Formulas and Health Thresholds: A Working Reference
These are planning ranges, not laws. They come from public industry studies and from patterns that repeat across sites, but a single figure can be off by a wide margin in a specific sector or country. Use them to spot outliers that deserve a closer look, then replace each one with your own baseline within a quarter.
Traffic and ranking metrics
| Metric | Formula | Typical healthy range | How to read it |
|---|---|---|---|
| Click-through rate | Clicks ÷ impressions × 100 | Roughly 25–32% at position 1, 8–10% at 3, 1.5–2% at 10 | Well below the norm for the same position means the title and description are being ignored |
| Engagement rate | Engaged sessions ÷ total sessions × 100 | Above 60% B2B and ecommerce, above 50% publishing | GA4’s primary quality signal; a sharp drop usually means intent mismatch |
| Bounce rate | 1 − engagement rate | Under 40% ecommerce, under 55% publishing | The inverse of the metric above; useful only as a trend |
| Pages per session | Views ÷ sessions | Above 2.0 content sites, above 3.0 ecommerce | Tests whether internal linking is actually working |
| Average engagement time | Engaged time ÷ engaged sessions | Above 90 seconds long-form, above 45 seconds product pages | Weak on its own; strong when it moves against a control page |
| Conversion rate | Conversions ÷ sessions × 100 | Roughly 1.5–3.5% ecommerce, 1–5% B2B lead capture | Always calculate per landing-page type, never site-wide as an average |
| Keyword coverage | Ranking keywords ÷ target keyword set × 100 | Should trend upward monthly | Shows whether the keyword strategy is being executed or just written |
| Top-ten share | Keywords in top ten ÷ all ranking keywords × 100 | Should trend upward monthly | Separates ranking breadth from ranking quality |
Core Web Vitals and technical health
| Metric | Threshold | Where to measure |
|---|---|---|
| LCP | Good under 2.5s, needs work 2.5–4s, poor above 4s | PageSpeed Insights, Search Console |
| INP | Good under 200ms, needs work 200–500ms, poor above 500ms | PageSpeed Insights, Search Console |
| CLS | Good under 0.1, needs work 0.1–0.25, poor above 0.25 | PageSpeed Insights, Search Console |
| TTFB | Under 600ms | PageSpeed Insights, GTmetrix |
| Index rate | Indexed pages ÷ submitted pages above 80% | Search Console Pages report |
| Crawl volume | Steady or rising; a sudden drop is the signal, not the absolute number | Search Console Crawl Stats |
Core Web Vitals are field metrics measured on real devices. A perfect lab score with poor field data describes a fast site for nobody.
Links
| Metric | Typical range | Why it matters |
|---|---|---|
| Referring domains | Upward trend month over month | Far more resistant to manipulation than raw link count |
| Branded anchor share | Roughly 30–40% | Too high suggests an unnatural profile; too low suggests nobody knows the brand |
| Exact-match anchor share | Under 5–10% | Concentrated exact-match anchors are the classic over-optimisation signal |
| Dofollow share | Roughly 70–85% | A profile that is entirely dofollow looks engineered |
| Toxic link share | Under 5% | Above 10% is worth a manual review before considering disavow |
Treat all five as diagnostics. Anchor-text ratios describe how other people chose to write about you; they are not inputs you should be tuning directly.
Generative engine metrics
| Metric | Definition | Planning target |
|---|---|---|
| Mention rate | Prompts where the brand appears ÷ prompts tested × 100 | Upward trend; a first-year goal of 40% on core category prompts is ambitious but realistic |
| Citation rate | Answers linking to your domain ÷ prompts tested × 100 | Upward trend, tracked separately from mentions |
| Share of model | Your mentions ÷ all competitor mentions × 100 | Top three in the category |
| Net sentiment | Positive descriptions − negative descriptions | Above +60% |
| AI referral sessions | Sessions from chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com | Upward trend, reviewed against total organic |
| Branded search lift | Change in branded query volume during a GEO campaign | Should move in the same direction as mention rate |
| Attribution accuracy | Factually correct claims ÷ claims made about you | Above 95%; anything lower is a correction task, not a content task |
AI referral traffic is usually small in absolute terms and unusually high in conversion rate, because the assistant has already done the qualification. Judge it by revenue per session, not by volume.
Commercial metrics
| Metric | Formula | Note |
|---|---|---|
| SEO ROI | (Organic revenue − total SEO cost) ÷ total SEO cost × 100 | Total cost includes salaries, tools, content, and link acquisition |
| Organic CAC | Total SEO cost ÷ new organic customers | Should fall over time; if it doesn’t, the content is not compounding |
| Revenue per session | Organic revenue ÷ organic sessions | The cleanest single measure of traffic quality |
| LTV to CAC | Customer lifetime value ÷ acquisition cost | Three to one is the conventional floor for a sustainable channel |
When organic CAC rises while spend stays flat, the usual cause is a content library that has stopped growing. Compounding requires new contributions; an ageing library decays as competitors publish.