The baseline ranking is a conditional result

The baseline is a declared reference calculation. It brings together a frozen candidate set, the current evidence interpretations, the stated scales and the authorised weight vector. Its purpose is to make a recommendation inspectable and to provide a consistent starting point for challenges. Calling it the baseline should not imply that it is a neutral view of the world. It contains choices about scope, measurement and preference. Those choices should be visible before the committee interprets the ordering.

A baseline presentation should show the raw performance beside the score. Otherwise a committee sees only the aggregation and cannot tell whether an option leads because it is less costly, more strategically attractive, easier to deliver or less exposed to execution difficulties. Raw values also preserve the distinction between a favourable position within the comparison set and an acceptable absolute performance. The least costly candidate in an unaffordable set is still unaffordable. The most ready candidate in an immature set may still need development.

The illustrative baseline uses fictional proposal descriptions and supplied attribute values. The names organise different intervention stories; they do not represent measured projects or verified engineering alternatives. Cost is labelled in monetary units by the app, while impact, readiness and risk are supplied ratings requiring a meaningful rubric in a real decision. This paper uses the fixture to expose calculation behaviour. It makes no claim that a particular intervention has the cost, impact or implementation characteristics assigned to its illustrative name.

The next reading is contribution. A contribution is the portion of the total preference score associated with an attribute after transformation and weighting. It is not the monetary benefit attributable to that attribute, nor a causal estimate of how the intervention creates value. Its useful role is explanatory: it shows which represented differences and tradeoffs support the ordering. The committee can then challenge the criterion that actually matters to the result instead of debating a total whose internal structure remains hidden.

A low contribution may have different explanations. The candidate may perform poorly within the current range, the criterion may carry little weight or every candidate may perform similarly. The raw values, normalised values and weights distinguish these explanations. A contribution should not be described as evidence that the underlying outcome is unimportant. It expresses how much that criterion adds to the particular weighted comparison. Its meaning remains conditional on the scale and the other alternatives defining the range.

Under these illustrative inputs, Reliability architecture leads at 72.13 preference points, followed by Commercial learning at 62.86. The leader receives 22.92 points from cost, 12.50 from impact, 21.71 from readiness and 15.00 from risk. Its advantage is distributed across represented attributes. This is an explanation of the supplied preference calculation, not a forecast of the intervention's return or a recommendation to transact.

The capacity proposal illustrates a different profile. It carries the strongest represented impact but weak positions on the other criteria under the current comparison. Its lower baseline position therefore does not mean the impact claim was ignored. The claim contributes according to the declared tradeoff, while the other represented concerns reduce the total relative to competing options. If the organisation believes strategic capacity should dominate this decision, it should examine an explicitly different mandate. It should not pretend that the balanced calculation already embodies that preference.

The commercial-learning proposal illustrates the opposite tension. Lower initial cost and greater readiness can support a strong preference score even when its represented impact is less ambitious. The committee should ask whether the readiness rating reflects the same commitment stage as the larger intervention and whether the impact criterion fairly describes the different scope. A pilot and a capacity commitment may belong in different decision envelopes. The arithmetic can compare their supplied ratings, but the charter must explain why that comparison is legitimate.

The energy proposal may remain relevant despite not leading. A ranked list does not establish that only its first entry has value. If the actual decision permits multiple compatible interventions, different timings or separate funding envelopes, the committee needs a portfolio-selection analysis beyond this app. The baseline can help explain individual preferences. It cannot decide whether an attractive lower-ranked option complements the leader, competes for the same delivery resources or provides a hedge against a shared exposure. Those relationships require an explicit extension of the decision architecture.

Pareto membership adds another reading. An option can remain non-dominated because it possesses some represented advantage, even if its weighted score is lower. That is a reason to examine the tradeoff, not a promise that every non-dominated option should be funded. The set may contain too many legitimate compromises for the available budget. Conversely, eliminating a dominated option does not resolve the choice among the remaining alternatives. Dominance narrows a particular kind of redundancy; weights or another authorised choice rule still express preference among tradeoffs.

The gap between displayed scores should be handled carefully. It is meaningful within the supplied numerical transformation, but it does not have an automatic interpretation as a proportionate difference in strategic value. A score cannot be converted into a probability of success or an expected financial return. The ratings themselves may be uncertain or ordinal in a real workflow. The team should use the gap to identify which changes could matter to the preference calculation, then test those changes against credible evidence and authorised judgement.

The baseline should lead to a proposition that can fail. For example, a fictional committee could prefer the reliability intervention for a bounded next-stage investigation if its operating accountability is confirmed and the balanced mandate remains authorised. That statement identifies the action, its basis and a condition capable of changing it. A proclamation that the top-ranked proposal is best gives little guidance when new evidence arrives. A conditional recommendation tells the next analyst what to revisit and tells the sponsor what the present approval actually permits.

Before presenting the proposition, an analyst should perform a result-to-input walkthrough. Select the leader, identify its important contributions, reopen the underlying raw values and check their evidence perimeter. Then compare the nearest relevant challenger and identify the differences supporting the ordering. This walkthrough can reveal an inconsistent rating definition, a scope mismatch or a mistaken direction of preference. It is more informative than merely confirming that the engine reproduced its own totals. Computational reproducibility and interpretation review serve different purposes.

The committee should preserve the baseline even when it rejects the recommendation. It may choose a different mandate, stage an alternative or request a richer analysis. The original result remains a useful record of what the stated model implied before those changes. Keeping it prevents later revisions from rewriting the basis of the earlier discussion. The archive should include the inputs, engine version and decision rationale, with a separate record for the adopted action. A model output and a committee decision are related artefacts, not interchangeable records.