Variance is the average of the squared distances between each value in a set and the set’s mean — the raw measure of spread from which standard deviation (its square root) is taken.
| Term | Variance |
|---|---|
| Category | Statistics and Evidence |
| Also known as | Statistical Variance |
| Where it appears | Statistical interpretation |
What it means in RankGear
When RankGear compares a set of pages on a single factor — word count, link velocity, a provider score — variance quantifies how tightly or loosely those pages cluster around their average. A factor where the top results all sit near the same value has low variance; a factor where they scatter widely has high variance. RankGear surfaces this in its statistical interpretation so you can tell a stable, agreed-upon signal from a noisy one before you act on the average alone.
How to interpret it
Variance is always zero or positive, and it grows with the square of the spread, so it is exquisitely sensitive to outliers — one extreme page can inflate it dramatically. Because the value is in squared units it is hard to read directly; most people convert to standard deviation for an intuitive sense of scale, then reason about variance when comparing consistency between factors or runs. Read it alongside sample size: with only a handful of SERP results, a variance figure is itself unstable and can swing on the next crawl.
| Variance reading | What it tells you |
|---|---|
| Near zero | Ranking pages agree closely on this factor — the average is a dependable target. |
| Moderate | Some genuine spread; treat the average as a center of gravity, not a precise mark. |
| High | Wide disagreement or an outlier is present; the average alone can mislead, so inspect the distribution. |
Example
Two RankGear runs on the same keyword both report an average of roughly 1,800 words for the top ten pages. The first run has low variance — nearly every page lands between 1,650 and 1,950 — so 1,800 words is a confident target. The second run shows the same average but high variance, with pages ranging from 600 to 3,200; here the mean is an artifact of a few very long and very short outliers, and copying it blindly would be a mistake.
Important considerations
- Variance is reported in squared units, which have no intuitive meaning on their own — convert to standard deviation when you need a human-scale sense of spread.
- Its squaring makes it highly reactive to a single extreme page, so pair it with an outlier check before trusting it.
- Small SERP samples produce unstable variance that can change materially between crawls; more results give a steadier figure.
- A low-variance factor shows that ranking pages agree, not that the factor causes ranking — statistical association is not causation, and provider scores are comparative indicators, not Google’s own scores.
Related terms
Part of the RankGear glossary · how RankGear measures · the 870 factors.