The scan and the health score

A scan walks your Freshdesk knowledge base through the API, in every portal language, runs eleven checks over what it read, and stores one row per finding. It writes nothing. A read-only key is all it ever needs.

This page explains what the checks are, how the number at the top of the page is worked out, and — just as important — the cases where a scan deliberately refuses to tell you something.

Full scans and daily top-ups

A full scan reads every category, folder and article in every language your portal serves. It is the only kind of scan that produces totals and a score, and the only one that can conclude an article has been deleted — because it is the only one that has seen everything.

A daily top-up, on plans that include it, re-reads only the articles Freshdesk says changed since the last scan. It refreshes the findings of those articles — including marking them fixed — and leaves every other article's findings exactly as the last full scan left them. It has no score of its own; the dashboard keeps the last full scan's numbers.

When the two are due together, the full scan wins, and it counts as that day's top-up too.

The eleven checks

A few of these are fussier than they look, on purpose. "Old year" ignores copyright lines, version numbers, release notes, ISO standards, dates in URLs, "since 2019", "Windows Server 2019" and about twenty other shapes that were false positives when we ran the rules over twenty real help centers. "Near-duplicate" refuses to flag an article and its own translation, titles that differ by a number, or two pages that merely share a title.

How the score is worked out

For every published article version, each check that fired costs points by severity: high 3, medium 2, low 1. Two findings of the same kind on one article are the most that kind can cost — 40 broken links in one article still count as two — and one article can never cost more than 6 points in total, so one dreadful article cannot sink the whole help center.

The score is the average article's penalty measured against that 6-point ceiling, subtracted from 100. An article at the ceiling scores 0; a knowledge base with no findings scores 100.

Three details follow from that:

• Drafts and folders never move the score. A draft is not a page a reader can reach, and an empty folder belongs to no article. Both are still in the list.
• Translations count as articles. Every published version in every language is a page somebody reads, so each one is scored.
• The score is always measured over a set of checks, and that set is stored with it. Two scores are only comparable over the same set.

That last point is why a trend sometimes shows a dash instead of a number. If your last scan could measure eleven checks and this one could measure eight, the difference between the two scores is not a change in your help center, and printing it would be a lie. So we do not print it.

What a scan refuses to claim

Three checks — link to a dead article, empty folder, draft never published — work by absence: they conclude something is wrong because an article is not there. They only run when the scan can vouch that it read the whole knowledge base. They are switched off, and the scan says so in plain words, when:

• Your plan's article limit stopped the scan early. Links pointing at articles it never read are not reported as dead.
• Freshdesk did not return every page in a language. Missing and outdated translations are not reported for that language, and existing findings for it are left open rather than marked resolved.
• Freshdesk no longer supports a language your portal used to have. Nothing is scanned in it.
• It was a daily top-up, which by definition did not look at everything.

The same caution applies in reverse: a scan that did not run a check may not mark that check's old findings as fixed either. We never tell you an article is gone when we simply did not read it.

Links we couldn't check

A separate tab from the issues, and never counted in the score. A URL lands there when nothing came back that proves anything: a 401 or 403, a 429, a 5xx, a timeout, a sign-in wall, a domain that does not exist, or an address on an unusual port we do not open. Each line says which article it came from, what happened, and when it was last tried.

Many of them are fine — an intranet page, a partner portal behind SSO. Some are genuinely broken for your readers too. The point is that we do not guess.

Running a scan yourself

The Scan now button on the health scan page starts a full scan immediately. There is a five-minute cooldown after the previous one, only one scan of a connection runs at a time, and a manual scan counts as that day's scheduled run, so it never costs you an extra one.

When a scan can't start, the page says which of those it is.

Freshdesk and Freddy are trademarks of Freshworks Inc. TidyKB is an independent product and is not affiliated with, endorsed or sponsored by any company named on this page.