Enterprise KM dimension: Operational Knowledge Fitness (OKF)¶
Research note on one measurable dimension for enterprise knowledge management (EKM), how it relates to established frameworks, and how a platform like openKMS can support assessment. Written for product, KM, and compliance stakeholders (2026-06).
Related: Goals (vision) · Development plan · Evaluation · RAGFlow vs openKMS
Why this dimension¶
Enterprise KM programs are often judged on activity (pages created, licenses, training hours) or technology (search deployed, RAG pilot launched). Those inputs do not answer the question executives and frontline staff actually care about:
Can we use what is in the system for real work—today—with acceptable risk?
Goals frames the same tension as user adoption vs org governance. Operational Knowledge Fitness (OKF) is a single outcome-oriented dimension that compresses that tension into something you can score, trend, and tie to product capabilities—without replacing full program assessments (strategy, culture, funding) that frameworks such as APQC’s KM CAT still cover.
OKF is deliberately not “RAG accuracy” or “search relevance” alone. It is fitness for operational use by humans and agents under permission, lifecycle, and compliance constraints.
Definition¶
Operational Knowledge Fitness (OKF) — the extent to which an organization’s managed knowledge corpus (documents, articles, wiki, indexed KB content, governed terminology, and linked structured data) supports:
- Authorized discovery — the right people find the right material quickly.
- Verifiable use — answers and deliverables can cite source, version, and scope.
- Situational understanding — newcomers and cross-functional staff can orient (what to read first, what terms mean).
- Currency — “current for purpose” is knowable; stale or superseded material is visible before misuse.
- Reuse — knowledge contributed once is actually reused, not trapped in private chats or individual drives.
OKF is low when staff still default to ask-a-colleague, personal AI without citations, or uncontrolled file shares for work the KMS was meant to support—matching user pains in Goals (e.g. search friction, answers without evidence, knowledge kept private).
OKF is high when the KMS is the first stop for operational questions and updates flow back into the corpus with measurable quality feedback.
Position in common frameworks¶
OKF is not a replacement for APQC, ISO 30401, or SECI-style models. It is a lens on outcomes that cuts across several of their building blocks.
| Framework | Where OKF sits |
|---|---|
| APQC KM CAT (12 capabilities, 5 maturity levels) | Concentrates on Knowledge flow processes, Content management process, KM approaches & tools, and Measurement—not on budget or change management alone. |
| ISO 30401:2018 (KMS requirements) | Aligns with performance evaluation and continual improvement: the organization must define indicators of KMS effectiveness; OKF proposes a structured indicator set. |
| Classic KM lifecycle (identify → collect → organize → share → use → improve) | OKF measures whether the “use” and “improve” stages close the loop, not whether repositories exist. |
flowchart TB
subgraph program [KM program enablers]
S[Strategy / funding]
P[People / culture]
G[Governance]
end
subgraph okf [OKF dimension — outcome]
D[Discoverability]
V[Verifiability]
N[Navigability]
C[Currency]
R[Reuse / contribution]
end
program --> okf
okf --> Outcomes[Shorter time-to-answer · Fewer policy errors · Higher agent trust]Facets (sub-scores)¶
OKF should be reported as one headline score or level plus five facet scores so teams know what to fix.
| Facet | Question it answers | Maps to Goals pains |
|---|---|---|
| Discoverability | Can authorized users find relevant material in one governed place? | Search friction (“can't find”) |
| Verifiability | Can they prove what they relied on (passage, file, version, date)? | Lack of verifiable sources (“can't trust”) |
| Navigability | Can they understand domain structure and entry paths? | Lack of orientation (“don't know where to start”) |
| Currency | Is it clear whether content is still valid for the task? | Obsolete content unnoticed |
| Reuse | Does contributed knowledge get used again by others (or agents)? | No time to contribute · answers kept private |
Facets are partially independent: a corpus can be discoverable but untrustworthy (stale PDFs, no citations in AI answers) or trustworthy but unfindable (good policies buried in email). That is why a single “we have search” metric is insufficient.
Maturity rubric (1–5)¶
Use evidence-based levels (similar spirit to APQC Level 1–5). Advance only when tangible evidence exists, not when tools are merely purchased.
| Level | Label | OKF summary | Typical evidence |
|---|---|---|---|
| 1 | Ad hoc | Knowledge lives outside the KMS; tool is optional. | No shared taxonomy; no citation habit; shadow IT and oral norms dominate. |
| 2 | Accessible | Content is uploaded; search exists; trust is uneven. | Global search or folders used sometimes; few lifecycle fields; AI answers often without traceable sources. |
| 3 | Traceable | Structure, permissions, and citations are standard for critical domains. | Channels/maps defined; ACL on sensitive trees; Q&A or articles cite sources; versions on key documents. |
| 4 | Adaptive | Currency and quality are measured; gaps drive remediation. | Lifecycle/supersedes used; evaluations or audits on KB/wiki; connector/sync for major sources; role-based dashboards. |
| 5 | Agent-ready | Humans and agents share the same governed context layer with closed-loop improvement. | Agents use same ACL and “current for RAG” semantics; eval regressions block releases; contribution paths after Q&A are routine. |
Level 3 is a realistic 18–24 month target for many enterprises; Level 5 requires product and process maturity beyond “RAG demo.”
Indicators and metrics¶
Split indicators into leading (predict OKF) and lagging (confirm business effect). Prefer ratios and time-bounded samples over raw counts.
Discoverability¶
| Indicator | Type | How to measure |
|---|---|---|
| Median time-to-first-relevant-asset | Lagging | Task-based study or instrumented search sessions |
| % operational queries resolved without escalation to expert | Lagging | Ticket tags / survey after KB or global search |
| Search success rate (click or “useful” within top-k) | Leading | Search logs + explicit feedback |
| Coverage: % of declared “golden” topics with ≥1 current asset | Leading | Knowledge map / taxonomy audit |
Verifiability¶
| Indicator | Type | How to measure |
|---|---|---|
| % KB/Q&A answers with ≥1 inspectable source (chunk or page) | Leading | Agent or KB logs (sources, retrieval_debug) |
| % published customer/regulatory responses with linked internal source | Lagging | Sample audit of outbound packs |
| Chunk/document edit rate after parse (human correction load) | Leading | Version/commits per ingested page |
Navigability¶
| Indicator | Type | How to measure |
|---|---|---|
| New-hire time-to-complete onboarding reading list | Lagging | HR/L&D milestone |
| % glossary/map terms with live links to channels or spaces | Leading | Knowledge map + glossary completeness |
| Wiki/graph or map usage vs raw search-only sessions | Leading | Analytics |
Currency¶
| Indicator | Type | How to measure |
|---|---|---|
% policy-tagged documents with effective_to / current-for-RAG set correctly |
Leading | DB query on lifecycle fields |
| Mean age of assets in “current” retrieval set vs known change cadence | Leading | Metadata + channel review |
| Count of dependents flagged when a source document is superseded | Leading | Lifecycle graph (when workflow exists) |
| Incidents traced to obsolete internal knowledge | Lagging | Risk/compliance register |
Reuse¶
| Indicator | Type | How to measure |
|---|---|---|
| Contributor/active-reader ratio on key channels | Leading | ACL-aware access logs |
| Repeat retrieval of same chunks/pages across users | Leading | Search/KB analytics |
| % expert Q&A sessions that produce a persisted FAQ/wiki/article | Leading | Product workflow (see User experience backlog) |
| Duplicate question rate in support channels | Lagging | CRM / chat taxonomy |
ISO 30401 does not prescribe these names; it expects the organization to select indicators appropriate to objectives and review them in management review cycles.
Assessment method¶
A practical OKF assessment blends three inputs (quarterly or per major release):
- Facet checklist interview — KM, compliance, and line-of-business reps score maturity 1–5 with one evidence link per facet.
- System sample — e.g. 30 random operational questions; trace whether the KMS could answer with citation and current scope.
- Automated snapshot — queries run against DB and product APIs (coverage, lifecycle completeness, eval pass rates).
OKF headline = weighted mean of facet scores (weights set per industry)
Suggested default weights (regulated enterprises):
Verifiability 25% · Currency 25% · Discoverability 20% ·
Navigability 15% · Reuse 15%
Suggested default weights (fast-moving ops teams):
Discoverability 30% · Reuse 25% · Verifiability 20% ·
Currency 15% · Navigability 10%
Document weights in the KM charter so scores are comparable year over year.
Sample scorecard (illustration)¶
| Facet | Score (1–5) | Evidence snippet | Priority action |
|---|---|---|---|
| Discoverability | 3 | Global search used; 40% topics lack map link | Link top 20 terms in knowledge map |
| Verifiability | 2 | KB Q&A cites chunks; email still used for audits | Mandate article export with source links |
| Navigability | 3 | Glossary live; onboarding list manual | Publish wiki onboarding hub |
| Currency | 2 | Lifecycle fields on docs; no impact workflow | Ship policy-change dependent queue |
| Reuse | 2 | Low wiki edits; high repeat Slack questions | Ask→contribute shortcut after KB chat |
Headline OKF (equal weights): (3+2+3+2+2)/5 = 2.4 → Level 2–3 (“Accessible” trending toward “Traceable”).
Mapping to openKMS (instrumentation today)¶
How the product supports each facet now vs on the roadmap (Development plan).
| Facet | Shipped signals in openKMS | Gaps / backlog |
|---|---|---|
| Discoverability | Global search; channel trees; knowledge map; hybrid KB search | Connector sync jobs; richer analytics |
| Verifiability | KB search/Q&A sources, chunk_index, retrieval_debug; document/article versions; printable article view |
Eval→fix loop; unified agent citations |
| Navigability | Wiki, glossaries, knowledge map, ontology browse | Onboarding paths; stronger map coverage metrics |
| Currency | Document/article lifecycle, relationships, is_current_for_rag, historical toggle in search |
Policy change impact workflow |
| Reuse | Wiki pages (UI, vault import, CLI / agent upsert); articles; evaluations | Ask→FAQ/wiki; contribution nudges; usage dashboards |
Evaluation module (evaluation.md) is the closest built-in OKF probe today:
| Evaluation type | OKF facet primarily exercised |
|---|---|
search_retrieval |
Discoverability + Verifiability (retrieval + judge) |
qa_answer |
Verifiability (end-to-end answer quality) |
wiki_content_coverage |
Navigability + Currency (does wiki support stated checklist) |
Running eval suites after corpus changes and comparing runs is a concrete Level 4 practice.
Security model (Security design, Data security) affects OKF measurement: metrics must be computed per access scope (ACL), or aggregates will overstate fitness for restricted domains.
What OKF does not measure¶
Keep scope honest to avoid false precision:
| Out of scope for OKF | Where to measure instead |
|---|---|
| KM program funding, executive sponsorship | APQC Strategy / Business case |
| Culture of psychological safety to share | People / Change management assessments |
| Raw LLM benchmark scores (MMLU, etc.) | Model evaluation outside KMS |
| Total cost of ownership of IT | FinOps / architecture reviews |
| Innovation revenue attribution | Portfolio metrics with causal design |
Using OKF in product decisions¶
When prioritizing openKMS backlog items, ask which facet moves most for target customers:
| If the customer pain is… | Prioritize facets | Example backlog themes |
|---|---|---|
| “Nobody can find anything” | Discoverability | Connectors, search, map linking |
| “Legal won’t let us use AI answers” | Verifiability + Currency | Citations, lifecycle, impact workflow |
| “Wiki is empty after launch” | Reuse | Ask→contribute, agent-drafted pages, eval feedback |
| “Agents hallucinate on old SOPs” | Currency + Verifiability | is_current_for_rag, dependent review, eval gates |
This aligns Strategic priorities (connectors, in-product agents, multimodal, evaluation, policy lifecycle) with a single reporting dimension executives can track.
References¶
| Source | Notes |
|---|---|
| APQC, Understanding the APQC KM CAT | Five maturity levels; 12 capabilities including knowledge flow and measurement |
| APQC Excellence in KM | Outcome framing (productivity, decision speed) |
| ISO 30401:2018 | KMS requirements; performance evaluation left to organization-defined indicators |
| Goals | User and organization pains; openKMS north star (vision doc) |
| Nonaka & Takeuchi, SECI (conceptual) | Socialization–externalization–combination–internalization—OKF emphasizes combination/internalization via managed artifacts |
Research note: OKF is an openKMS-internal working dimension for assessment and roadmap alignment, not a published industry standard. Adapt weights and rubric to your sector (regulated vs high-velocity operations).