TF/IDF is a page-local score that weighs your keyword’s vocabulary against how long the page is: RankGear takes a logarithm of the keyword-variation matches and divides it by the logarithm of the body word count. It approximates term prominence relative to page length — it is not corpus-wide textbook TF/IDF, and it relates to position within a measured result set rather than causing a ranking.
| Factor ID | RG-KWD-025 |
|---|---|
| Family | Keyword Usage & Density |
| Measurement | Page-local TF/IDF approximation |
| Measured zone | Raw body HTML without scripts |
What it measures
This factor normalizes a logarithmic keyword-variation count by the logarithm of the page’s raw-body word count. In plain terms, it asks whether the relevant terminology on the page is proportionate to the page’s length, rather than simply counting how often a term appears.
How RankGear measures it
RankGear calculates 1 + ln(variation matches) and divides it by ln(word count), where words are 3-to-64-character tokens in the script-free body. A zero word count returns 0 before division.
score = (1 + ln(variation_matches)) / ln(word_count)
word = token of 3–64 characters in the script-free body
if word_count == 0: score = 0How to optimize it
Create focused, comprehensive content whose relevant terminology is proportionate to its length and supports the user’s task. Treat this as an observation of balance, not a target to inflate: padding a page with keyword variations, or trimming words to move the ratio, works against the reader and against what the signal is meant to reflect.
Important considerations
- This is not corpus-wide textbook TF/IDF — it is a page-local approximation using RankGear’s own methodology.
- Raw markup can affect both the variation matches and the word tokens, since the measured zone is the raw body HTML with scripts removed.
- Zero matches can produce a nonfinite numerator, so a page with no variation matches will not score meaningfully.
- Use the result comparatively — against pages measured with the same RankGear methodology — not as an absolute quality number.
- Correlation is not causation: this score relates to position within a measured result set; it does not make a page rank or guarantee a position.
Related factors
Part of the Factors reference · how RankGear measures · glossary.