Concept Generation and Selection
Diverge Widely, Then Decide Honestly
Concept generation is about producing many genuinely different approaches before judging any of them. Useful techniques: brainstorming with deferred criticism, morphological analysis (list functions, list means for each, combine), functional decomposition, benchmarking, and analogical transfer from other industries.
Selection is where exams concentrate. Know these tools:
- Screening matrix (Pugh) - compare concepts against a datum concept using +/−/same, to eliminate weak ones cheaply.
- Weighted decision matrix - score each concept against criteria, multiply by criterion weights, sum. The winner is a decision aid, not an oracle.
- Pairwise comparison - derive weights when criteria are numerous.
Traps
- Weight sensitivity: if swapping two criteria weights flips the winner, the result is not robust - say so rather than declaring a winner.
- Scale abuse: scores of 1–5 are ordinal, not cardinal. Summing them is a convenience, not a proof.
- Anchoring on the datum: in a Pugh matrix, a concept that beats the datum on every criterion is suspicious - check whether criteria are complete.
- Premature convergence: selecting before enough concepts exist. Also, a concept that scores poorly may be fixable; elimination should consider potential, not just current score.
Rule of Thumb
Weight the criteria before you see the scores. If weights are set after scoring, the matrix will mysteriously favor whatever the team already liked.
Worked example: Two concepts, criteria weight A = 0.6, B = 0.4. Concept 1 scores 8 and 4 (sum 0.6×8 + 0.4×4 = 6.4). Concept 2 scores 6 and 9 (0.6×6 + 0.4×9 = 7.2). Concept 2 wins. Note that Concept 1 wins on the higher-weighted criterion - the result hinges on how much you trust that weight, which is exactly the sensitivity check a good answer mentions.