Glossary

The current reference scenario. When you compare scenarios, changes are measured relative to the baseline. You can promote any scenario to become the new baseline.

An edge that is either fully on or fully off — the solver treats it as a 0/1 integer variable. Use binary decisions for on/off choices like "open this line" or "use this supplier."

A constraint that is fully used up in the optimal solution. It has zero slack. Relaxing a binding constraint improves the objective; relaxing a non-binding one does nothing.

The binding constraint with the highest impact on the objective. SSPLAX identifies bottlenecks using shadow prices and stress testing.

A constraint limiting total CO₂ emissions across the network. Each edge can carry a carbon rate (tCO₂ per unit of flow).

The tonnes of CO₂ emitted per unit of flow on an edge. Feeds into carbon cap constraints and the carbon minimization objective.

A plain-language warning about model quality, generated from the assumption ledger. Caveats flag missing sources, low confidence, or placeholder values.

What flows through edges. Usually "throughput" but can be any named quantity. The objective and constraints reference commodities.

A rating (High, Medium, Low, Unspecified) on each constraint assumption indicating how reliable the number is.

A limit on the system: a budget cap, capacity ceiling, minimum delivery, or resource limit. Constraints are what make an optimization problem non-trivial.

USD per unit of flow on an edge. When minimizing cost, the solver routes flow through the cheapest paths. Cost rates also feed into budget constraints.

Piecewise-linear cost curve on an edge. Each segment defines a breakpoint and cost rate. The solver picks the optimal volume on each segment.

A structured summary of the recommended plan, binding constraints, scenario comparison, sensitivity rankings, and caveats. Designed for sharing with stakeholders.

A narrative explanation of why the solver chose one plan over another, including which constraints and flows drove the choice.

A flow path between two nodes. Carries a commodity with min/max bounds, cost rate, carbon rate, yield ratio, and optional resource consumption.

A model is feasible when at least one plan satisfies all constraints simultaneously. Infeasible means the constraints contradict each other.

The quantity moving through an edge in the optimal solution. The solver decides flow values to meet the objective while respecting constraints.

Synonym for slack — how much room is left before a non-binding constraint becomes binding.

A natural-language what-if question that SSPLAX translates into constraint and edge changes, then re-solves to show the impact.

A 0-100 sensitivity score showing how strongly the result depends on a constraint. Confidence, source, and value type are shown separately and used to flag validation priorities.

No plan can satisfy all constraints. SSPLAX computes the minimum relaxation: ranked single-limit relaxations that could restore feasibility.

A suggested constraint or edge change that would improve the outcome. Interventions are named actions tagged by the part of the system they affect.

Synonym for shadow price — how much the objective improves per unit of constraint relaxation.

When a model is infeasible, ranked single-limit relaxations that could restore feasibility. Each option shows one constraint that could be changed, by how much, and in which direction.

Solving the same network across multiple time periods with demand growth, ramp rates, and optional inventory carry-over between periods.

A stage in your system — a source, service, queue, or sink. Nodes are connected by edges.

What you want to optimize: maximize throughput, minimize cost, or minimize resource usage. The solver finds the best feasible plan for the chosen objective.

The set of efficient tradeoff points between two competing objectives. Each point on the curve represents a plan where you can't improve one objective without worsening the other.

An assumption value type indicating the number is a rough guess that needs validation before the model can support a real decision.

A normalized sensitivity metric (0-100). Measures how much the objective drops when a constraint is tightened under stress. Higher = more sensitive.

A region in parameter space where the same set of constraints is binding and the solver recommends the same structural plan. Regime boundaries are where small changes flip the strategy.

A 2D sweep over two constraints. Each point is colored by its regime. Shows how the optimal strategy changes across the parameter space.

A custom consumable attached to edges (FTEs, GPU hours, water, energy). Total consumption can be capped with a resource_limit constraint.

How much the objective improves per one-unit relaxation of a constraint. A high shadow price means this constraint is expensive to keep tight. Only valid within a range.

A node where flow exits the system. Represents delivered output, final destination, or completed work.

How much room is left before a non-binding constraint becomes binding. Zero slack = the constraint is active.

A node where flow enters the system. Represents raw material, inbound demand, or the starting point of a process.

Monte Carlo simulation that draws constraint values from uncertainty ranges and checks how often the plan stays feasible. Reports a feasibility rate and which constraints cause failures.

Tightening each constraint by a small percentage and re-solving to see how the objective degrades. Identifies which constraints are most dangerous when they move.

The structural definition of a model: nodes, edges, constraints, objectives, and resources. Templates can be starter templates (pre-built) or custom.

A horizontal bar chart ranking constraints by impact. Longer bars = higher leverage. Combines shadow price and stress test results into a normalized pressure score.

A chart showing how the objective changes as you sweep a single constraint from tight to loose. Shows the marginal return of relaxation.

The range of constraint values over which a shadow price holds. Beyond this range, a different constraint becomes binding and the shadow price changes.

The fraction of input that becomes output on an edge. A yield ratio of 0.72 means 28% is lost. Output = input × yield.