A Rubric for Consistent Anchor Classification
Every anchor audit depends on classification, yet practitioners bucket inconsistently — one analyst's 'partial-match' is another's 'exact-match'. Without a written rubric, your before/after comparisons are noise. A reproducible scheme:
— Exact-match: the anchor equals the target query with no additional words ('cheap car insurance').
— Partial-match: the target query appears inside a longer phrase ('compare cheap car insurance here').
— Branded: contains the brand token and no commercial keyword.
— Generic: contextless calls to action ('read more', 'click here').
— Naked URL and image/empty as separate buckets.
The boundary cases decide everything. Document your rules for them in advance: Is a brand-plus-keyword anchor branded or partial? Is a near-synonym exact or partial? Pick a rule and apply it identically across every audit so the deltas are real.
The value is methodological, not algorithmic — Google's internal classification is unknown and surely more sophisticated. Your rubric's purpose is internal consistency, letting you measure change in your own profile reliably.
On one hand, finer buckets capture more nuance. On the other, more buckets multiply boundary disputes and reduce reproducibility — a precision/reliability trade-off.
Limitation: any human rubric is a coarse proxy for whatever semantic model Google actually applies.
Open question: should near-synonyms (stemming, close variants) count as exact-match, given search engines treat them as equivalent for ranking?
Anchor Theory
@AnchorTheory
A Rubric for Consistent Anchor Classification
Этот пост опубликован в Telegram-канале Anchor Theory. Подписаться можно по ссылке: @AnchorTheory.