EDITORIAL PROCESS

Reference Checks says which version of a record is worth your time, and shows the evidence. This page is how that gets done, and how to tell me when I get it wrong.

Who makes it

I’m Craig Jenest. I write the editions, make the calls and carry the mistakes.

The research runs through a framework I built for this work called the Pressing Council. It assembles release data, mastering chains, personnel credits, plant identification and provenance into a structured dossier for each edition. That dossier is the input. The edition is what I write from it.

The research workflow uses AI tooling for source aggregation, cross-referencing across Discogs, MusicBrainz, label catalogs, engineer interviews and archival press, and for organizing release identifiers into something comparable. Assembly is tooled. Verification happens by hand against primary sources before anything publishes. The editorial decisions, what qualifies as reference-level, which version takes the top pick, what comes out of a draft, are mine.

The chain I listen on is named on the About page. It is calibration context and it stays out of every verdict. The best pressing is the best pressing regardless of what you play it on.

I recommend the version that best carries the recording, judged on provenance, mastering chain and measurement. Where an artist made or endorsed a different version and that preference still stands, I name it, say what it costs against the pick, and leave the choice to you. Where nobody’s preference is on the record, which is most of the catalog before about 1970, I say nothing about intent I can’t source.

The eight gates

Every edition passes these before publish. Gates 1 through 4 also clear before an edition goes to anyone outside the publication, including a preview sent to an artist, producer or label. A preview is held to the publish standard, because the person reading it can’t tell the difference and shouldn’t have to. They’re here so that anyone with a correction, a primary source or a flag knows where it lands.

1. Personnel, session and anecdote claims. Any claim naming a specific person on a specific track needs two independent primary sources before it goes into prose. Primary means liner notes, official label or platform credits, AllMusic, Discogs personnel fields, or a first-person statement from the producer or musician by interview, podcast, verified post or direct correspondence. Wikipedia counts as one of the two only when it transcribes liner-note credits directly. Aggregator sites don’t count. Neither does plausibility based on who someone usually worked with.

Biographical and role claims sit inside this gate too. When somebody joined a label, which campaigns they ran, which releases they were around for: same two-source burden. This class of claim is the one most likely to arrive as ornament, a subordinate clause added for rhythm that asserts a fact nobody sourced. So the rule reads: every descriptive clause attached to a named person is a claim, and it’s verified or it’s cut. A sentence isn’t exempt from the gates because it wasn’t meant as reporting.

Direct quotations work differently. A pull quote needs one reliable published primary source, transcribed word for word, attributed correctly, with enough context that the meaning survives. A second source is required when the quote is contested or extraordinary, or when it also asserts a personnel fact, which puts it back under the two-source rule. Fabricated quotes, paraphrase dressed as quotation and single-blog-sourced quotes never qualify.

2. Prices. Every price in a verdict table traces to a live page lookup or a directly verified visible price at the time of publish. Search-result summaries are not verification. What a platform usually charges is not evidence of what it charges today.

3. Search discipline. Before I say a primary source can’t be found, I exhaust the search: site-restricted queries, URL pattern guessing for canonical archives, name-variant testing. AI-generated summaries are research leads that require primary-source confirmation, never the confirmation itself.

4. Voice and style. The editorial voice is the same for every reader, free or paid. A pre-flight audit catches machine phrasing, hollow warmth, consultant vocabulary and internal jargon before a draft reaches anyone.

5. Visual system. Locked palette, locked canvas dimensions, retina export. Graphic headings come out when the article header already labels the section.

6. Visual redundancy. Candidate graphics get checked against each other and against the prose. The asset carrying a structural insight earns the embed. The one restating something already covered gets cut.

7. Visual data anchoring. Any graphic representing real geography, real measurements or real chain-of-custody data is built from the actual data. Freehand stylization doesn’t run.

8. Producer and musician preview. Where the relationship allows, I send a private preview to the album’s producer or a named session musician before publish. Gates 1 through 4 clear before it goes out, and this gate never stands in for them. It’s the most valuable check available precisely because it runs last, against work that’s already been verified. The first Boxer edition caught a misattribution that way, before anyone read it.

Where evidence is thin

Where the evidence is partial, the edition says so. A claim is labeled as inference or it comes out. “The lacquer was cut by X” and “the lacquer was reportedly cut by X” are different sentences and Reference Checks writes the one that’s true.

Corrections

When new evidence changes a published claim, I update the edition, note the change at the bottom of the post, and log it on the Corrections page with the date and whoever flagged it.

Corrections are public and they stay public. A publication that has never printed one is either lucky or not looking.

If you have something

A primary source I should see. A name to re-verify. An anecdote I got wrong. A measurement I read incorrectly. All welcome, and credited unless you’d rather not be.

Reply to any edition, comment on it, or write to craig@twelvebar.studio. Flags run through the relevant gate and I’ll come back with what I find, including when the answer is that the original claim holds.

If you’re a label, engineer, archivist or other primary source disputing a factual claim, same address. Documented corrections get reviewed and the edition gets updated when the evidence warrants.

Feedback about whether a recommendation matched what you heard on your copy is welcome at the same address. I read all of it. It isn’t treated as a factual dispute, because individual listening experience doesn’t trigger a correction.

What this covers, and what it doesn’t

Reference Checks evaluates editions, not individual copies. The work identifies which mastering chain, which pressing, which transfer is worth a serious listener’s time. It doesn’t inspect the specific copy you buy, doesn’t warrant any seller’s grading and can’t promise a given pressing arrives clean. Sourcing risk is yours.

This is editorial commentary on recordings. It’s informational, it isn’t investment or financial advice, and it makes no claim about future market value, collectibility or resale.

Contact

Editorial: craig@twelvebar.studio

Reference Checks is a Twelve Bar Studio production.