Significance Threshold is the minimum absolute correlation a factor must reach before RankGear treats it as statistically supported within the current sample. Below that cutoff, the relationship is considered too weak to distinguish from noise.
| Term | Significance Threshold |
|---|---|
| Category | Statistics and Evidence |
| Also known as | Correlation Threshold, Threshold |
| Where it appears | Diff report and factor filtering |
What it means in RankGear
When RankGear analyzes a set of ranking pages, it measures how strongly each of the 870 factors correlates with position across that sample. The Significance Threshold is the line that separates factors worth acting on from those whose correlation is small enough to be sample noise. In the Diff report and in factor filtering, it works as a gate: factors whose absolute correlation clears the threshold are surfaced and flagged, while weaker ones are set aside so the report stays focused on differences that are actually supported by the data in front of you.
How to interpret it
Read the threshold as a confidence gate, not a ranking weight. It uses the absolute value of the correlation, so a strong negative relationship (a factor where less is associated with better positions) can clear it just as a strong positive one does. A factor sitting just above the line is being included on thin evidence, and a small SERP sample raises the odds that such a factor crossed the line by chance. When you have room to be selective, raising the threshold keeps only the factors with the firmest support; lowering it widens the net but lets in weaker signals. Always weigh a factor that clears the threshold against the size of the sample it was measured on.
| Factor correlation vs. threshold | What it tells you |
|---|---|
| Absolute correlation well above the threshold | The association is strong and consistent enough within this sample to treat as a supported signal. |
| Just above the threshold | Included, but on marginal evidence; verify against sample size before acting. |
| Below the threshold | Filtered out as too weak to distinguish from noise in the current sample. |
Example
You run a Diff report for a 12-result SERP with the Significance Threshold set to 0.30. A factor measuring content word count shows an absolute correlation of 0.52, so it clears the gate and appears in the report as a supported difference. A factor for exact-match anchor text comes in at 0.19 and is filtered out. Because the sample is only 12 pages, you nudge the threshold up to 0.40 to be stricter; the word-count factor still qualifies, giving you more confidence that it is worth prioritizing rather than an artifact of a small run.
Important considerations
- Clearing the threshold shows statistical association only. It does not establish that the factor causes rankings, and it is never a Google score, only a comparative indicator within your sample.
- Small SERP samples are unstable, so a factor can cross the threshold on a handful of pages and drop below it on the next run. Treat marginal factors as provisional until a larger or repeated sample confirms them.
- The threshold is applied to the absolute correlation, so it captures strong negative relationships as well as positive ones.
- A stricter (higher) threshold reduces false positives but can hide real, weaker signals; a looser one surfaces more factors at the cost of more noise. Choose the level to match how much you trust the sample.
Related terms
Part of the RankGear glossary · how RankGear measures · the 870 factors.