Compensation can hide unacceptable performance

Some conditions define whether an alternative may be considered at all. They do not describe a benefit that a sufficiently strong performance elsewhere can offset. A committee may require a lawful operating arrangement, an acceptable environmental and social assessment, an available delivery owner or a minimum service capability. An attractive strategic story does not necessarily compensate for failure on such a condition. The decision process should identify these requirements before aggregation, with authority and evidence appropriate to the condition being imposed.

Eligibility and preference answer different questions. Eligibility asks whether the option belongs in the admissible choice set at the specified commitment stage. Preference asks which admissible option better serves the stated objectives. A proposal can be eligible and undesirable, or potentially desirable and currently ineligible. Confusing these states makes remediation difficult. The team may need to redesign an attractive but inadmissible proposal, while an eligible weak proposal may simply deserve no further attention. The recommendation should state which problem exists rather than treating every low score as the same diagnosis.

A gate needs an accountable owner. The analyst should identify who may define the condition, who can verify its satisfaction and who can authorise an exception if exceptions are permitted. An invented numerical threshold is not a substitute for that governance. A risk rating in the app is an illustrative preference attribute; it does not establish compliance with an applicable standard or permit. A screening team must use the relevant assessment process and retain the evidence supporting its conclusion. The weighted model can then compare options that have passed the required checks.

The condition also needs a timing statement. A feasibility-study commitment may reasonably proceed while some operating requirements remain unresolved, provided the study is designed to resolve them and does not itself require those conditions to be satisfied. A construction commitment may require a different evidence standard. Calling a proposal ineligible without naming the decision stage can prematurely exclude a useful development option. Calling it eligible for a study does not authorise implementation. The register should distinguish the present commitment from the future project state.

The committee should avoid hiding a gate inside a very large weight. A high risk weight still participates in a compensatory sum. Enough benefit on other represented attributes can offset poor risk performance, depending on the scales and the candidate set. If a condition is truly non-compensable, represent it outside that sum. This preserves the meaning of both mechanisms. The gate records the required condition; the weights record tradeoffs among admissible performances. A large weight can express strong preference, but it should not masquerade as a prohibition.

Dominance is a separate useful challenge. In the canonical app, a candidate is Pareto-dominated when another candidate is no worse on every represented raw criterion and strictly better on at least one, taking the declared directions into account. Lower cost and risk are favourable; higher impact and readiness are favourable. This test does not use the weights. It identifies an option that lacks a represented performance advantage over a comparator. Before retaining that option, the committee should ask whether an omitted attribute, different scope or data error explains the apparent disadvantage.

A dominance result depends on the represented attributes and their evidence. It cannot prove that an option is inferior on every consideration that matters to the organisation. A proposal may possess an unrepresented contractual right, geographic fit or strategic option. The proper response is to examine whether that feature belongs in the decision architecture, rather than dismissing the test or inventing an unexplained override. If the feature matters, document it explicitly. If it does not, the dominated option usually needs a reason to consume further analytical attention.

Dominance and an additive ranking can disagree in an instructive way when a criterion receives no weight. The weighted preference then deliberately ignores differences on that criterion. A candidate may tie on every positively weighted attribute while being worse on ignored attributes. The app reports the raw dominance result alongside the weighted tie. It does not silently modify the declared weights. This makes the tension visible and lets the committee decide whether its stated indifference is intentional, whether the ignored attributes belong in gates or whether the vector needs reconsideration.

In this illustrative calculation the weighted scores are 50.00 and 50.00. The raw Pareto set contains B cheaper proven system. The result does not license the literal text display leader to receive capital. It asks the committee whether indifference to cost, readiness and risk is truly authorised, or whether those concerns are meant to be handled elsewhere. Equal weighted scores and unequal raw performance can both be correct statements under the declared model.

A useful review sequence therefore begins with eligibility, proceeds to a dominance challenge and then examines preference. The sequence does not require excluding every dominated proposal automatically. It requires an explicit explanation for keeping one. A development proposal may be retained because an uncertain attribute could change after targeted investigation, because its scope is being redesigned or because the represented comparison omits a material feature. Each reason should identify the work needed to resolve it. Retention without a reason weakens the discipline of the option set.

Gates themselves can be gamed. A sponsor might define a technical requirement so narrowly that only its preferred design satisfies it, or declare a delivery feature indispensable without showing how it serves the objective. The gate review should test necessity, proportionality and decision authority. Where different technologies can achieve the same required outcome, an outcome-based condition may preserve more legitimate alternatives than a design-specific condition. This is a governance judgement informed by technical evidence, not a task the scoring app can settle automatically.

Exceptions deserve visible treatment. If the organisation has an authorised exception process, the committee should record the condition, rationale, compensating measures and person responsible for the residual decision. An exception should not appear as a favourable score adjustment. That would conceal both the departure and the authority that accepted it. Where an exception is unavailable, the analyst should not create one by allowing an otherwise attractive total to override the failed condition. The method must follow the applicable decision rules.

The final recommendation can be conditional without being evasive. It can identify the preferred admissible option while stating that commitment depends on a particular unresolved gate, or recommend a study because the eligibility evidence is incomplete. The condition should be specific enough to change the action when it fails. A vague promise to manage risks later provides little decision control. A named approval, accountable assessment or demonstrable capability makes the next step reviewable and preserves the boundary between preference and permission.