Strategy
Entity analysis
An entity is a thing a search engine holds a record of: a company, a person, a place, a product, a condition, a piece of legislation. It is not a keyword. "Blackburn" is nine characters. The town in Lancashire, the town in West Lothian and the suburb of Melbourne are three separate entities that happen to share those characters, and a search engine resolves the string to one of them before it decides what to show.
Entity analysis is the work of establishing which entities a page ought to be about, which ones it actually covers, and which ones it has left out.
Why it is worth doing
Keyword research tells you what people type. It does not tell you what a page has to contain in order to be a credible answer. Two pages can target the same phrase, use it the same number of times, and differ completely in whether they mention the regulator, the competing method, the cost, or the reason someone would choose the alternative.
That difference matters more as answers get assembled rather than listed. When a search engine or a language model builds a summary, it is pulling facts about entities and the relationships between them. A page that names the relevant entities plainly, and states how they relate, is easier to pull from than a page that circles the topic in general language.
How the work is done
Start with the subject and establish its entity, not its label. Check how it is represented in the public knowledge sources: Wikidata, Wikipedia, industry registers, company records. Those records are where a lot of machine understanding of a subject begins.
Then take the pages that already rank for the query and extract the entities each one names. Natural language processing tools will do the extraction and return a salience score, which is an estimate of how central each entity is to the document. Compare that set against your own page.
The gap is the finding. It usually falls into three shapes: entities that everyone else covers and you do not, entities you mention once in passing where others treat them properly, and entities you name ambiguously so that a machine cannot tell which one you mean.
What you do with the output
Fill the genuine gaps in the copy. Disambiguate the ambiguous mentions by naming the entity fully at least once. Link internally between the pages that cover related entities, because the link is a statement that the two are connected. Where a relationship can be stated in structured data, state it, so it is not left to be inferred from prose.
What it does not tell you
Entity coverage is not a dial. There is no count of entities that earns a position, and salience scores are one vendor's estimate rather than a reading of what any search engine actually stores.
Treat the analysis as a way of finding what a page failed to say, which is a question editors are bad at answering about their own work, and not as a score to raise.