A Pugh matrix compares concepts with a named baseline, often called the datum. Each judgment means better, the same, or worse against that baseline on a defined criterion. The discussion behind each cell helps a team decide what to develop next.
The example below screens three fictional desk sign holders and proposes a fourth. Its judgments are supplied for the exercise; they are not prototype test results.
Start with a baseline everyone can picture
The team needs a reusable holder for a small desk sign. These are its concepts:
- A — datum: a flat, wide base with a fixed slot for the sign.
- B — folding holder: a hinged base that folds flat and allows angle adjustment.
- C — compact holder: a narrow, heavy base with a removable clip.
Before comparing them, define the criteria. “Convenient” is too broad to support a useful disagreement.
| Criterion | Meaning in this review |
|---|---|
| Packs flat | The holder collapses into a flatter shape for storage. |
| Angle adjustment | The sign angle can be changed without modifying the holder. |
| Desk footprint | The base occupies less desk surface. |
| Setup steps | Fewer separate actions are needed to put the sign in place. |
Holding the intended sign is a mandatory requirement, separate from those preferences. That requirement remains unverified for every concept. A favorable matrix cannot settle it.
Compare each concept with A
Use + for better than A, S for the same, and − for worse. Each mark needs a short reason.
| Criterion | A: datum | B: folding holder | C: compact holder |
|---|---|---|---|
| Packs flat | Datum | +: hinged base folds | S: no folding feature proposed |
| Angle adjustment | Datum | +: hinge permits adjustment | S: no angle adjustment proposed |
| Desk footprint | Datum | S: no smaller footprint proposed | +: narrower base proposed |
| Setup steps | Datum | −: unfold and set the angle | −: attach the separate clip |
B has two pluses, one same, and one minus: a simple net of +1. C has one plus, two sames, and one minus: a net of zero. Those counts summarize the supplied judgments. They do not measure the size of a benefit, and they do not make B a proven winner.
For example, “packs flatter” says which way the comparison goes. It does not say how much storage space will be saved. A five-point score would ask reviewers to estimate something different. Keep the relative judgment and its explanation together so nobody mistakes a symbol for a measurement.
MIT OpenCourseWare's concept-evaluation lecture describes this sequence: clarify criteria, choose a datum, compare concepts, then use their strengths and weaknesses to develop further concepts. It also distinguishes relative comparison from estimating the magnitude of performance. Read the MIT lecture slides.
Use the matrix to propose D
B offers a folding hinge. C offers a removable clip and a narrower base. The team proposes D: a folding holder with a removable clip.
That is a design question to investigate. Do not fill D's column by copying every favorable mark from B and C. The combination could introduce new problems: attaching the clip may add a setup action, and its effect on stability is unknown.
The review record for D should therefore say:
- Proposed features: folding hinge and removable clip.
- Comparison ratings: unknown until the combined concept is reviewed.
- Open questions: how the clip attaches, how many setup actions are required, and whether the holder supports the intended sign.
- Next work: clarify the combined design and gather evidence for those questions.
Keep the original B and C judgments visible. Otherwise, a speculative improvement can quietly replace the evidence that led to it.
Try the working example
Select a cell to see its comparison and rationale. Check the plus, same, and minus counts, then open the hybrid proposal to inspect its unknowns. Compare the relative marks with the explanation of a numerical rating before choosing which approach fits your review.
For a team session, start a blank matrix, name the datum, and write the criterion definitions before entering marks. Add temporary discussion notes where reviewers disagree. The reset control asks for confirmation before clearing those notes.
The example records a concept discussion. It does not generate a CAD model or establish that a design will work.
Copy this prompt to make your own
Replace the bracketed inputs for your own review, or retain the worked case to rehearse the method.
Create a complete self-contained HTML Pugh concept-review matrix.
My project: [design problem]. Datum: [baseline concept and features].
Alternatives: [concept descriptions]. Criteria: [definitions and what
better means]. Mandatory requirements and current evidence: [details].
Use only supplied judgments; leave missing comparisons unknown.
Include this separate fictional desk-sign-holder worked case:
A datum = flat wide base with fixed slot; B = fold-flat hinged base
with adjustable angle; C = narrow heavy base with removable clip.
Criteria in order: packs flat, angle adjustment, smaller desk footprint,
fewer setup steps. B comparisons are +, +, S, −; C are S, S, +, −.
Each cell must open its reason: B folds, adjusts, has no proposed
smaller footprint, and adds unfolding/angle-setting; C does not fold,
has no proposed angle adjustment, has a narrower base, and adds clip
attachment. These are illustrative judgments, not prototype tests.
Show B counts +2/S1/−1/net+1 and C +1/S2/−1/net0. No winner.
Explain that relative marks do not measure benefit magnitude.
Propose D combining B's folding hinge with C's removable clip, but
leave all D ratings unknown. Ask about extra setup work, stability,
and whether it holds the intended sign. The mandatory hold-sign
requirement remains unverified for all concepts. Preserve B/C marks.
Allow cell inspection, count checking, hybrid inspection, a blank
working matrix, and temporary notes that survive switching views.
Confirm before resetting notes. Show missing datum/criterion feedback
in the blank version without replacing the user's entries.
Use labeled keyboard-accessible controls with visible focus, text at
least 16px, and readable 360px layout without horizontal scrolling.
Put CSS/JavaScript inline, use system fonts, and load no dependencies.
Return complete HTML, not CAD, a prototype test, or certification.
Make the disagreement precise
If reviewers keep arguing over “desk footprint,” settle what the criterion means before another comparison. In NOS, clarify the instruction above the blank working row so it names the datum and the direction that counts as better. Keep the comparison reason beside the mark; that gives the next design discussion a precise starting point.