Methodology
What RankGear measures — and what a score does not promise
RankGear scores your page against the pages already winning a query. Here is exactly what it measures, how the numbers are produced, and the line it will never cross: correlation is evidence for what to prioritize, not proof of what will rank.
The field
It starts with the field, not a rulebook
The field is the set of pages actually ranking for your exact query — at your chosen depth (Top 10 to Top 100), country, language, device, and location. RankGear measures that field first, then places your page inside it. A generic checklist tells you what “good SEO” looks like in the abstract; the field tells you what is winning this query, right now. Your tracked page is excluded from its own benchmark so it can’t skew the comparison.
You choose the deficit strategy — the statistical target a gap is measured against: the field average, the maximum after an outlier guard, or the competitor mean plus 1, 1.5, or 2 standard deviations. The strategy changes the target, never the measured values.
Depth you can audit
870 factors, 856 measured, all reconciled to stable IDs
RankGear catalogues 870 factors across 22 families and measures 856 on a standard run (some factors are minted at runtime and depend on the page set). Families range from JSON-LD structured data (269) and keyword usage & density (101) to headings, entities, links, schema, and technical signals. Each factor is tested for how it correlates with position across the field, using Spearman and Pearson correlation alongside competitor usage, distributions, BM25F, and topical coverage.
Every factor keeps a stable Factor ID across runs, so Diff and Trends can compare the same measurement over weeks — you’re watching one number move, not re-deriving it each time.
Correlation, not causation. A high optimization score means your page now resembles what’s winning the field. It does not guarantee the ranking will move, and RankGear will never tell you it does.
Reading the numbers
What each headline number means
- Optimization score
- How much of the supported deficit your page has closed against the field and your chosen strategy — shown in degrees on an Analysis Run, out of 100 on an On-Page Audit. It is a coverage measure, not a Google quality score or a probability of ranking.
- Correlation (r)
- The strength of the relationship between a factor and position within the measured field. Evidence for where to look first — not a claim that the factor moves rankings.
- BM25 “would rank”
- Where a draft’s relevance sits inside the current comparison corpus. A placement in the field, not a forecast of a live Google position.
- Topical Heatmap coverage
- Exact term coverage versus the field’s consensus, with a documented formula. A comparison, not a density target or an authority score.
Trust
Your data, your keys, your reports
RankGear runs on your Mac or PC. Reports are timestamped files on your own disk — nothing is uploaded. Where a feature needs external data, it uses your provider accounts (SERP, entity, AI); keys live in a protected local store and the interface only ever sees that a key is present, never its value. Generated content and AI interpretation are optional, billed to your own provider account, and always labeled for review — they never become the measurement.
See the evidence in motion on the testing-results blog, where we publish controlled single-variable experiments with their method, sample, and date.
Questions, answered
Does RankGear tell me Google’s algorithm?
No. It measures relationships inside the result set you select. Correlations and competitive patterns are evidence for prioritization — not proof of algorithm weights or causation.
What is a deficit strategy?
The statistical target used to calculate goal, deficit, score, and priorities: field average, maximum after an outlier guard, or the competitor mean plus 1, 1.5, or 2 standard deviations. Your tracked page is excluded from the benchmark.
Where does “870 factors” come from?
It’s the catalogue RankGear measures a page against; 856 are measured on a standard run, with a few minted at runtime depending on the page set. Each maps to a stable Factor ID so runs stay comparable over time.
See the evidence for yourself.
Run one analysis and read the field, the gaps, and the number behind every recommendation.
macOS 14+ (Apple silicon) · Windows 11 (x64/ARM64) · bring your own provider accounts