A metric exists because a question needs it

Answers Modality · Derivation · Epistemology · Deontics

An Intelligence row is a standing question area that sits between a Goal and the Metrics that answer it, following Goal-Question-Metric. A Signal informs a question area by contributing evidence; it is not classified by it. The kind of a conversion - a first, a repeat, a return - is read from the condition the move it triggers leaves, and is never declared beside the event as a second fact. A question area advises the campaign purposes whose campaigns act on its answers, so measurement reaches the decision it exists to inform.

What goes wrong without it

Goal-Question-Metric was formulated against a specific failure: measurement chosen because a system already produces it, rather than because a stated question needs it. The order it insists on is goal, then question, then measurement. Without the middle term a dashboard grows without limit, and which of its numbers would change a decision stops being answerable.

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 integritytyped relationships
  Every relationship has declared endpoint kinds and a governed cardinality.

Derived from — 4

• Basili, Caldiera and Rombach, University of MarylandThe Goal Question Metric Approach
• Media Rating CouncilOutcomes and Data Quality Standards
• Kimball GroupDeclaring the Grain
• Basili, Trendowicz, Kowalczyk, Heidrich, Seaman, Münch and Rombach, Springer, 2014Aligning Organizations Through Measurement: The GQM+Strategies Approach

Kinds it licenses

Relationships it licenses

informs intelligence · is evaluated by · answers goal · advises purpose

Where this model stands

Every Intelligence row names at least one Metric that evaluates it, so no question area is stated without the means to answer it.

Signal to Intelligence is an Association carrying 1..*: a signal may inform several question areas without being typed by any of them.

A conversion is read as a first, a repeat or a return from the move it triggers, so a count of conversions by kind is a reading over moves, never a second event kind and never a vocabulary storing what the machine derives.

Every question area names the goal or goals it answers, so Goal-Question-Metric is a path the reader can walk end to end rather than a shape the model only describes: a goal reaches its question areas, and each of those reaches the metrics that answer it.

Every question area advises at least one campaign purpose and every purpose is advised by at least one, so the loop the model draws closes through a reading rather than a picture: what a goal's questions answer is what decides the campaigns that act on the goal, which is the alignment GQM+Strategies adds to Goal-Question-Metric. The advice is stated once, from the question area to the purpose; a move no longer recommends a purpose and a purpose no longer names the conditions it may address, because both were opinions about causation stored where facts belong.

A goal is an outcome the effort exists to advance, never a way of reading one. Two rows that named methods of reading - connecting actions to their source, and dividing revenue among touches - stood among the goals; they were question areas wearing a goal's name, and the readings that served them now serve the outcomes they bound: conversion and efficiency.

RUL-013