Gaps and contradictions: the two failures your docs cannot see
A content gap is demand with no page behind it. A contradiction is two pages that disagree. Neither is visible to a tool that only looks inside one silo — and both are expensive.
Once you have a model of the whole estate rather than one silo, two things fall out of it that no single-silo tool can compute. They are worth naming separately, because they have different causes and different fixes.
A gap is demand with no supply
Your community is a demand signal that nobody reads as one. When forty threads discuss proxy timeouts and the documentation has no page on the subject, that is not an anomaly — it is a measurement. The readers told you what was missing, at length, in public, and the information stayed trapped in a channel the docs team does not treat as an input.
Kenaptic computes this across buckets: topics with real demand in one part of the estate and no coverage in a part you could publish to. On one mid-sized estate we analysed recently, the community carried more than 5,000 topics with no counterpart anywhere else — every one of them a question somebody asked that the documentation never answered.
This matters commercially because of what happens next. Gartner finds customers fully resolve only 14% of their service and support issues in self-service (August 2024), while the same firm benchmarks the median self-service contact at roughly $1.84 against about $13.50 assisted. A gap is a guaranteed escalation from the cheap channel to the expensive one — the reader cannot self-serve, because the page does not exist.
A contradiction is two pages that cannot both be right
Fast-shipping teams produce these constantly and almost never catch them. The changelog says a default changed; the tutorial still teaches the old one. The knowledge base says tokens last 24 hours; the reference says 90 days. Both pages were correct when written. Neither is wrong in a way that any linter, test or reviewer of a single page could detect, because the error only exists in the relationship between two documents that nobody reads together.
Detecting these well turns out to be mostly a matter of restraint. Our first instinct was to flag numeric and version differences between similar passages — it produced a 66% flag rate at approximately zero precision, because most numeric differences between two pages are not disagreements at all. What works is far narrower: sample passages that carry explicit change language — deprecated, no longer, replaced by, breaking change — pair each with its nearest cross-bucket neighbour, and judge that pair. Fewer candidates, dramatically better findings.
We would rather report five contradictions that are all real than fifty that a reader has to triage. A detector people stop trusting is worse than no detector, because it consumes the attention that the real findings needed.
Both are reported honestly, including when we cannot answer
If your estate has no writable destination configured, Kenaptic cannot tell you what is missing from it — and it says exactly that, rather than reporting zero gaps. This sounds like a small distinction. It is not. “We found nothing wrong” and “we were unable to look” are completely different statements to put in front of a team, and a product that conflates them will eventually be trusted at precisely the wrong moment.
The same principle runs through the dead-link detector: it reports only on links inside the estate we actually crawl, and states what it excluded. Calling a working link broken costs more trust than saying nothing.
What you do with them
A gap becomes a drafted page you can accept, edit or discard. A contradiction becomes a proposed correction at the source, routed through the same review gate and the same one-click retraction as every cross-link. Nothing about a finding gives it a shortcut past a human.
Kenaptic proposes; a person decides. Every link is reviewed before it publishes and retractable in one click — see Security.
