Skip to content

Discrete Commitment Optimization

Choose among indivisible options or commitments when partial allocation is impossible.

Essence

Discrete Commitment Optimization is the intervention pattern for decisions where the real action is not a smooth allocation but a set of whole commitments. A city does not build 37 percent of a clinic in one neighborhood and 63 percent in another; a team does not assign half a person to a shift if the shift requires a qualified worker; a procurement board cannot buy a fractional vendor contract. The archetype makes those all-or-nothing or whole-number commitments explicit, then selects a feasible bundle under objectives and constraints.

The key move is to stop evaluating each option as if it stood alone. The best feasible bundle may exclude a high-scoring individual option because it blocks a needed dependency, consumes too much shared capacity, violates a coverage rule, or creates an unfair assignment pattern. The archetype therefore treats the combination as the decision object.

Compression statement

When decisions are discrete rather than continuous, optimize the combination of yes/no, integer, assignment, siting, or bundle choices under explicit objectives and constraints instead of pretending that fractional allocation or simple ranking is implementable.

Canonical formula: Given indivisible options I, discrete decision variables x, objective O, coupling constraints C, and feasible combinations F, choose a commitment bundle x* in F that best advances O while preserving hard constraints and implementation legitimacy.

When to Use This Archetype

Use this archetype when choices are indivisible, when whole-number quantities matter, or when assignments and placements must be made as complete commitments. It is especially useful when a simple ranked list keeps producing infeasible or politically contested results because the selected items interact with each other.

It also fits when the decision space grows combinatorially. If every project, site, contract, feature, or assignment can be either selected or not selected, the number of possible bundles can become too large for intuition. The archetype does not require a formal solver, but it does require explicit representation of the choice units, constraints, feasible combinations, and selection rule.

Structural Problem

The structural problem is false continuity. Decision-makers may reason as if options can be partially funded, smoothly adjusted, or independently ranked, while the real world demands whole commitments. A fractional or locally attractive answer then fails at implementation: projects cannot be half-launched, coverage requirements are missed, dependencies are broken, or a rounded answer violates capacity.

A second structural problem is bundle interaction. Each option may look good alone, but the selected set can be poor because choices share budgets, staff, space, deadlines, political legitimacy, or technical dependencies. Discrete commitment problems therefore require combination-level feasibility, not just option-level scoring.

Intervention Logic

The intervention begins by naming the indivisible units: projects, sites, people, shifts, vendors, routes, features, assets, or slots. It then represents each unit as a binary, integer, assignment, or categorical decision. The objective defines what a good bundle means, while constraints define what bundles are feasible.

The decision process then compares feasible bundles. In some domains this is done with an integer programming model or solver. In others it is done through a transparent project-selection matrix, staged narrowing, constraint satisfaction search, or a human review board. The mechanism can vary, but the intervention logic remains the same: choose implementable commitments as a coherent set, not as isolated preferences.

Key Components

Discrete Commitment Optimization is built around the fact that some decisions cannot be averaged or interpolated, so its first job is to represent the choice in implementable units. The Indivisible Option Set names the projects, sites, assignments, or contracts that can only be taken whole. The Binary Decision Variable records in/out for each such option, while an Integer Constraint handles cases where the question is whole-number quantity rather than yes/no. The Objective Function declares what a good bundle improves — coverage, equity, throughput, cost, strategic value — and must remain inspectable because it usually encodes contested judgments. Together these four components reframe the problem so that the decision object is a combination of commitments rather than a ranked list of isolated options.

The remaining components govern which combinations are admissible and how one of them is selected and made real. Coupling Constraint captures the dependencies, exclusions, shared budgets, and minimum-coverage rules that make a bundle stand or fall as a unit; these constraints are precisely why simple ranking can fail. The Combinatorial Feasible Set is the conceptual space of bundles that satisfy all hard constraints, and the Selection Rule — exact solver, heuristic search, review board, or hybrid — picks one feasible bundle and provides enough rationale for the stakes involved. The Feasibility Audit re-checks the chosen bundle against constraints and implementation realities after any rounding or heuristic step, and the Implementation Commitment Record documents what was selected and rejected, by whom, under what assumptions, and when the decision should be revisited so that the lock-in created by discrete choices remains visible and reviewable.

ComponentDescription
Indivisible Option Set ↗The indivisible option set defines what can be selected only as a whole: a project, site, contract, worker assignment, feature, route, or asset. Without this component, the decision can slip back into false fractional thinking.
Binary Decision Variable ↗A binary decision variable records whether a candidate commitment is in or out. It is not the archetype itself; it is a representation component that helps make yes/no commitments explicit.
Integer Constraint ↗An integer constraint requires whole-number quantities. It matters when the decision is not merely yes/no but still cannot be fractional, such as choosing the number of vehicles, servers, beds, staff positions, or batches.
Objective Function ↗The objective function states what the selected bundle is meant to improve. It may be impact, coverage, throughput, equity, cost, reliability, or strategic value. It should remain inspectable because objectives often encode value judgments.
Coupling Constraint ↗A coupling constraint captures interactions among choices: dependencies, mutual exclusions, shared budgets, compatibility, timing, or minimum coverage. Coupling constraints are the reason a simple ranked list can fail.
Combinatorial Feasible Set ↗The combinatorial feasible set is the set of bundles that satisfy all hard constraints. It may be too large to list manually, but it must be conceptually explicit: some combinations are possible and others are not.
Selection Rule ↗The selection rule says how the final feasible bundle is chosen. It may be exact optimization, heuristic search, a deliberative procedure, or a hybrid. The rule must preserve feasibility and provide enough explanation for the stakes of the decision.
Feasibility Audit ↗The feasibility audit checks whether the proposed bundle actually respects constraints, assumptions, and implementation realities. It is especially important after rounding, heuristic search, or solver output.
Implementation Commitment Record ↗The implementation commitment record documents what was selected, what was rejected, why, by whom, under what assumptions, and when the decision should be revisited. Discrete choices often create obligations and lock-in, so documentation is part of the intervention.

Common Mechanisms

Integer programming models and solvers are common mechanisms when the problem can be formalized. They can search large discrete spaces and return candidate solutions, but they do not decide whether the objective, constraints, and candidate universe are legitimate.

Branch-and-bound procedures are search mechanisms. They help explore large decision spaces by pruning regions that cannot improve on current candidates. They implement part of the search logic but should not be confused with the archetype.

Assignment models and matching algorithms implement the assignment variant. They are useful when agents, tasks, slots, or resources must be paired under compatibility and capacity constraints.

Facility location models implement the siting variant. They help choose which sites to open, close, place, or reserve under cost, coverage, access, and capacity constraints.

Project selection matrices and selection review boards are governance mechanisms. They make commitments, constraints, dependencies, and rationales visible for human review, especially when solver precision would be inappropriate or insufficient.

Solver dashboards are interface mechanisms. They help inspect solutions and constraint reports, but they are not substitutes for the archetype’s objective, constraint, feasibility, and accountability logic.

  • Assignment Model
  • Branch-and-Bound Procedure
  • Constraint Satisfaction Search
  • Crew Scheduling Model
  • Facility Location Model
  • Integer Programming Model
  • Integer Programming Solver
  • Project Selection Matrix
  • Selection Review Board
  • Solver Dashboard

Parameter / Tuning Dimensions

Important tuning dimensions include option granularity, constraint hardness, objective weighting, search exactness, commitment reversibility, fairness guardrails, and re-selection cadence. Poor granularity can hide meaningful alternatives; weak constraints can produce infeasible bundles; excessive exactness can make an uncertain model look more authoritative than it deserves.

The most important tuning question is whether the representation matches the implementable action. If the real system can only approve whole projects or assign whole people, the decision representation must preserve that indivisibility.

Invariants to Preserve

The first invariant is indivisibility preservation: the selected answer must be made of units that can actually be implemented. The second is full-bundle feasibility: all selected commitments must coexist under shared constraints.

The third invariant is objective and constraint traceability. Observers should be able to see why this bundle was selected and what rules shaped the decision. The fourth is implementation accountability: selected commitments create obligations, costs, owners, and dependencies that must be visible after selection.

Target Outcomes

The archetype should produce commitment bundles that are feasible, implementable, and better aligned with stated objectives than ad hoc or rank-only selection. It should reduce hidden dependency failures, rounding failures, and bundle-level contradictions.

It should also improve legitimacy where decisions are contested. Even if stakeholders disagree with the final bundle, they should be able to inspect the candidate set, constraints, objective, and selection rationale.

Tradeoffs

Discrete commitment optimization can improve global fit, but it can also reduce transparency if the mechanism is too technical. Exact solvers may produce strong answers for well-modeled problems, but they can launder hidden value judgments when objectives and constraints are not reviewable.

There is also a granularity tradeoff. Smaller units create more flexibility but more complexity; larger bundles simplify the decision but may hide important alternatives. Finally, discrete commitments enable decisive action but may create lock-in and reversal costs.

Failure Modes

A common failure mode is false divisibility: a continuous or fractional solution is accepted even though the real system requires whole commitments. Another is greedy local selection, where the highest-ranked individual options are selected even though a different bundle would perform better.

Hidden constraint violation is especially common in assignment, scheduling, procurement, and public-service decisions. A bundle may satisfy visible budget constraints but violate compatibility, fairness, timing, or implementation constraints. Objective laundering is another risk: a model can hide contested value judgments inside weights, scores, or solver settings.

Neighbor Distinctions

Constrained Resource Allocation distributes divisible resources across uses. Discrete Commitment Optimization chooses whole commitments or whole-number quantities. The question is not just “how much goes where?” but “which commitments are made at all?”

Network Flow Optimization routes flow through nodes and edges subject to capacity and conservation rules. Discrete Commitment Optimization may choose a route, site, or network asset as a whole commitment, but it is not primarily about continuous flow allocation.

Bounded Search Pruning is a search logic that may support this archetype. It removes branches that cannot dominate. Discrete Commitment Optimization is the broader commitment-selection pattern that may use pruning, solvers, heuristics, or review boards.

Priority-Based Admission orders access to limited opportunities or capacity. Discrete Commitment Optimization selects a feasible bundle or assignment set, often with multiple coupling constraints beyond priority ordering.

Assignment / Matching Optimization is recorded here as a candidate variant and possible second-wave promotion candidate. It should not be drafted separately until saturation review confirms that its matching-specific components and failure modes require their own archetype.

Cross-Domain Examples

In capital budgeting, a government selects a feasible bundle of infrastructure projects under budget, coverage, mandate, and contractor-capacity constraints. In workforce operations, a hospital assigns nurses to shifts while preserving skills, rest requirements, and minimum coverage.

In product strategy, a team chooses a feature release bundle where dependencies and shared engineering capacity matter. In procurement, a company selects vendor contracts or license bundles that are indivisible and mutually constrained. In education, a school assigns students to course sections while preserving prerequisites, capacities, and fairness rules.

Non-Examples

A proportional budget split across departments is usually Constrained Resource Allocation unless the split funds indivisible projects. Buying an integer-programming solver is not this archetype; it is acquiring a mechanism.

Rejecting an option because it violates a categorical safety or legal rule is not optimization; it is feasibility exclusion. Routing continuous flow through a network is not this archetype unless the main decision is selecting whole routes, facilities, or assets.

Abstractions this archetype builds on — directly (a source ingredient) or as a related pattern. Links follow the typed catalog namespace.

Built directly on (3)

Also references 6 related abstractions

Variants

Narrower or domain-specific specializations that share this archetype's core structure. Recognized variants are established; candidate variants are provisional.

Binary Project Selection · subtype · recognized

Choose which indivisible projects, proposals, features, or initiatives to accept when each candidate is essentially yes-or-no.

  • Distinct from parent: The parent also covers integer quantities and assignments; this variant centers on binary inclusion/exclusion.
  • Use when: {'condition': 'Each option must be accepted, rejected, funded, built, hired, purchased, or approved as a whole.'}; {'condition': 'The number of possible bundles is large enough that simple ranking may miss feasible or better combinations.'}; {'condition': 'Budget, staffing, timing, dependency, eligibility, or risk constraints bind across the bundle.'}.
  • Typical domains:
  • Common mechanisms: Project Selection Matrix, Integer Programming Model

Integer Quantity Commitment · subtype · recognized

Choose whole-number quantities of units, slots, staff, licenses, batches, or assets when fractional quantities are impossible or meaningless.

  • Distinct from parent: The parent covers all discrete choices; this variant emphasizes integral quantities rather than binary selection.
  • Use when: {'condition': 'The decision variable can take more than two values but only in whole units.'}; {'condition': 'Unit counts interact with capacity, staffing, timing, compatibility, or minimum-scale constraints.'}; {'condition': 'Rounding a continuous allocation would produce infeasible or misleading commitments.'}.
  • Typical domains:
  • Common mechanisms: Integer Programming Model, Capacity Sizing Model

Assignment / Matching as Discrete Commitment · subtype · candidate

Assign agents, tasks, slots, resources, or roles to one another when each match is a discrete commitment constrained by capacity and compatibility.

  • Distinct from parent: The parent covers discrete commitments generally; this variant focuses on pairings or slot assignments.
  • Use when: {'condition': 'Entities must be paired, matched, or assigned rather than fractionally allocated.'}; {'condition': 'Compatibility, capacity, fairness, availability, or preference constraints shape feasible matches.'}; {'condition': 'Naive first-come or local matching creates overload, mismatch, or unfairness.'}.
  • Typical domains:
  • Common mechanisms: Assignment Model, Matching Algorithm

Facility Location Commitment · domain variant · recognized

Choose which physical or logical sites to open, close, place, or reserve when each site is an indivisible commitment with network effects.

  • Distinct from parent: The parent is not inherently spatial; this variant adds location, coverage, and access structure.
  • Use when: {'condition': 'Potential sites are discrete and cannot be partially opened in a meaningful way.'}; {'condition': 'Each selected site changes coverage, access, cost, routing, or downstream capacity.'}; {'condition': 'Siting decisions have durable implementation consequences.'}.
  • Typical domains:
  • Common mechanisms: Facility Location Model, Site Selection Review

Near names: Integer Optimization, Integer Linear Programming, 0-1 Optimization, Combinatorial Selection, Assignment Problem, Project Portfolio Selection, Facility Location, Integer Programming Solver, Assignment Algorithm.