Contextual Vector, Hierarchy and Domain in Semantic SEO

CONTEXTUAL VECTORS

A contextual vector is the angle from which a URL approaches its attribute, set by the order of its headings and the questions they answer. It differs from the topic, which is what the page is about, and from search intent, which is what the searcher wants done. Every URL carries one contextual vector, and two URLs that approach one attribute from the same angle compete with each other. Headings set the vector through question order, where answering one question raises the next. Four related concepts complete the model: contextual hierarchy orders contexts by importance inside a document, contextual domain is the field in which a word is read, contextual borders mark where a page stops, and contextual flow carries meaning from one sentence to the next. One attribute, cost of retrieval, shows four vectors as four pages, a vector is assigned in five steps, and four mistakes break it. Holistic Radar’s own pages follow the same rules, and the questions at the end cover long pages and AI search.

What a contextual vector is

A contextual vector is the angle from which a URL approaches its attribute. Two pages can cover the same entity-attribute pair, such as the cost of retrieval of a website, and still differ completely: one defines it, one explains how to lower it, one measures it, and one compares it with page speed. Each is a different vector on the same attribute.

Definitionwhat it isProcesshow to lower itMeasurementhow it is scoredServicewho lowers itCost ofretrievalone attributeEach angle is a contextual vector, and each gets one URL
Figure 1 · One attribute, several angles

The word vector describes direction. The page starts from its attribute and moves in one direction through its headings, and every section moves further along that direction. A definitional page moves from what the attribute is, through its parts, toward how it is recognized. A procedural page moves from the starting state, through steps, toward the result.

A contextual vector also decides what a page leaves out. A definitional page about cost of retrieval does not list tools, prices or a step-by-step method, because those belong to other vectors on the same attribute. The omission is deliberate: it keeps the page’s passages pointing in one direction, so any one of them can be retrieved and still read as part of a definition.

Contextual vector is a planning concept before it is a writing one. The topical map assigns each URL its attribute and its vector together, so the writer receives a page that already has a direction.

Contextual vector, topic and search intent

A contextual vector is often confused with the topic and with search intent, and the three answer different questions.

ConceptQuestion it answersExample for cost of retrieval
TopicWhat is the page about?Cost of retrieval
Search intentWhat does the searcher want done?Understand it, or reduce it
Contextual vectorFrom which angle does the page approach it?Definition, process, measurement or comparison
Topicwhat it is aboutSearch intentwhat the searcher wants doneContextual vectorthe angle the page takes
Figure 2 · Three questions about one page

The vector follows from the intent but is not the same thing. A searcher who wants to reduce the cost of retrieval may be served by a process vector, a checklist vector or a case vector, and the map chooses one per URL.

One contextual vector per URL

Each URL carries exactly one contextual vector. A page that defines an attribute in its first half and sells a service for it in its second half has two vectors, and it serves neither reader fully. A search engine reading it cannot tell whether the page is a definition or an offer, so it ranks the page as a weak version of both.

CLUSTERDifferent vectorsDefinition page and process pageHeadings cannot be swappedEach owns its queriesCANNIBALIZATIONSame vectorTwo definition pagesHeadings could be swappedQueries split between themvs
Figure 3 · Same attribute: cluster or cannibalization

The rule also prevents competition between URLs. Two pages on one attribute with different vectors are a cluster; two pages on one attribute with the same vector are cannibalization. The test is simple: if the two pages’ heading lists could be swapped without anyone noticing, they share a vector, and one of them should be merged or rewritten.

How headings set the contextual vector

Headings set the contextual vector because each heading is a question the page answers, and the order of the questions is the direction of the page. A definitional sequence asks what the attribute is, what it consists of, how it differs from its neighbours and how it is recognized. A procedural sequence asks what is needed, what comes first, what comes next and how the result is checked.

What is it?definitionWhat raises it?causesHow is itmeasured?stagesHow is itlowered?next vector:bridgeWhen the next question belongs to another vector, the page has reached its border
Figure 4 · Each answer raises the next question

The sequence follows inquisitive semantics: answering one question generates the next question a reader would ask. After “what is the cost of retrieval”, the next question is “what raises it”; after that, “how is it measured”; after that, “how is it lowered”. A heading that breaks the chain, jumping from definition to pricing, forces the reader and the search engine to change direction mid-page.

Question order also decides where the page ends. When the next question a reader would ask belongs to a different vector, such as “who can do this for me” at the end of a process page, the page has reached its border, and the answer is a bridge to the page that owns that question rather than another section.

The extractive summary previews the same sequence in the same order, so the vector is stated once in the first paragraph and then followed heading by heading. The question types article covers which questions belong to which vector.

Contextual hierarchy

Contextual hierarchy is the order of contexts inside a document by importance. The macro context, the page’s attribute and vector, comes first and receives the most space; micro contexts, the sub-questions and related entities, follow in descending order of importance; supplementary content comes last, after a bridge heading.

01Macro contextattribute and vector, most space02Micro contextssub-questions, related entities03Bridge headingmain to supplementary04Supplementary contentrelated pages, last
Figure 5 · Contexts ordered by importance

Hierarchy is visible in three places. The extractive summary names the macro context in its first sentence. The first heading answers the main question before any sub-question. The supplementary section, such as related pages or a closing bridge, sits below the main content, never between its sections. A page that answers a sub-question before the main one inverts its hierarchy, and passage-level retrieval then finds the sub-answer where the main answer should be.

Contextual domain

A contextual domain is the field of meaning in which a word, an entity or an attribute is read. “Authority” in an SEO contextual domain means demonstrated coverage of a subject; in a legal domain it means a source of binding law; in a public-health domain it means an official body. The same word carries a different attribute set in each.

01SEOauthority: demonstratedcoverage of a subject02Lawauthority: a source ofbinding law03Public healthauthority: an officialbodyState the domain early; every later word is read in it
Figure 6 · One word, three domains

A page states its contextual domain early, usually in its first sentence, so every later word is read in that domain. On a site with one central entity, the domain is stable across pages, which is one reason a site with a clear central entity is easier for a search engine to interpret: its words keep the same meanings from page to page. Ambiguous terms that cross domains are resolved where they first appear, as described in lexical semantics.

Contextual borders and bridges

A contextual border is the point where a page stops covering its own attribute and a neighbouring attribute begins. The border keeps a page on its vector: material that belongs to a neighbour is not explained, only named. A contextual bridge is an internal link between two sections that carries a justification passage naming the concept that connects them.

A border without a bridge isolates a page; a bridge without a border lets the page drift into its neighbour. Used together, the border stops the page and the bridge tells the reader where the next attribute is covered, which is how a site keeps every page focused and every page connected at once.

Contextual flow

Contextual flow is the continuity of meaning from one sentence to the next. In a page with good flow, the object of one sentence often becomes the subject of the next: “A topical map assigns each attribute to a URL. Each URL then receives a content brief. The content brief fixes the headings.” Each sentence picks up what the last one left, so the reader never has to look back to find the subject.

CHAINED SENTENCESA topical map assigns each attribute to a URL. Each URL then receives a content brief.objectnext subject
Figure 7 · The object of one sentence becomes the next subject

Flow fails when a paragraph introduces a new subject without a link to the last one, or when a pronoun refers back across several sentences. Both raise the cost of reading for a person and the cost of extraction for a search engine. Flow is the sentence-level form of the contextual vector: the page moves in one direction, and so does each paragraph.

One attribute, four contextual vectors

The cost of retrieval of a website shows how one attribute divides into vectors. On a site that covers the attribute completely, each vector gets its own URL; the headings below are a planning example, not a page list.

VectorFirst headingLast headingSection
DefinitionWhat cost of retrieval isHow cost of retrieval is recognizedOuter
ProcessWhat raises cost of retrievalHow to verify the reductionOuter
MeasurementThe stages of retrievalReading the scoreOuter
ServiceWhat the work includesHow an engagement startsCore

The four pages share an attribute and link to each other, and none repeats another’s headings. A site that published the definition and the process on one URL would carry two vectors on one page; a site that published two definition pages would carry one vector on two URLs. The map prevents both by assigning vectors before writing starts.

How to assign a contextual vector in five steps

A contextual vector is assigned in the topical map, before the brief, in five steps:

  1. Take the entity-attribute pair the URL owns.
  2. Read the queries the pair receives and group them by what the searcher wants done: understand, compare, do or buy.
  3. Check which of those groups already have a URL on the site, so the new page does not repeat a vector.
  4. Choose the vector for the largest group without a URL, and record it in the map row.
  5. Write the heading sequence as a chain of questions in that direction, and the extractive summary as its preview.
Pairentity-attributeGroupqueriesby jobCheck URLsno repeatsChoosevectorrecord itQuestionchainheadings +summary
Figure 8 · Five steps in the map row

The third step is the one most often skipped, and it is where cannibalization starts. A vector chosen without checking the existing pages is a vector chosen twice.

Contextual vector mistakes

Four mistakes break a contextual vector, and each is visible in the heading list before the page is published.

  • Two vectors on one page. The page defines and then sells, or compares and then explains a process.
  • Drift mid-page. The first half follows one direction and the second half follows a neighbouring attribute.
  • One vector on two URLs. Two pages approach the same attribute from the same angle and split its queries.
  • A summary that disagrees with the headings. The first paragraph previews one order and the page follows another.

Each mistake has the same fix: return to the map row, confirm the vector, and cut or move every section that points elsewhere.

Contextual vectors on holisticradar.com

Every URL on holisticradar.com is assigned one attribute and one vector in the Map Template. The outer articles carry definitional, comparative and procedural vectors under the Know predicate; the core pages carry service vectors under Rank. Where an outer article and a core page share an attribute, such as topical maps, the article defines and the service page delivers, and each links to the other through the bridge table.

The audit that produced this article found the clearest test of the rule in Search Console: where two pages shared a vector, such as the YMYL article and the YMYL industry page, impressions for one query split between them. Separating their vectors in their titles and openings was the fix.

Contextual vector questions

Can a long page have more than one contextual vector?

A long page has one contextual vector and many micro contexts. Length comes from depth along the vector, not from adding a second direction.

Contextual vectors matter for AI search because answer engines retrieve passages, and a passage on a page with one vector states its subject the same way the rest of the page does, which makes it easier to quote without its surroundings.

Is a contextual vector the same as an embedding vector?

A contextual vector is not an embedding vector. An embedding is a numerical representation a model computes; a contextual vector is the direction a writer gives a page. A page with one clear contextual vector does tend to produce passages whose embeddings sit close together.

Contextual Vector in the semantic SEO topical map

Contextual Vector is the direction attribute of the page semantics entity inside the Holistic Radar topical map, and it sits in the outer section under the Know predicate.

The outer section holds the page because it defines how a page is planned without selling the planning, and its bridge passes session signals toward Content Network Build, where every URL receives its attribute and its vector before it is written.

→ Content network build: every URL assigned one attribute and one vector.

→ The ten question types: which questions belong to which vector.

→ Content cannibalization: what happens when two URLs share a vector.

→ Lexical semantics: how word meaning holds a page in one domain.

Get my free Radar ScanFree Radar Scan