site will never do

• No protocols. No dosing. Not for any substance, not ever.
• No disease-treatment claims, including implied ones. Naming a condition in a heading above a product block makes a claim about that product even if the sentence does not.
• No fabricated credentials, reviewers, testimonials, or first-hand experience. If we have not used something, the page does not imply we have.
• No presenting a serum DAO test as diagnostic, and no presenting a personal fermentation tracker as a SIBO test.
• No review or rating structured data. We have no real reviews, so we do not emit markup claiming we do.
• No affiliate links on about, this page, the corpus, or disclosures. That restraint is the point of those pages.

How we make money

Currently we do not. No affiliate programme on this site is live, no commercial link renders anywhere, and no partner has approved us. When that changes we will name every relationship on the disclosures page before the first link goes live, not after.

Corrections

We log them publicly rather than quietly editing. The correction log.