Skip to content

Editorial policy

How we source our numbers.

We publish 18 guides and 63 glossary entries about money that other people are responsible for. That deserves a stated standard rather than an implied one.

Every number is either computed or cited

Worked examples in our guides are exactly that — arithmetic you can reproduce from the inputs shown. We label them as examples and show every input, so you can check them or substitute your own numbers.

Where we state a fact about the world rather than a calculation — an industry range, a government price series, a competitor's pricing — it carries a source or it doesn't get published. We do not cite statistics we cannot trace to a primary source, and we don't repeat widely-quoted figures whose origin we can't find.

Ranges are presented as ranges

Benchmarks like "food cost runs 28–35% in full service" are planning ranges drawn from common operating practice, not measured research findings. We say so, and we consistently push readers toward deriving their own target from their own rent and labour rather than adopting a borrowed number.

We use AI, and a human is accountable for every page

We build AI software; it would be strange to pretend we don't use it. AI assists with drafting, structuring and checking our guides.

What that does not change: every published page is reviewed by a person who is accountable for it, every calculation is verified, and no page ships with a statistic that hasn't been traced. AI assistance is a drafting tool, not a source. Where a claim couldn't be verified, the claim was removed rather than softened.

Competitor facts are dated and sourced from the competitor

Every fact on a comparison page comes from the competitor's own public pages, and carries the date we checked it. Pricing changes; a stale comparison is both a credibility problem and a fairness problem.

Where a competitor's pricing or capabilities aren't publicly verifiable, we don't publish a comparison page for them at all. We have skipped comparison pages for this reason and will again.

We don't make claims about our own customers that we can't document

No customer counts, no "trusted by" claims, no aggregate savings figures until they're documented and the customers have agreed to it. Product-truth math only: if we say a $2-per-plate gap at 100 covers a week is $200 a week, that's arithmetic, and it's labelled as arithmetic.

Corrections are made in place, and dated

When we get something wrong we fix the page and update its modified date rather than quietly editing history. If a correction materially changes the advice, we say what changed.

Found an error? Email support@marji.co. We would genuinely rather be corrected than be wrong in public.

Nothing publishes that an operator wouldn't find useful

Volume is part of our strategy; filler is not. Our own build process fails the site build if a guide or glossary entry lacks a formula, a worked example or a data table, or falls below a minimum length. That check exists because the temptation to publish thin pages at scale is real and we'd rather be stopped by a test than by our own restraint.

Where the numbers come from

Commodity prices on this site come from US Bureau of Labor Statistics average-price series, retrieved through public data APIs. The methodology, series identifiers and revision policy are documented on the index methodology page, and the underlying data is downloadable so you can check our arithmetic yourself.