Vocabulary
The terms used in these notes, and where each one comes from.
Why this page exists
A method is only adopted if it can be spoken about. These notes therefore use a small, fixed vocabulary, and this page is where each term is defined.
Two disciplines govern it. Coin sparingly — a name is claimed only where no adequate one exists. Never rename — where a concept already has an established name, that name is used unchanged, however tempting a fresher one might be. Renaming a seventy-year-old idea is the fastest way to lose the readers who know the literature, and it is not a contribution.
Each entry states whether the term is new, adapted, or standard.
Judgment — how priorities are established
Declared weight
Standard · ratio-scale weights follow from AHP (Saaty, 1977)
The relative importance an organization has stated for a strategic driver, arrived at through structured comparison rather than typed in directly. Declared weights are ratio-scale: a weight of 40 against a weight of 8 asserts five times the importance, not merely a higher rank.
The range
Adapted · sensitivity intervals are standard in decision analysis; reporting them on an executive surface is not
The interval within which a declared weight sits, given the precision of the judgments that produced it. Every weight has one; almost no tool reports it.
Calibration price
New
The cost of departing from a judgment you have already declared — expressed, recorded, and shown at the moment of departure. Overriding a declared weight is permitted; doing so silently is not.
Selectivity
Adapted · used here critically
The practice of applying an exponent or curve to declared weights to sharpen or flatten the resulting priorities. Selectivity preserves rank order while destroying ratio meaning, which is why these notes treat it as a defect rather than a feature.
Selection — how a portfolio is chosen
Distance to flip
Adapted · this is a stability radius, standard in parametric optimization
How far the world must move before an initiative's status changes — expressed in the units of whatever moved: dollars of cost, dollars of envelope, or shift in declared priority. What is proposed is its place on an executive screen, not the measure itself.
The boundary
New as a presentation · the underlying sensitivity is standard
The region of a portfolio where the decision is genuinely uncertain, comprising four bands: held, borderline in, borderline out, and out. A portfolio boundary is a zone, not a line.
Priced lever
New
A stated, finite change that would alter an initiative's status: reduce its scope by a given percentage, increase the envelope by a given amount, or force it in and accept named displacement. Ratings are never a lever.
Extensive and intensive quantities
Standard · borrowed from physics and thermodynamics
An extensive quantity scales with the size of the portfolio — total value, total cost. An intensive quantity does not — value per dollar, a ratio. The objective coefficient of a selection problem must be extensive. If what you want to maximise is a rate, it belongs in a constraint.
Pipeline supply
New as a term · the underlying comparison is not
The share of proposed spend that actually stands behind each declared driver, compared against that driver's declared weight. A driver with declared weight and no initiatives exists on paper only.
Decision — how the choice is made and recorded
Decision surface
New
A screen whose purpose is to support a specific executive decision, presenting the properties of the solution rather than the properties of the inputs. A ranked list of scores is not a decision surface; it is a readout.
The desk
New as used here
The place where a committed decision is changed. Every amendment is priced before it is accepted, and the acceptance belongs to the person accountable, not to the software.
Inference firewall
New
The separation between what is computed and what is explained. Mathematics decides; language describes. No recommendation originates in a generative model.
Decision record
Adapted · from ordinary governance usage, with the last clause added
The durable account of what was decided, under what constraints, against which declared strategy, and what would have had to change for the outcome to differ.
Allocation and automation — how money moves and who decides
Maintain, deliver, build, retire
New as used here · an alternative to Run-Grow-Transform
What an investment's money does: keeps an asset running, delivers a change, builds a lasting capability, or retires something. Each investment is tagged once, and the split among the four comes out of the optimization rather than being set in advance.
Container
New as used here
Anything that receives an envelope of money or capacity before projects are chosen inside it: a business unit, a program, a capability, a value stream. Allocation happens at two levels, to containers and to projects, and prices link them.
Augmentation, supervised, autopilot
Adapted · from the levels of automation (Sheridan and Verplank, 1978)
Three ways a decision can be made with a machine. In augmentation, people decide and the machine informs. Supervised, the machine proposes and people review where the margin is thin. On autopilot, the machine prepares the decision within rules and people are called in when a rule trips. The mode belongs to the decision, not to the software.
Terms deliberately not coined
Several concepts in these notes already have established names, and those names are used unchanged:
- Opportunity cost — the value of the best alternative given up. A shadow price is the opportunity cost of a limit, measured exactly.
- Shadow price — the value of relaxing a binding constraint by one unit. Standard in linear programming since the 1940s. It is not renamed here, and any note discussing constraint economics uses this term.
- Dual value — used interchangeably with shadow price where the formal context calls for it.
- Sensitivity analysis — the general practice from which distance to flip is a specific presentation.
- Efficient frontier — the set of non-dominated portfolios.
- Knapsack problem — the formal structure of constrained selection.
- Algorithm aversion — the documented reluctance to accept algorithmic recommendations, established in the behavioural literature.
Cite as: Gruia, M. (2026). Vocabulary. Frontnovo Method Notes.