We Publishing your mistakes, which means we will sometimes get it wrong. When that happens the fix goes on a public corrections log. This is why we keep one, what is currently on it, and why we think a methodology page without a corrections log is closer to marketing than to method.
What is on ours right now
Three entries. None of them reached a published figure, and we log them anyway.
- Concurrency corrupted a full scan. Fetching robots.txt at 175 parallel requests made google.com return as unreachable. Nothing errored; rows simply came back empty. The entire run was discarded and rerun at 25 to 40 concurrent.
- Raw scores reported as percentages. Our nine checks total 110 points, not 100. The first analysis pass treated the raw total as a percentage, inflating every score by roughly ten percent. Caught when a page scored above 100.
- The sample was not a list of websites. The top of the Tranco list is dominated by DNS roots and CDN endpoints that serve no site and no robots.txt. Unfiltered, it would have produced the headline that most top sites have no robots.txt, which is false.
Why publish them at all
The obvious objection is that admitting errors undermines confidence in the data. We think the opposite, for a specific reason.
Anyone can publish a number. What makes a number usable is knowing how it was produced and what could have gone wrong with it. A corrections log is evidence that somebody was looking for the failure modes rather than hoping there were none.
It also creates a useful internal constraint. Knowing that a mistake will be published in detail, with the mechanism named, changes how carefully the analysis gets checked before release. That effect is worth more than the reputational cost.
The failure mode worth dwelling on
The concurrency problem is the one we would most want another researcher to read, because it produces confident wrong answers rather than obvious failures.
Nothing crashed. No exception was raised. No timeout was logged. The scan completed, produced a clean-looking dataset, and a meaningful share of rows were silently empty. Had one implausible row not been questioned — google.com appearing to have no reachable robots.txt — the study would have published.
That is the shape of the dangerous error in any scanning work: the tool reports success while the data is wrong. It is why we now include known-good control domains in every run and check them before analysing anything else.
What counts as a correction
The line has to be drawn somewhere, or the log becomes a changelog nobody reads.
- Logged — anything that would have changed a published number, or did
- Logged — discarded runs, scoring errors, sampling problems, figures we could not evidence
- Not logged — typos, wording changes, design updates, added context
If it does not change what a number says, it is an edit rather than a correction. Conflating the two makes the log useless, because a reader can no longer tell which entries matter.
How this connects to trust as a ranking concept
Google’s guidance on helpful content asks whether a page demonstrates first-hand expertise and whether it is trustworthy enough to act on. Those are hard things to assert and relatively easy to evidence.
- Show the method — how the data was collected, at what scale, with what limits
- Show the failures — what went wrong and what you did about it
- Ship the data — per-row results so anyone can check your arithmetic
- Name the author — a real person, accountable for the claim
None of those are ranking factors in a mechanical sense. All of them are what a reasonable reader — or a quality rater working to published guidelines — would look for before treating a claim as reliable.
What we would tell anyone publishing research
Three things, learned the expensive way.
- Include control cases. Known-good inputs whose expected result you can verify before trusting anything else in the run.
- Distrust clean data. A dataset with no anomalies is more often a broken collector than a tidy world.
- Publish the limits in the same document as the findings. Not in a footnote, and not later.
The full log is on our corrections page, and the method behind every study is documented on the methodology page, including sample construction and the weighting behind each check.
Sources and further reading
Build Greater Trust in Your Original Research
Publishing original research is not only about presenting strong findings. Readers, search engines, and AI systems also need to understand how the data was collected, which limitations affected the study, and how errors are handled when they are discovered.
A clear methodology page, public corrections log, named author, source documentation, and accessible per-row data can make research easier to verify and more credible to cite. These elements demonstrate that your conclusions come from a repeatable process rather than unsupported marketing claims.
MoxSEO helps businesses, agencies, publishers, and research-led brands strengthen the trust signals surrounding their original data. Our team can review your research methodology, sampling process, scoring model, control cases, source attribution, structured data, author information, and corrections policy to identify gaps that may weaken confidence in your findings.
Whether you are preparing a new industry study, publishing survey results, analysing website performance, or reviewing an existing research library, MoxSEO can help you create a clearer and more transparent framework. The goal is not to pretend that mistakes never happen. It is to show that errors are actively investigated, corrected, documented, and prevented from repeating.
A research consultation can also help you determine what should appear in a public corrections log, how to separate meaningful corrections from ordinary editorial updates, and how to present limitations without reducing the value of your findings.
Make your original research easier to inspect, understand, trust, and cite.
Contact MoxSEO for an Original Research and AI Visibility Audit
Frequently asked questions
Does publishing mistakes damage credibility?
It costs something short term and buys more back. A methodology page with no corrections log reads as either lucky or unexamined, and neither is reassuring.
What if someone finds an error in your data?
Tell us and we will check it. Every study ships per-row data specifically so that is possible, and external corrections get logged with the same detail as our own.
How do you decide what is significant enough to log?
Whether it would have changed a published number. That is the only test we apply, and it keeps the log short enough to stay readable.
Do you republish corrected figures?
Yes, and the correction stays on the log permanently. Removing it once the number is right would defeat the entire purpose.
Is a corrections log worth it for a normal business blog?
If you publish original data, yes. If you publish opinion and guidance, a clear last-updated date and a named author achieve most of the same thing.
Sakshi Kumari is an SEO Specialist at MoxSEO with expertise in keyword research, on-page SEO, content optimization, technical SEO, and link-building strategies. She focuses on improving organic visibility, strengthening website performance, and creating search-focused content that attracts relevant traffic. Through data-driven analysis and practical SEO execution, Sakshi helps businesses build stronger search presence and achieve sustainable digital growth.
