Concept generation and selection

Why it matters

The concept chosen early in development fixes most of a product's cost, performance and manufacturability. Teams that generate many concepts systematically, and then select among them with explicit criteria, avoid falling in love with the first idea and reduce expensive redesigns later. Interviewers and GATE both expect you to know the standard methods and to compute a selection matrix correctly.

Key ideas

Concept. An approximate description of the technology, working principle and form of a product — how it will satisfy the customer needs. It is usually a sketch or rough 3D model with a short explanation.

Concept generation — a five-step method.

1. Clarify the problem. Start from customer needs and target specifications. Break a complex problem into simpler sub-problems, typically by functional decomposition: the overall function (e.g. "drive nail") is split into sub-functions (store energy, convert energy to translational motion, apply force to nail) linked by flows of energy, material and signal.
2. Search externally. Lead users, experts, patents, published literature, and benchmarking of competitors' products.
3. Search internally. Individual and group creativity — brainstorming (quantity first, no criticism, build on others' ideas), brainwriting (6-3-5 method), analogies, SCAMPER, sketching.
4. Explore systematically. A morphological chart (concept combination table) lists alternative solutions for each sub-function; one solution is picked from each row to make a complete concept. The number of possible combinations is the product of the number of options in each row, so the chart must be pruned using engineering judgement. A concept classification tree divides the solution space into branches (e.g. electrical / chemical / mechanical energy source) to prune poor branches early.
5. Reflect on the results and the process.

TRIZ (theory of inventive problem solving), developed from the study of a very large number of patents, identifies technical contradictions (improving one parameter worsens another) and suggests inventive principles — segmentation, local quality, nesting, prior action and so on — that have resolved similar contradictions elsewhere.

Concept selection — two stages.

• Concept screening (Pugh method). Choose a reference concept (often an existing product or a strong candidate). Rate every other concept against it on each criterion as better (+), same (0) or worse (−). Net score = number of + minus number of −. Rank, then combine or improve concepts (merge the strengths of two), and eliminate the weakest. Criteria are not weighted at this stage.
• Concept scoring (weighted decision matrix). For the few survivors, assign importance weights to criteria (summing to 100% or 1), rate each concept on a finer scale (typically 1–5 or 1–10) against a reference value, multiply and sum. The highest weighted score wins, but check sensitivity — if two totals are close, small changes in weights may reverse the ranking, so refine or test both.

Other tools used alongside: the analytic hierarchy process (AHP) for deriving weights by pairwise comparison, prototypes and concept testing with customers, and simple cost estimates for each concept.

Good practice. Generate before you evaluate; keep criteria tied to customer needs and target specifications; document reasons so the decision can be revisited.

Formulas

• Number of combinations in a morphological chart: N = n₁ × n₂ × … × n_k
• Pugh net score: Net = (number of +) − (number of −)
• Weighted score of concept j: S_j = Σ wᵢ·rᵢⱼ, with Σ wᵢ = 1
• Normalising raw weights: wᵢ = Wᵢ / Σ W
• Normalised (percentage) score: S_j% = S_j / r_max × 100

Symbols: nᵢ = number of solution options for sub-function i; k = number of sub-functions; wᵢ = weight of criterion i (decimal); Wᵢ = raw importance rating; rᵢⱼ = rating of concept j on criterion i (on a 1–5 or 1–10 scale); r_max = top of the rating scale.

Worked examples

Example 1 (standard) — Pugh screening and morphological count. A manual citrus juicer is broken into four sub-functions with 3, 4, 2 and 3 solution options. Three new concepts are screened against the current model on six criteria (cost, ease of cleaning, juice yield, durability, ease of use, size):

• Concept 1: + + 0 − 0 +
• Concept 2: + 0 + + − +
• Concept 3: − 0 − 0 + −

1. Combinations in the chart: N = 3 × 4 × 2 × 3 = 72 possible concepts (before pruning).
2. Concept 1: 3 plus, 1 minus → net +2.
3. Concept 2: 4 plus, 1 minus → net +3.
4. Concept 3: 1 plus, 3 minus → net −2.
5. Ranking: 2, 1, 3. Concept 3 is dropped (or reworked); concepts 1 and 2 go forward to scoring, and the team checks whether Concept 1's ease-of-cleaning feature can be merged into Concept 2.

Example 2 (GATE level) — weighted scoring. Weights: ease of manufacture 25%, cost 30%, durability 20%, ease of use 15%, portability 10%. Ratings on a 1–5 scale:

• A: 3, 4, 3, 4, 2
• B: 4, 3, 4, 3, 4
• C: 2, 5, 3, 4, 3

1. S_A = 0.25×3 + 0.30×4 + 0.20×3 + 0.15×4 + 0.10×2 = 0.75 + 1.20 + 0.60 + 0.60 + 0.20 = 3.35.
2. S_B = 1.00 + 0.90 + 0.80 + 0.45 + 0.40 = 3.55.
3. S_C = 0.50 + 1.50 + 0.60 + 0.60 + 0.30 = 3.50.
4. B ranks first, but only 0.05 ahead of C. Sensitivity check: if cost's weight rose from 30% to 35% and ease of manufacture fell to 20%, S_B = 3.50 and S_C = 3.65 — C would win. The team should refine both before committing.

Common mistakes

• Evaluating ideas during brainstorming, which kills the number and variety of concepts.
• Adding instead of multiplying option counts in a morphological chart.
• Weights that do not sum to 1 (or 100%) — normalise them first.
• Treating the Pugh net score as a precise measure; it is a screening tool, and the reference concept strongly influences it.
• Accepting a winner by a tiny margin without a sensitivity check.
• Using criteria that are not linked to customer needs or that overlap (double-counting one attribute).

For GATE PI

Expect weighted-score numericals (sometimes with weights to be normalised), Pugh net scores, morphological combination counts, and MCQs that match a method to its purpose (generation vs selection; screening vs scoring). Practise doing the multiplication neatly in a column so an arithmetic slip does not pick the wrong concept.

Quick check

1. Five sub-functions with 2, 3, 2, 4 and 2 options give how many combinations?
2. A concept has 2 pluses, 3 zeros and 1 minus versus the reference. What is its net score?
3. Raw weights are 4, 3, 3. What are the normalised weights?
4. Is the Pugh matrix a generation or a selection tool?

Answers: 1. 96; 2. +1; 3. 0.4, 0.3, 0.3; 4. selection (screening).