Clarifying the Decision

Definition

Most failed measurement efforts fail before any data is collected: the "thing to measure" was never tied to a decision it would actually change. Hubbard's fix is to require that any real decision have four properties — two or more realistic alternatives, genuine uncertainty about which is best, potentially negative consequences (including opportunity loss) if the wrong one is chosen, and an identifiable decision maker. If a proposed measurement can't be traced to a decision meeting these criteria, the request to "measure X" is premature — the real first question is "what is your dilemma?"

In the Book

Chapter 4 opens the five-step Applied Information Economics (AIE) method with "define a decision problem and the relevant uncertainties" as step one, ahead of quantifying uncertainty, computing information value, or measuring anything. Hubbard catalogs the ways managers dodge this: asking to "measure project performance" without saying what action a progress reading would trigger; wanting to measure "the value of IT" or "the value of clean drinking water" as if abandoning IT or water were a real alternative (a false dichotomy); or wanting to "just start measuring" the way an impatient client wants to start pouring concrete before the architect has picked a building type. He illustrates the fix with a 2013 CGIAR (World Agroforestry Centre) project: soil scientist Dr. Keith Shepherd's team had been measuring biodiversity and drought resistance metrics for their own sake, and only after building explicit "impact pathway" decision models — e.g., whether to fund small dams versus large dams — did they discover which metrics actually had value. He also discusses corporate dashboards as a recurring failure mode: dials and charts assembled without a predefined action threshold, so managers either miss the moment to act or waste time improvising a response that could have been worked out in advance.

Why It Matters

Tying measurement to a specific, well-formed decision is what lets you later compute whether a measurement is worth its cost (its information value) — a question that is meaningless without a decision to inform. It also short-circuits the most common form of wasted analysis: elaborate dashboards, KPI programs, and research efforts that produce numbers nobody's choices depend on. The discipline of asking "what would you do differently if you knew this?" applies to any field that collects data before it has decided what the data is for.