River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Internal Knowledge Base Template

Three documents and three sheets that name which topics only one person still understands, and which ones already failed a real incident this quarter.

Free download  ·  No account needed

Knowledge Gap Register

[Team], [n] documented topics

Checked against the current roster and [n] incidents this quarter

The register, in four counts

Named owners no longer on the team
Topics with a backup count of zero
Topics carrying both flags, the highest-risk tier
Highest-risk topics confirmed wrong by a real incident

The question a last-edited date can't answer

Not when a topic was last touched. Whether the person listed as its owner is still around, and whether anyone else has ever shown they could fix it if not.

Confirmed against real incidents

Every high-risk topic checked against what actually happened the next time someone relied on it, not just against a calendar.

Nearly every guide to keeping an internal wiki current lands on the same fix: name an owner, add a 'last verified' date, and review it on a cadence. Scribelet's own diagnosis goes furthest, naming the owner field itself as a claim that decays, since a departure never bounces a calendar invite. Sync-o's maintenance guide adds an escalation path for a quiet owner. Neither checks that name against who is still on the team, or ties staleness to whether the page held up the last time someone relied on it.

This pack checks both. On the worked example, Iveston Systems, an invented platform engineering team with 84 documented topics, cross-referencing every named owner against the current roster found 11 who had already left. Checking how many other people had ever substantively edited or reviewed each topic found 23 with a backup count of zero. Seven topics carried both flags, the highest-risk tier, ranked ahead of anything sorted by edit date alone.

The Staleness Report is the confirmation layer. This quarter's incident log showed 37 of 53 incidents consulted a topic from the register, and post-incident review flagged 9 of those 37 as not matching production. Two of the nine came from the seven-topic highest-risk tier, a 28.6 percent failure rate against 9.1 percent for everything else, 3.1 times higher from a slice that is only 8.3 percent of the register. Getting that knowledge out of one person's head before it ever reaches this register is the SME interview pack.

7 of 84 topics carry both ownership rot and a zero backup count. This quarter, those seven failed a real incident 3.1 times as often as everything else

The Topic Register, the Knowledge Gap Register and the Staleness Report.

Topic Register

Iveston Systems, an invented platform engineering team. 84 documented topics. Owner status checked against the current roster. Five of 84 shown.

TopicOwnerStatusBackup
Staging environment reset procedureM. FerreiraDeparted0
Kafka consumer lag remediationO. FitzgeraldDeparted0
Database failover procedureR. SilvaActive0
Payments webhook retry runbookD. OkaforActive2
On-call escalation policyP. NathanActive3

11 of 84 named owners no longer appear on the roster. 23 of 84 have a backup count of zero. Neither flag depends on when the page was last edited.

Knowledge Gap Register

Ranked by ownership rot and backup scarcity together, then checked against this quarter's incident log. Top rows shown.

TopicRisk tierConsultedConfirmed wrong
Staging environment reset procedureHighYesYes
Kafka consumer lag remediationHighYesYes
Vendor invoice dispute processHighNoUnverified
Certificate rotation, ingress gatewayMediumYesYes

7 of 84 topics carry both flags, the High tier, just 8.3% of the register. This quarter, 2 of those 7 were confirmed wrong by a real incident, a 28.6% failure rate against 9.1% for every other topic.

Staleness Report

Every incident that consulted a topic this quarter, cross-referenced against when that topic was last edited.

IncidentTopicEdited <90 daysOutcome
INC-2118Staging environment resetYesDid not match production
INC-2124Kafka consumer lag remediationYesDid not match production
INC-2109Payments webhook retry runbookYesMatched production

37 of 53 incidents this quarter consulted a topic from the register. 9 of those 37 did not match production, and 5 of those 9 had been edited inside the last 90 days, exactly the window a last-edited dashboard reads as fresh.

What's in the pack

01

Topic Register

Every topic in one place, with a named owner, a backup count and a review cadence sized to how fast it actually changes.

02

Knowledge Gap Register

Every topic ranked by ownership rot and backup scarcity together, the two structural signals a last-edited date cannot show.

03

Staleness Report

Every consulted topic's real outcome, cross-referenced against incidents and support cases, so a stale flag rests on evidence rather than a calendar.

04

Runbook Set

The procedures the Knowledge Gap Register flagged as highest risk, written up first. Where they belong in your navigation is a separate pass, the information architecture and navigation pack.

05

Ownership Note

What counts as still owning a topic, the handoff window for a departing owner, and a review cadence set by topic type, not a flat default.

06

Contribution Standard

How anyone besides the named owner contributes a change that actually counts toward the backup count, not just a typo fix.

07

A space rule every prompt reads first

A topic's risk comes from who still knows it and whether it held up in a real incident, never from when it was last edited.

How to use it

  1. 1

    Open it in River, or download it

    Open the pack and the agent cross-references your own roster and incident log before ranking a single topic, or download the blank register and sheets.

  2. 2

    Send your scattered notes and your roster

    Whatever your knowledge currently lives in, plus who is actually on the team today, since that is what a named owner gets checked against.

  3. 3

    Rank by rot and scarcity, not by edit date

    Every topic checked for a departed owner and a backup count of zero, then ranked by that combination rather than by when it was last touched.

  4. 4

    Confirm the ranking against a real incident

    The next time a topic actually gets used, log whether it held up, so the highest-risk tier is backed by evidence rather than a hunch.

Frequently asked questions

Is this template free?

Yes. Download the whole pack as Word documents and CSV sheets, no signup and no credit card. Edit with AI is a separate, optional path for anyone who wants the agent to cross-reference your own roster and incident log. Nothing happens until you send it your scattered notes.

How is this different from just adding a "last verified" date field?

A last-verified date still depends on someone remembering to click it, and it says nothing once the person who verified it leaves. This pack checks the named owner against your actual current roster automatically, and confirms the ranking against real incidents instead of a self-reported click.

Is a backup count of zero the same as a low bus factor?

Related, not identical. Bus factor is usually measured across a whole codebase or project from commit history. Backup count applies the same idea to one documented topic at a time, tied to a specific named owner, so it can flag a single risky runbook without needing to compute risk for the entire wiki at once.

What if we don't have an incident log or usage data yet?

Build the Knowledge Gap Register from the two structural flags alone, ownership rot and backup scarcity, and mark every ranking as unconfirmed rather than inventing an outcome. The Staleness Report gets built the moment real usage data exists, and not a day before.

Do we need to change our review process for the Contribution Standard to work?

Only if editing a topic you don't own is currently discouraged in practice. The standard names what counts as a real contribution and how a non-owner proposes one; wiring that into a review procedure your team already uses is the docs as code workflow pack.

How does this relate to auditing our published documentation?

Different target. The documentation audit checks reader-facing pages against your shipped product. This pack checks your internal wiki against your own team roster and your own incident history, which is a different failure mode with a different fix.

What format are the downloaded files?

Three Word documents and three CSV sheets, zipped. The documents open in Word, Pages and Google Docs; the sheets open in Excel, Numbers and Sheets. Add a PDF query string to the download if you want to circulate the register rather than fill it in yourself.

Find out which topic only one departed person ever understood

Download the blank Topic Register, Knowledge Gap Register and Staleness Report as Word and CSV, or have River check them against your own roster and incidents.

Edit with AI