Research hub
Corrections policy
The commitment
Bitcoin Data Guide will get things wrong. A source will be revised, a transcription will slip, a figure will be attributed to the wrong year, or a claim will turn out to rest on less than it appeared to. When that happens, the correction is made on the page and described, rather than applied silently. A reader who quoted the old figure deserves to know that it changed and why.
The standard the site holds itself to is simple: a reader should never be misled by something here, and should never have to guess whether a number is current. Where those two things come into conflict with making the writing look tidy, the writing loses. This page is the operational detail behind that promise.
How to report an error
Corrections are reported by email to hello@bitcoindataguide.com. That is the only channel, and it is monitored by the people who write and maintain the site rather than by an automated system. There is no ticketing queue and no support desk: a person reads what arrives.
A report that can be acted on quickly carries four things: the page you were reading, the figure or sentence you believe is wrong, what you believe the correct value or wording to be, and the source you are comparing it against. A correction that arrives with a citation can usually be checked and applied the same day. A message that says only that a number looks off is much harder to act on, because it gives no starting point for the check.
You do not need to be certain that the site is wrong in order to write in. A figure that looks inconsistent with another page, a source that has since been revised, or a claim that appears to overstate what the data supports are all worth reporting. The contact page sets out what else the address is monitored for.
How a report is reviewed and verified
Every report is checked against the primary source before anything is changed. The steps are the same whether the error was found by a reader or by the publisher.
1. Locate the claim and its source
The report is traced back to the dataset or sentence it concerns, and from there to the source recorded in the register on the data sources page. A claim with no traceable source is itself a defect, and is treated as one.
2. Re-fetch or re-check the source
Where the source is public and re-fetchable, it is pulled again and compared with what the site recorded. Where it is a protocol fact, it is checked against the block chain. Where it is a compiled historical series, it is compared with a second independent source.
3. Decide whether the site or the report is wrong
Not every report is a correction. Sometimes the site is right and the report rests on a different source, a different window or a different definition. In that case the reply explains the discrepancy, and the page is clarified if the confusion is reasonable.
4. Record the outcome
A confirmed error is logged with the date it was reported, the date it was verified, the source used to verify it, and the version of the figure before and after. An unconfirmed report is logged with the reason it was not actioned.
Verification is deliberately conservative. Where two sources disagree and neither can be shown to be authoritative, the site records the disagreement rather than picking a winner or averaging the two together. A figure that cannot be verified against a source is removed rather than left standing with a caveat.
How corrections are published and labelled
A correction to a data table moves the vintage label on the affected table forward, so the version a reader is looking at is always identifiable. The table itself carries the source and the period it covers, and the vintage range is the record of when it was last republished. A reader who needs to know whether a figure has changed can compare the vintage on the page with the one they captured earlier.
A correction to prose is handled according to how material it is. A wrong percentage, a misattributed date or a claim that overstated the evidence is corrected in the text and described where the error was material to the argument, so a reader who quoted the old wording can see what changed. A typo, a broken link or a formatting slip is fixed quietly, because a correction notice for a misspelling would bury the notices that matter.
Where a correction changes the conclusion a page draws, the page says so explicitly rather than editing the conclusion into place. The reasoning that led to the original conclusion is left visible, with a note on what the corrected figure shows instead. A reader should be able to see not only what the site now says, but what it said before and why it changed.
How material updates are dated
Every dataset on the site carries a vintage: the period it covers and the point at which it was last republished. A material update — a revised source, a corrected figure, an added year — moves that vintage forward and is dated. The date is the date of the update, not the date of the original publication, so a reader can tell at a glance how current the material is.
The live series are the exception, and the site is explicit about it. The BTC/USD spot price and the daily candles are fetched from the Coinbase Exchange public API, cached for a few minutes and re-fetched on demand. They are readings rather than settled records, they change between visits, and each one carries the timestamp of the moment it was fetched. They are not revised, because they are not stored as history: each fetch replaces the cached value and the previous reading is gone. If you need a permanent record of a particular day's price, take it from the curated daily candles rather than the spot panel, and note the date you captured it.
The data sources and methodology page sets out the vintage of every dataset in full, along with the periods where the compiled record is genuinely uncertain.
No silent alteration of published analysis
The site does not quietly rewrite published analysis. Once a page has drawn a conclusion from the record, a later change to that conclusion is made visibly: the page states that it has been revised, what the earlier version said, and what the corrected material shows. The earlier reasoning is not deleted, because a reader who followed it deserves to see where it led and why it no longer holds.
This applies to the site's own analysis as much as to a reader's report. When the publisher revisits a page and finds that a conclusion no longer follows from the data — because a source was revised, because a longer window is now available, or because the original reasoning was simply wrong — the revision is described on the page rather than applied in place. The record of what the site said before is part of the record.
The editorial standards page sets out the sourcing and review standards behind this policy, and the disclaimer states the educational purpose of the site and the limits of what it offers.
Continue reading
Related reading
- Research HubEvery dataset on the site, with methodology and provenance.
- Altcoin ResearchAltcoins measured against Bitcoin: design intent, consensus, execution, scaling and market structure.
- The ETH-BTC Correlation RecordHow the correlation is measured, how it behaves across windows, and where it breaks down.
- The ETH/BTC RatioWhat the ratio measures, how to read its trend, and why it is not a forecast.