Dominance Screening¶
Screening method — instantiates Pareto Frontier Navigation
Removes every option that another feasible option matches or beats on all objectives, shrinking a listed field to its non-dominated set before any value choice is made.
Before anyone debates which option is best, some of the options on the table are simply worse — beaten by another option on every dimension that matters. Dominance Screening is the purely negative first pass that throws those out. Working from an enumerated list of candidates already scored on each objective, it checks, pair by pair, whether some other feasible option is at least as good on every objective and strictly better on at least one; if so, the loser is dominated and removed. What survives is the non-dominated set — the frontier — and nothing more is claimed. Its defining discipline is restraint: it decides only what to eliminate, never what to prefer. That is exactly why it is powerful early — it clears away the inferior compromises that clog a decision without touching the genuine value question of which surviving sacrifice profile to accept.
Example¶
A mid-size firm runs an RFP to replace its warehouse-management software and gets eleven vendor bids. Each is scored on five objectives: licensing cost, functional fit to their picking workflow, implementation time, vendor support rating, and security posture. Before Dominance Screening, the selection committee applies the feasibility gate: two bids are dropped outright because they cannot meet a hard data-residency requirement — they were never eligible, efficient or not. That leaves nine.
Now the screen runs. Bid F costs more than Bid C, fits worse, takes longer to implement, and has a weaker support rating — C beats it on four objectives and ties the fifth, so F is dominated and removed. Two more bids fall the same way, each strictly beaten by a cheaper-and-better rival. Four bids resist elimination: the cheapest one has mediocre fit; the best-fitting one is expensive; a third is fastest to deploy but weaker on security; a fourth is balanced but slow. None of these four beats another across the board, so all four stay. The committee walks out with four real contenders instead of nine, and — crucially — with the knowledge that the five it dropped were not judgment calls but strictly inferior. The hard argument about cost-versus-fit is still ahead; the screen just made sure it happens over options worth arguing about.
How it works¶
The method is a filter, applied in a fixed order so it never eliminates the wrong thing:
- Gate for feasibility first. Apply hard floors — legal, safety, budget, capacity — and drop ineligible options before dominance is tested, so an efficient-but-unacceptable option can never survive on efficiency alone.
- Fix the comparison direction. For each objective, state which way is better and how ties and near-ties are treated, so "worse on everything" has an unambiguous meaning.
- Compare pairwise. For every pair, ask whether one option is at least as good on all objectives and better on at least one. If yes, the other is dominated.
- Remove and stop. Discard the dominated options; keep the rest as the non-dominated set. The screen makes no ranking among survivors — it hands them on untouched.
Tuning parameters¶
- Dominance strictness — whether only strictly-worse-on-all options are cut, or "near-dominated" ones (worse on all but by trivial margins) are also cut. Looser strictness shrinks the field faster but risks discarding a legitimate option on measurement noise.
- Tie handling — how equal or near-equal scores on an objective count. Treating ties as "at least as good" removes more; treating them as blocking dominance keeps more.
- Objective completeness — how many dimensions enter the comparison. A screen run on too few objectives will confidently eliminate options that were actually strong on an omitted dimension.
- Feasibility threshold placement — how tight the pre-screen hard floors are set, which decides how many options are eligible before dominance even runs.
- Measurement tolerance — the noise band below which a score difference is treated as a tie rather than a real advantage.
When it helps, and when it misleads¶
Its strength is decisiveness without overreach: it can prove that a whole set of familiar compromise options are strictly inferior[1] and never worth another meeting, replacing political defense of a legacy choice with an inarguable comparison. It shines exactly when a decision is cluttered with too many options and someone needs to cut the field before the real deliberation.
Its central failure mode is false dominance from a thin objective set: an option that looks beaten on cost and speed may have been quietly strong on maintainability or resilience that nobody scored, so the screen eliminates it on incomplete evidence. Noisy or inconsistently-scaled measurements produce the same error, cutting an option that a re-measurement would have spared. And the screen is silent on the hardest question — every survivor is efficient, none is thereby acceptable or best. The guarding discipline is to audit the objective set for missing dimensions before screening, treat small score gaps as ties rather than eliminations, and remember that surviving the screen is a floor, not a verdict.
How it implements the components¶
Dominance Screening fills the archetype's eliminate-the-inferior machinery — the filtering half, and only that:
option_set— it operates on the enumerated collection of candidate options, and its output is that set reduced to its non-dominated members.feasibility_constraints— it applies the hard eligibility floors first, ensuring only feasible options are ever compared for dominance.dominance_criterion— its core rule: the explicit definition of when one option is strictly worse than another across all relevant objectives, with tie and tolerance handling specified.
Unlike Multiobjective Optimization Model, which searches a feasible space to build the frontier_map, this screen only prunes options already on the table; and it hands the survivors on without applying a preference_or_priority_rule or writing a tradeoff_rationale — those belong to Weighted Scoring Overlay and Knee Point Analysis.
Related¶
- Instantiates: Pareto Frontier Navigation — it performs the dominance-filtering step that produces the frontier the rest of the navigation chooses on.
- Sibling mechanisms: Efficient Frontier Plot · Tradeoff Curve Visualization · Knee Point Analysis · Multiobjective Optimization Model · Weighted Scoring Overlay · Pareto Frontier Analysis · Scenario Sensitivity Sweep · Stakeholder Frontier Review
Editorial Notes¶
Form Classification¶
Form family: Decision, Gate & Allocation
Rationale: Dominance Screening operates as a case-specific gate, selection, routing, prioritization, or resource disposition because it removes every option that another feasible option matches or beats on all objectives, shrinking a listed field to its non-dominated set before any value choice is made.
Independent corroboration: The frozen evidence defines Dominance Screening as 'Removes every option that another feasible option matches or beats on all objectives, shrinking a listed field to its non-dominated set before any value choice is made', so its operative form is Decision, Gate & Allocation.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Operations Research
Origin pattern: Single lineage
Present-day reach: Multi-domain
Rationale: Multi-criteria decision analysis cohered screening a feasible option set down to its Pareto frontier before any value tradeoff is imposed.
Related originating lineages:
- Economics & Finance — Welfare economics supplied Pareto efficiency and non-dominated alternatives.
Review resolution: Operations research cohered the explicit screening procedure and economics supplied its Pareto criterion; this is a single optimization lineage with genuine economic provenance and multi-domain reach.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
The screen is only as trustworthy as the dimension list it runs on. Its most dangerous error is invisible: an option eliminated because a value it was strong on was never made an objective. When an elimination feels wrong, the honest response is not to override the screen but to ask what objective is missing — and re-run it, rather than smuggling the option back in as a hunch.
References¶
[1] Miettinen, K. Nonlinear Multiobjective Optimization. Springer New York (1998). Formalizes Pareto dominance for identifying alternatives that are no better on every modeled objective and worse on at least one. registry ↩