Evaluate ideas after exploration
Judge proposals against the brief before ranking them. A constraint violation, unsupported factual premise, or infeasible dependency is not offset by a high novelty score. Keep uncertainty separate from low quality; an untested assumption is not a measured failure or success.
For each serious candidate, inspect the causal mechanism, specific beneficiary, feasible first version, adoption route, closest substitute, main risk, and cheapest disconfirming test. Use a simple qualitative comparison unless the task has an explicit decision model. Do not invent precise weights or empirical scores.
Differentiation
Describe what changes relative to the closest alternative and why it could matter. A renamed dashboard is not a new mechanism. Conversely, an established mechanism may be useful in a different context without being globally novel. A novelty claim must state its search/evidence scope. Explain the difference in natural prose; there is no required contrast sentence or slogan.
Portfolio
When the request calls for exploration, compare mechanism, audience, adoption, risk, and timescale. Do not select three near-duplicates because they scored well, or include a weak long shot merely to fill a bold-bet slot. A requested single winner should be supported by the decisive reason and uncertainty, without an unrequested catalogue of alternatives.
Tests and decisions
Match the experiment to the uncertain claim. User interviews can uncover language and unmet needs; they do not establish willingness to pay. A prototype can test interaction or feasibility; it does not prove acquisition. Identify what a test could distinguish, the least costly credible observation, and the action that would follow a positive, negative, or ambiguous result.
For research, include the proposed mechanism, alternative explanations, measurement, controls, and failure signal. For product/service work, name the first user, first context, first version, and route to adoption. For names/campaigns, evaluate the agreed audience and brief without turning the task into product redesign.
Keep experiments ethical and authorised. Do not launch pages, spend advertising budget, contact prospects, charge cards, or represent an unbuilt capability as available merely because an idea card contains a test plan.