{
  "slug": "practical-effect-magnitude",
  "name": "Practical magnitude",
  "aliases": [
    "Effect size",
    "Minimum detectable effect",
    "Confidence interval"
  ],
  "definition": "How large a measured change is in terms that matter to a decision, an effect size, a minimum detectable effect, or an interval around the estimate.",
  "category": "Statistics",
  "type": "statistical verdict",
  "unit": "not produced",
  "direction": "neutral",
  "verification": "review",
  "grain": "Would be one value per test if it were produced.",
  "dimensions": [],
  "requiredFilters": [],
  "sources": [
    "search-console",
    "web-analytics",
    "google-ads",
    "rank-tracking",
    "lighthouse"
  ],
  "reports": [
    "ai-referral-traffic-report",
    "cannibalization-report",
    "content-decay-report",
    "content-performance-report",
    "internal-linking-report",
    "keyword-ranking-report",
    "monthly-executive-search-report",
    "paid-organic-overlap-report",
    "paid-search-performance-report",
    "search-market-share-report",
    "search-to-revenue-report",
    "serp-volatility-report",
    "weekly-search-report"
  ],
  "skills": [
    "significance-referee",
    "anomaly-investigator"
  ],
  "verbs": [],
  "rungs": [
    "R5"
  ],
  "levers": [],
  "questions": [
    "Is this change big enough to act on?",
    "What is the confidence interval?",
    "What change could we even have detected?"
  ],
  "caveats": [
    "No tool produces an effect size, a minimum detectable effect or a confidence interval.",
    "The absolute change and the relative change are available and are the closest thing to a magnitude read, but neither carries uncertainty around the estimate.",
    "No multiple-comparison correction is applied anywhere either, so a batch of significant results has no adjusted view to fall back on.",
    "When a magnitude question is asked, the answer is what exists, the change, the sample size and the verdict, plus a plain statement that the sizing statistics are not available."
  ],
  "freshness": "varies by site, priority URLs can run nightly; most pages far less often. The collection date rides on the card rather than being assumed.",
  "failureModes": [
    "Presenting a small p-value as evidence that a change is large.",
    "Computing an interval or an effect size outside the tools and presenting it as a product output.",
    "Reading the significance threshold as a minimum effect the test was designed to detect."
  ],
  "notSameAs": [
    {
      "slug": "significant-flag",
      "why": "The flag is the p-value below a threshold and nothing else. It carries no effect size, and none is computed anywhere, so significance can never stand in for magnitude."
    },
    {
      "slug": "significance-p-value",
      "why": "A p-value answers whether a difference is distinguishable from noise. Magnitude answers whether it is worth acting on. The tools produce the first and not the second."
    },
    {
      "slug": "real-noise-verdict",
      "why": "Real means distinguishable from noise, not large enough to act on. No effect size is computed anywhere, so the verdict says nothing about magnitude.",
      "mirroredFrom": "real-noise-verdict"
    },
    {
      "slug": "pearson-correlation",
      "why": "A coefficient describes the tightness of a relationship, not its business size, and no interval is computed around it.",
      "mirroredFrom": "pearson-correlation"
    }
  ],
  "related": [
    "significance-p-value",
    "significant-flag",
    "real-noise-verdict",
    "rate-change-percent"
  ],
  "workflows": [],
  "lastVerified": "2026-08-04"
}