{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "$id": "/schemas/enums/attribution-methodology.json",
  "title": "Attribution Methodology",
  "description": "How attribution between ad exposure and outcome events was computed. Used as a `qualifier.attribution_methodology` key on `committed_metrics`, `missing_metrics`, package-level delivery values (`metric_values` or `vendor_metric_values`), and `performance-feedback.metric` to disambiguate the same outcome metric reported under different methodologies — `conversion_value` measured deterministically (matched purchase IDs) is not the same number as `conversion_value` measured probabilistically (modeled match) and should never be summed across methodologies. The retail-media closed-loop pattern typically reports under `deterministic_purchase`; MMM and clean-room outputs typically report under `modeled` or `probabilistic`; panel-based measurement (Nielsen, comScore, Edison) reports under `panel_based`.",
  "type": "string",
  "enum": ["deterministic_purchase", "probabilistic", "panel_based", "modeled"],
  "enumDescriptions": {
    "deterministic_purchase": "1:1 deterministic match between exposed users and purchasers, typically via shared identifier (loyalty card, logged-in account, hashed email). Standard in retail-media closed-loop measurement (Walmart Connect, Kroger Precision, Amazon DSP attribution).",
    "probabilistic": "Probabilistic match using statistical models over device graphs, IP/UA fingerprinting, or behavioral signals. No 1:1 identifier link.",
    "panel_based": "Panel methodology — outcomes measured against a recruited respondent panel and projected to the population. Standard for Nielsen, comScore, Edison Research, and similar panel providers.",
    "modeled": "Modeled output — Media Mix Modeling (MMM), Multi-Touch Attribution (MTA), incrementality testing, or other statistical inference that produces an estimated outcome rather than a directly-observed match. Includes most clean-room outputs."
  }
}