Make outcome quality, calculation method and attribution inspectable

Answers Ontology · Derivation · Epistemology · Semantics · Deontics

An outcome or efficiency claim needs declared inputs, exclusions, quality controls, attribution window and an activation join key. Metric meaning is distinct from its versioned Formula. A Formula consumes governed Signal or Metric inputs and must not split a Metric across a second direct-input path. An attributing Metric names the Attribution Model it divides credit by, because the rule is part of the method rather than a presentation choice.

What goes wrong without it

An outcome number is only as good as the inputs, exclusions and window that nobody wrote down. Media measurement standards exist because outcome claims circulate detached from their method, and a figure with no stated method cannot be challenged, reproduced or compared against anyone else's. Keeping what a measure means apart from how it is computed is what makes either one reviewable.

What proves it — 1

Each is a gate the build runs: a check inside a ring, a whole ring, or an eval beside them. A build stops at the first that refuses.

• registry integritycalculation and projection integrity
  Formula inputs, metric composition, cardinality, and plan projections satisfy their declared contracts.

Derived from — 8

• Media Rating CouncilOutcomes and Data Quality Standards
• Kimball GroupDeclaring the Grain
• IABMeasuring Visibility in the AI Era
• Shao and Li, KDD 2011Data-driven multi-touch attribution models
• Dalessandro, Perlich, Stitelman and Provost, ADKDD 2012Causally motivated attribution for online advertising
• Gordon, Zettelmeyer, Bhargava and Chapsky, Marketing Science 38(2), 2019A Comparison of Approaches to Advertising Measurement: Evidence from Big Field Experiments at Facebook
• Herbert E. Krugman, Journal of Advertising Research 12(6), 1972Why Three Exposures May Be Enough
• Media Rating Council, 2017Digital Audience-Based Measurement Standards

Kinds it licenses

Relationships it licenses

counts in numerator · counts in denominator · takes as numerator · takes as denominator · computed from formula · takes input · measured by · attributed to tactic · captured on · implemented in platform · attributed by · improves when · reads as

Where this model stands

Attribution to a tactic is optional and evidential; interpretation runs Signal → Metric → KPI.

A Formula names governed inputs, and a Metric takes either direct inputs or a Formula, never both.

A Metric or KPI whose source does not yet exist says Requires source rather than inheriting proof from a neighbouring calculation.

Where a measurement standard names a reading this model already takes, the model takes the standard's name for it, so a number leaving here can be set beside one that came from somewhere else.

A measure that attributes names the rule it divides credit by. The same outcome under two rules gives two different numbers, and a figure carrying no rule cannot be set beside any other.

Where a single-touch rule is used, the measure says so in words as well as in the edge, because other exposures almost certainly contributed and a number that stays silent about that reads as though they did not.

Every KPI states which way it has to move to be a better result, and what arithmetic form its reading takes. Both were carried only in the prose of a definition before, which a reader could follow and nothing else could: whether a rising cost or a rising hallucination rate is a gain is now a fact in the graph rather than an inference from a sentence.

Attribution divides credit among exposures by a stated rule; it does not establish that an exposure caused the outcome, and observational attribution has been shown to differ widely from experimental lift, so an attributed figure is read as a method's output and never as incrementality.

Reach and frequency are read as the audience standard defines them, and a frequency reading is judged against a band rather than a direction, because exposure below the band does not register and exposure above it buys nothing.

What it does not claim

A case-backed activation key and structured metric contract are required before attributed revenue, modelled conversion, ROI or efficiency claims become evidenced.

The lookback window stays undeclared. It has to be empirically supported and logical for a campaign's objective, length and buying cycle, which makes it a fact about a campaign rather than a value this model can fix in advance; declaring it is the first thing a real measurement contract adds.

RUL-007