Skip to content
Browse courses

Entities & Knowledge Graphs

Entity-Attribute-Variable (EAV) Model

In shortA framework that breaks topics into entities, their attributes and the variables (values) those attributes take, e.g. dog, breed, Labrador.

Updated 24 Oct 2026· 2 min read

The Entity Attribute Variable (EAV) model describes a topic as three layers: entities (the things people search for), attributes (the ways those things are described) and variables (the specific values each attribute can take). For the entity “dog”, “breed” is an attribute and “Labrador” is a variable. In database design the same structure is called entity-attribute-value, and MLforSEO uses it as a practical framework for semantic keyword research.

How it works

EAV started outside SEO as a way to store sparse data, such as medical records where most patients have values for only a few of thousands of possible measurements. Facts are stored as triples (entity, attribute, value), and only the ones that exist are recorded. Knowledge graphs use the same triple shape: “Apple Inc., founded, 1976” is one fact.

For keyword research you:

  1. Pick the entity your content or product is about.
  2. List the attributes that matter to searchers.
  3. List the variables each attribute can take.
  4. Combine them to generate and validate query patterns.

MLforSEO stresses the difference between finite variable sets (breeds, countries, months) and effectively infinite ones (people’s names, slogans), because only finite sets can be enumerated sensibly.

How it’s used in SEO and marketing

  • Programmatic keyword universes. Generate patterns such as [variable] [entity] [attribute] and check them against search data.
  • Gap analysis. Compare which attributes competitors cover for an entity and which they skip.
  • Site architecture and filters. Attributes often become facets or category filters, and variables become the filter values.
  • Structured data. Attributes map naturally to schema.org properties.
  • AI search. Query fan-out tends to split a question by attribute, so covering the attributes that matter makes a page useful for more sub-questions.

Example

Entity: dog. Attributes: breed, life stage, size, food type. Variables: Labrador, puppy, large, grain-free. Combining them gives queries such as “best grain-free food for Labrador puppies”, and checking which combinations have real demand tells you which deserve a page and which belong as a section of a broader page.

Limitations and things to watch

  • Combinations multiply fast. Most will have no demand, and publishing a page for each risks thin, near-duplicate content.
  • Every generated pattern needs validating with search data and the live results page.
  • The model maps the information landscape; you still need original content to fill the gaps it reveals.

Learn it on MLforSEO

Where Entity-Attribute-Variable (EAV) Model shows up on MLforSEO

Learn it in a course

Read about it on the blog

Try it yourself

Sources and further reading

Other terms from Entities & Knowledge Graphs

See all 23 Entities & Knowledge Graphs terms →