Publishers have always accumulated assets.
A newspaper builds an archive. A retailer builds a product catalog. A healthcare publisher develops a library of articles, guides, directories, comparisons, and reference pages. A manufacturer may spend years building technical documentation around its products.
Those publications have real value. They can establish authority, attract audiences, explain difficult subjects, support customers, earn links, generate revenue, and create a body of work that becomes increasingly difficult for a competitor to reproduce.
Nothing about knowledge engineering changes that.
What changes is our understanding of what may exist beneath those publications.
If many pages depend on the same identities, facts, concepts, relationships, provenance, and contextual rules, then the organization possesses another asset alongside its content inventory.
It possesses knowledge.
The distinction can seem abstract until something changes. A specification is revised. A Medicare plan enters a new plan year. A source dataset is updated. A product relationship changes. An error is discovered in a calculation.
At that moment, the difference between maintaining publications and maintaining knowledge becomes very practical.
20.1 Content Is an Expression of the Asset
At one point, a company I worked with had twenty-seven articles about Medicare and dental coverage.
Twenty-seven.
No one had deliberately designed a twenty-seven-document Medicare dental architecture. The collection had accumulated naturally. People asked different questions. Search data revealed different formulations of those questions. Each represented a legitimate information need, and over time those needs became articles.
The individual decisions were reasonable.
The resulting system was not.
The articles inevitably overlapped. The same Medicare rules appeared in different explanations. Definitions were repeated. Distinctions between Original Medicare and Medicare Advantage appeared again and again. Exceptions surfaced in several places. One article approached the subject through coverage, another through costs, another through eligibility, another through a particular dental service.
From a content-inventory perspective, the company had built an impressive asset.
From a maintenance perspective, it had created twenty-seven places where substantially overlapping knowledge could become inconsistent.
When an underlying fact changed, the publishing question was no longer simply whether we knew the new fact.
We had to determine where we had said the old one.
That could mean searching twenty-seven articles, deciding which passages were affected, updating them consistently, and hoping another expression of the same knowledge was not hiding somewhere else on the site.
The problem was not that the company had published twenty-seven useful articles. Some questions genuinely deserved their own explanations, and those explanations could remain valuable to readers.
The problem was that the factual knowledge beneath those articles had been maintained as though each expression were independent.
The organization did not possess twenty-seven different bodies of knowledge about Medicare and dental coverage.
It possessed one comparatively bounded body of knowledge that had been expressed through twenty-seven information products.
Once that distinction becomes visible, the definition of the asset changes.
The articles remain valuable.
But the durable factual asset is the maintained knowledge they share.
20.2 Knowledge Compounds Differently Than Content
The two assets grow differently.
When a publisher creates another article, it gains another publication. That article may become extremely valuable. It may rank well, attract links, build authority, generate revenue, or become the definitive explanation of an important subject.
But the article remains an information product with its own maintenance requirements.
Knowledge can compound in another way.
Return to the camera catalog from Chapter 2.
Suppose the retailer begins with ten consistently represented attributes for every camera. Those attributes already support many possible questions and comparisons.
Now the retailer adds another useful attribute across the catalog: weather sealing.
It has not merely created one additional answer.
Weather sealing can now participate in combinations with price, weight, sensor format, lens mount, stabilization, video capability, and every other dimension already represented.
A customer can ask for weather-sealed cameras below a certain weight.
Another can add a price constraint.
Another can require a particular lens mount.
Another can compare only models with in-body stabilization.
The new knowledge acquires utility through its relationship to knowledge already present.
Relationships can compound in the same way.
If the publisher establishes that the Apex X5 Mark II uses the Apex Z lens mount, questions involving compatible lenses become possible. If compatible lenses already have Knowledge Objects of their own, the new relationship connects previously separate bodies of knowledge.
Corrections can compound too.
Suppose the publisher discovers that the weight of one camera was recorded incorrectly.
In a publication-centered system, the organization may need to find every place the old weight was written and correct each occurrence. A missed buying guide or comparison table can preserve the error long after the primary product page has been fixed.
The maintenance problem looks roughly like this:
Fact changes
↓
Find every published expression
↓
Correct each one
↓
Hope nothing was missedA maintained Knowledge Space creates the possibility of a different relationship:
Reality changes
↓
Maintained knowledge changes
↓
Published expressions followThat does not happen automatically. A publisher still needs systems capable of producing and maintaining those expressions.
The important point is architectural rather than technical.
The correction belongs upstream.
The same principle applies to provenance. If a new source version replaces the one supporting a set of assertions, the publisher can update the knowledge at the level where that provenance actually applies.
If a new resolution condition is discovered, that condition can improve future uses of the affected assertions rather than being explained independently every time the problem appears.
If a relationship changes, the maintained relationship can change.
This is the compounding utility of the Knowledge Space described earlier in the book. Improving the maintained knowledge can increase the number and quality of information needs that the publisher can support without requiring an independently authored factual answer for each one.
The economic implication is not that knowledge engineering makes publishing cheap.
Maintaining authoritative knowledge can be expensive.
The difference is where the investment accumulates.
Instead of every new information need necessarily producing another independently maintained factual artifact, some investments improve the underlying asset from which many information products can draw.
20.3 Organizations Already Own More Knowledge Than They Publish
Once we began thinking this way, another realization followed.
Much of the Knowledge Space already existed.
We simply had not considered all of it part of the publication.
The Plan-ID problem made this unusually clear.
Our systems knew that a parent Medicare plan could contain multiple segments. They knew the premium associated with each segment. They knew the geographic service areas. They knew that geography determined which segment applied.
The public resource could expose the premiums while the rule connecting geography to the applicable segment remained inside application logic.
From an operational perspective, that arrangement made sense. The application asked for geography and returned the appropriate result.
From a knowledge perspective, something important had been divided.
The organization knew the facts.
It also knew how their applicability was determined.
Only part of that knowledge was being published.
The same pattern exists in many organizations.
A manufacturer may publish product specifications while its internal systems contain compatibility rules connecting products, accessories, configurations, and model generations.
A retailer may publish thousands of product pages while its catalog systems know that several differently named attributes represent the same underlying concept.
A university may publish information about programs while its operational systems know how programs relate to campuses, academic years, prerequisites, and degree requirements.
A financial publisher may present calculated metrics while the methodology connecting source filings to those metrics exists elsewhere.
In each case, the organization possesses knowledge that extends beyond what appears in its prose.
Some of it lives in databases.
Some exists in application logic.
Some is enforced by data pipelines.
Some appears in editorial policies.
Some is documented in methodology.
And some exists primarily in the experience of people who understand the domain.
Earlier in the book, we described this as operational knowledge. The organization already knows distinctions that an external consumer may otherwise have to reconstruct: which identifier represents which entity, which configuration matters, which source is authoritative, when a value applies, and how one thing relates to another.
The organizational implication does not require a new department or an elaborate enterprise program.
It begins with a simpler question:
What does this organization know that its public information architecture does not yet preserve?
The answer will differ by publisher.
For some, the most important missing knowledge may be product relationships. For others, it may be terminology, provenance, geographic applicability, temporal state, or the methodology behind derived assertions.
Not all internal knowledge belongs on the public Web. Organizations possess confidential information, operational detail, private customer information, proprietary processes, and many other things that should remain internal.
The point is narrower.
Organizations often possess publicly useful knowledge about the subjects on which they already publish that remains implicit, fragmented, or trapped inside systems built for other purposes.
Recognizing that knowledge as part of the publishing asset changes what becomes worth maintaining.
20.4 Steward the Knowledge, Not Just the Pages
Once the Knowledge Space becomes an asset, maintenance begins to look different.
Publishers have always maintained pages. Editors correct errors, update articles, revise explanations, replace outdated statistics, repair links, and respond when the underlying world changes.
Those responsibilities remain.
Knowledge stewardship asks what changed underneath the publication.
Suppose a Medicare premium changes with a new plan year.
The publishing task is not merely to find the page displaying the old premium and replace one number with another. The new value belongs to a different temporal state. The old value may remain perfectly valid as historical knowledge. The source supporting the new assertion may be different. Other representations of the current plan may need to reflect the change.
Or suppose a publisher changes the methodology used to calculate a county average.
The visible result may change from $15.26 to $15.41.
From the reader’s perspective, that looks like a number changing.
From the knowledge perspective, the derivation changed. The provenance chain changed with it.
Stewardship means preserving those distinctions rather than treating every update as a text-editing problem.
The same applies to corrections.
If the Apex X5 Mark II has always weighed 658 grams but the publisher accidentally recorded 685 grams, the publisher is correcting an error in its representation of reality.
If a successor model weighs 685 grams, the publisher is representing a different subject.
If the manufacturer revises the official measurement methodology and publishes a new weight for the same product, the source state has changed.
All three situations can produce a changed number on a webpage.
They are not the same knowledge event.
Thinking at that level does not require every editor to become a knowledge engineer. It requires the publishing system to respect the distinctions that matter to the integrity of what the organization knows.
This is also where the asset argument becomes more durable than any particular distribution strategy.
Pages can lose rankings.
Queries can disappear.
Search interfaces can change.
Answer engines can change how they retrieve and cite sources.
A social platform can rise and fall.
A format that performs well today may matter much less five years from now.
Publishers cannot make any one distribution system permanent.
The authoritative knowledge they possess can survive those changes.
A camera manufacturer will still need accurate knowledge about its products if search interfaces change. A university will still need coherent knowledge about its programs. A healthcare publisher will still need to know what its sources establish, what it derives from them, and where those assertions apply.
The representations will change.
The distribution systems will change.
The information needs will change.
The underlying responsibility remains remarkably stable: maintain an accurate, coherent account of what the organization knows about the things on which it claims authority.
That makes the Knowledge Space different from a content archive.
An archive records what the publisher has published.
A maintained Knowledge Space represents what the publisher currently knows, what it knew within relevant historical contexts, where that knowledge came from, and the conditions under which it remains valid.
Both have value.
But they are different assets.
The Web will continue changing how information is discovered, assembled, and delivered. Publishers cannot know which interface will dominate the next stage of that evolution.
They can decide what they bring into it.
The durable publishing asset is therefore not only the collection of things the publisher has written.
It is the maintained body of knowledge from which those publications—and information products not yet imagined—can be produced.