A site that asks institutions to show their working has to show its own.

• Corrections are made in public. The original text is struck rather than deleted. Anyone who read the wrong version can see exactly what they read and exactly what changed.
• The correction is credited. Whoever supplied it is named, unless they ask not to be. Institutions that correct the record are named as a matter of course, and being right in public is worth something.
• A qualifying retrievable source, captured and tested against the published method, can move a row. Not a rebuttal, not a statement, not a phone call. And retrievable is necessary but not sufficient: a document that a third party can find still has to pass the scope, unit and reconstructibility tests set out on the Method page. Ethiopia's National Bank annual report is retrievable and has not yet moved a row for exactly that reason.
• A challenge triggers re-adjudication, not an automatic change. The row is re-tested against the controlling rule book. It may move, it may stay, and it may move to a different state than the one proposed. What is promised is that the challenge is answered in public, not that it succeeds.
• Every public state change is logged with its full provenance: the old state, the new state, the effective date, the rule book revision under which it was decided, and the decisive evidence. A reader should be able to tell not only that a figure changed but which rule and which document changed it.
• Internal adjudication changes are not corrections. A determination reached, revised, or withheld before anything was published is ordinary working process. Recording it here would inflate this log with activity that never misled anyone.
• Substantive corrections are logged here. Typographical fixes are not, and pretending otherwise would just bury the real ones.
• Failure is reported. If a collection effort produces nothing usable, that is published, with the reasons. A monitor that only reports its successes is not a monitor.

No corrections have been made. This site launched recently and has had few opportunities to be wrong. That will change, and when it does the record of it will be here rather than quietly absent.

If you have found an error, or a first-party publication that could change a row of the coverage index or an entry in The Record, send the source to wondwossen.mezlekia@coffeemonitor.com. Coffee Monitor will capture it, test it under the controlling rule book and publish the outcome. If it changes the public record, the change and the source are credited and logged. If it does not qualify, the reason is recorded.

What a state change records

The coverage index and the Observatory carry dated states rather than prose, so a change to one is a change to data and is logged as such. Every public state change records the row, the state it held before, the state it holds now, the date the change takes effect, the rule book revision it was decided under, and the decisive evidence.

Internal adjudications are not corrections. A determination reached but not yet published has changed nothing a reader relied on, and recording it here would inflate this log with work rather than with error. A correction is logged when a public state, figure or claim that someone could have acted on turns out to have been wrong.