Scores and formulas

Glimana shows its working. Every number below is the actual configuration the engine runs with, so you can predict what a change will do before you make it.

SEO score

A weighted blend of six parts, each 0–100:

A part that cannot be measured is left out and the remaining weights are rescaled, so a new site without CrUX data is not punished for missing data. The grade words: 85+ Excellent, 70–84 Good, 50–69 Fair, below 50 Poor.

Health score

Each open issue carries a severity weight — critical 10, high 5, medium 2, low 0.5, info 0. The weighted sum is divided by the number of indexable pages and passed through a curve with k = 60, so that health = 100 × 60 ÷ (60 + weighted issues per 100 pages). Informational findings never move the score. Ignored rules and rules marked not running are excluded.

Task priority

(impact + 10) × confidence × urgency ÷ effort, then × the site's task weight for the family.

• impact = measured organic clicks of the affected pages, last 28 days
• confidence: issue 0.9 · site health 0.95 · CTR opportunity 0.7 · decay 0.6 · internal link 0.6 · cannibalization 0.5 · AI visibility 0.4 · custom 0.5
• urgency: 1 normal · 1.2 due soon · 3 site health
• effort: S 1 · M 2.5 · L 6

A closed task is not recreated for the same rule and page for 90 days.

Expected CTR and CTR opportunities

Glimana fits a CTR-by-position curve from your own queries with at least 50 impressions and blends it with a global curve (position 1: 27.8%, 2: 15.4%, 3: 9.9%, 4: 6.7%, 5: 4.9% … 10: 1.7%, 20: 0.6%) when a position has fewer than 30 of your data points. A query is a CTR opportunity when it has ≥ 500 impressions, sits in positions 4–10 (the top 3 are skipped because SERP features explain most gaps there), its CTR is below 70% of expected, the gap is at least 1 point and the potential gain is at least 20 clicks. At most 5 pages per week are proposed so the list stays actionable.

Content decay

A page is flagged when, over a 28-day window, its clicks are at least 35% below a peak window of at least 100 clicks, the peak window ended at least 28 days before the current one, and the drop is at least 15 points worse than the site's own change. Pages younger than 90 days are not evaluated; a seasonality check needs 396 days of history and the task says when that history is not available. The diagnosis compares position (worse by ≥ 1), impressions (down ≥ 20%), CTR (down ≥ 20% with ≥ 100 impressions) and lost queries (≥ 25% of the page's queries gone) to name the most likely cause.

Cannibalization

Two or more pages of the site swap the lead for the same queries: ≥ 300 impressions in 90 days, ≥ 3 lead changes, pages within 5 positions of each other, and the trailing page holding ≥ 10% of the clicks. Groups with fewer than 10 clicks in 90 days are collapsed into one informational line. Intent-protected pairs (such as amount or location variants of the same tool) are excluded.

Traffic anomalies

Weekly organic clicks are compared with an eight-week baseline. A drop alert needs a z-score below −3 and a relative drop of at least 25% on at least 50 clicks; a spike needs z above 4. Discover uses z below −4.5 and a 50% drop. The last two days are skipped because Search Console has not published them.

Internal PageRank

Computed over the site's own links with damping 0.85 for 30 iterations. Links are weighted by where they sit: body content 1.0, navigation 0.7, sidebar 0.5, footer 0.3. The internal-link suggestions use it together with text similarity (≥ 0.75) and suggest at most 5 links per page and 50 per site per week, targeting pages that rank between positions 4 and 15.

Core Web Vitals thresholds

LCP good ≤ 2.5 s, poor > 4.0 s · INP good ≤ 200 ms, poor > 500 ms · CLS good ≤ 0.10, poor > 0.25. All at p75 over 28 days, from CrUX.

Content thresholds

Thin content: under 300 words of main text. Title too long: wider than 580 px in Google's result font. Description too long: wider than 920 px. Near-duplicate pages: SimHash distance ≤ 6 (≤ 3 counts as a copy).