How can I keep one summary score from hiding an important trade-off?

Treat the total as an index, not a verdict

A summary score answers a narrow question: given these criteria, weights, ratings, and arithmetic, which option has the larger total? It does not show whether every criterion belongs in the comparison, whether a serious downside has been averaged away, whether the ratings are well supported, or whether an important effect was left outside the grid.

The University of Minnesota Extension criteria grid shows how a total can make several options easier to compare and how weighting can clarify close choices. The 2026 Green Book draws the necessary boundary: an initial ranking by summary metrics is not sufficient for a decision. Its balanced judgment also considers unmonetisable effects, risks, uncertainty, financial impact, distributional effects, and whether an option meets the objective. These are formal group and public-policy sources; the low-stakes transfer is to keep the arithmetic visible without letting it impersonate the whole judgment.

Put the MARGIN card beside the score

MARGIN is a Forever Free Compass framework, not an external research finding. Complete it after calculating a comparison score and before naming a preferred option.

1. Must-pass boundaries: list any requirement that cannot be traded away for gains elsewhere; an option that fails one should not be rescued by its total.
2. Aggregate total: record the score, the scale, and the weights used without adding precision the evidence cannot support.
3. Row-level reasons: identify the one or two criteria that create most of the lead and state what each rating is based on.
4. Gaps outside the score: name one important effect, person, place, time horizon, or uncertainty that the total does not represent well.
5. Instability test: change the weakest important rating or weight across a plausible range and note whether the ranking reverses.
6. Narrative decision: write one sentence explaining the preferred option, the accepted downside, the remaining uncertainty, and the next review point.

Find the trade-off that creates the lead

Compare the options criterion by criterion, not just total by total. Mark the largest meaningful difference and ask whether it reflects evidence, a forecast, or a preference. Then inspect what the leading option gives up. A two-point advantage spread across convenience and familiarity may not offset a failure on a boundary; several small gains may also conceal one large downside that deserves a separate decision.

The Government Analysis Functional Standard says decisions should assess options against defined criteria while taking external context and uncertainty into account. It also calls for assumptions, limitations, applicability, caveats, and transparent reasons to accompany analytical results. For a personal comparison, that supports keeping the basis of the ratings and the reason for the final judgment beside the total.

Show what cannot be safely collapsed

Keep a separate note for hard boundaries, important unscored effects, who bears a downside, and evidence that is too uncertain to express as one confident number. Do not quietly turn unlike considerations into interchangeable points. If a trade-off would be unacceptable no matter how attractive the other criteria look, treat it as a screen rather than a weight.

The 2026 Green Book recommends an appraisal summary table precisely to avoid reducing a comparison to one metric: it presents summary measures alongside unmonetisable effects, financial impact, distribution, risks, and uncertainty. The 2026 MHCLG Appraisal Guide likewise says appraisal is not a decision algorithm and places non-monetised impacts, assumptions, uncertainty, switching values, and residual risk beside its summary measures. MARGIN is only a plain-language adaptation of that presentation discipline.

Make the final choice in words

Finish with a sentence such as: ‘Option A leads because it better serves the stated objective, even after the rating for time is reduced; I am accepting the higher effort, the effect on another commitment remains outside the score, and I will review after the first bounded step.’ If you cannot explain the choice without quoting the total, the comparison is not ready to carry the decision.

Use MARGIN only for ordinary, low-stakes reflection. It is not a validated scoring, multi-criteria decision analysis, statistical, safety, legal, medical, financial, employment, housing, or safeguarding method. Do not use a personal grid to trade away consent, rights, safety, fixed obligations, or professional requirements; seek qualified guidance for consequential choices.