01 · Definition and the baseline
Definition and the baseline
A Pugh matrix compares candidate concepts with a reference concept against agreed criteria. A simple version records better (+), similar (0), or worse (−) for each comparison. The symbols are ordinal judgments: they express direction relative to the baseline, not an exact amount of value. The purpose is to expose trade-offs and improve concepts, rather than manufacture a mathematically precise winner from incomplete information.
Use it when several feasible designs address the same problem and the team needs to understand their relative strengths. Prerequisites include a clearly described baseline, comparable concept definitions, and criteria that matter to the decision. Candidates should first satisfy must criteria. A baseline may be the current process or a well-understood concept. Choosing an unstable or poorly defined baseline makes every judgment ambiguous and can distort the comparison.
02 · Define criteria and comparison rules
Define criteria and comparison rules
This worked matrix is an earlier concept-development snapshot before final must-criteria screening. Unresolved recovery concerns are deliberately visible here; they are resolved and screened in the later eligibility decision. The shared screen is an information component, not a complete identity solution eligible for selection on its own.
Fictional Northbank Dispatch uses its current staging process as the baseline. The team compares fixed carrier lanes, early assignment with scan confirmation, and a shared information screen. Criteria are search delay, wrong-lane prevention, response to reassignment, operator burden, and ease of recovery during an outage. The team avoids using both “speed” and “search time” as separate criteria unless they capture distinct consequences; otherwise a preferred feature receives repeated credit.
For each row, members write what “better” means. Lower search delay is desirable. Stronger prevention of wrong-lane loading is desirable. Lower operator burden is desirable. A concept receives 0 when the available evidence does not establish a meaningful direction, with a note identifying uncertainty. This convention is local and must be distinguished from a confident judgment of equivalence. Separate notes can mark “similar” and “unknown” when that distinction is important.
03 · Populate with reasons, not just symbols
Populate with reasons, not just symbols
For fixed lanes, Northbank assigns + on search delay, 0 on wrong-lane prevention, − on reassignment response, + on operator burden, and 0 on outage recovery. The reassignment minus reflects the risk that a carton remains in a lane chosen from outdated information. For early assignment with scan confirmation, it assigns +, +, +, −, and −. The burden minus reflects added scanning; the recovery minus reflects an unresolved offline procedure. For the screen it assigns +, 0, +, 0, and 0.
Every cell has a short reason and evidence status. Observed walking patterns support the fixed-lane search judgment; a proposed software feature supports the scan concept’s reassignment judgment only as a design hypothesis. A minus is not automatically fatal unless it corresponds to a failed must criterion. Conversely, several pluses cannot compensate for a traceability failure. The matrix makes the concern visible so the design can be revised.
04 · Worked concept improvement
Worked concept improvement
A superficial count gives fixed lanes two pluses and one minus, scan confirmation three pluses and two minuses, and the screen two pluses with no minuses. Northbank does not interpret these counts as exact net utility. One minus on outage recovery could matter far more than one plus on routine search. The team reviews the meaning of each trade-off and discovers that the scan concept’s two disadvantages are designable rather than unavoidable.
It develops a single confirmation scan at loading instead of repeated scans during every movement, reducing operator burden. It also creates a controlled offline manifest with reconciliation before normal service resumes. The revised concept is reassessed against the baseline and must criteria before selection. The screen remains an information component and cannot be selected as a complete stand-alone dispatch identity solution. The matrix has helped generate a stronger design. It has not justified selecting the screen merely because its unweighted symbol count appeared tidy.
05 · Interpretation and sensitivity
Interpretation and sensitivity
Judgments should be revisited when new evidence changes a mechanism. If a pilot shows scan confirmation causes long queues, the burden comparison may worsen despite faster carrier search. If a carrier frequently changes after staging, fixed lanes may need a visible change signal. The baseline and concept versions remain recorded so the team does not compare a revised candidate with an old, implicit reference.
Some teams use weighted variants. If Northbank does so, it declares the weights, symbol-to-score convention, and sensitivity checks as a local decision aid. Converting +, 0, and − to 1, 0, and −1 does not make the intervals equal or the result objectively measured. A weighted total can organize discussion, but the decision must still address critical requirements, uncertainty, and whether the recommendation changes under plausible alternative weights.
06 · Weak versus strong use and advanced adaptation
Weak versus strong use and advanced adaptation
Weak practice fills cells from preference, omits reasons, counts symbols as though every criterion has the same consequence, or changes the baseline to make a favorite look better. Strong practice defines the reference, states the meaning of every comparison, preserves evidence gaps, and uses disadvantages to improve designs. It distinguishes an unknown comparison from demonstrated similarity and keeps must failures outside compensating scores.
Advanced use compares concepts against more than one sensible baseline. A design that looks strong only against the current troubled process may be less compelling against a simple fixed-lane alternative. Northbank can also perform a separate comparison for routine operation and outage recovery, because averaging those contexts hides a serious weakness. The best outcome may be a hybrid concept whose elements come from several candidates. That hybrid requires a new review and pilot; it does not inherit proof from the matrices that inspired it.
07 · Operating conditions and evidence
Operating conditions and evidence
The matrix should describe concept maturity. A fixed-lane design with a completed layout review cannot be compared fairly with an undefined future software package simply by assigning confident symbols to both. Northbank records which features are demonstrated, quoted, proposed, or unresolved. Where the direction depends on an unbuilt feature, the cell remains a hypothesis with a validation action.
The team also checks whether the baseline is being idealized or unfairly weakened. Current recovery may be slow but well understood, while the new design may have faster routine operation and uncertain outage behavior. Both deserve explicit treatment. If baseline performance changes during the project, comparisons are refreshed rather than retaining a convenient old reference. A good Pugh review is repeatable in its logic even though reasonable experts can differ in judgment. The recorded reasons make those differences discussable and allow pilot evidence to resolve them.
08 · Decision record and next step
Decision record and next step
A practical Pugh review ends with design actions rather than a celebratory total. Northbank assigns the outage recovery concern to the systems owner and the added scanning burden to the dispatch owner. Each action states what evidence would change the comparison. A tested fallback can improve the recovery judgment; a measured queue can establish whether burden is acceptable. This makes the matrix an agenda for engineering and operational learning.
The team does not compare benefits and disadvantages as if all symbols represented equal quantities. Wrong-carrier prevention and operator convenience differ in consequence and may have different acceptance conditions. Even in a locally weighted version, a severe requirement remains a gate. Northbank records the final decision in plain language: the revised scan concept is selected for testing because it addresses assignment and identity mechanisms, while its burden and recovery risks are specifically controlled. The explanation is stronger than a claim that one symbol total won.
| Comparison | Interpretation | Decision use |
|---|---|---|
| + | Better than the stated baseline on this criterion | Retain the reason and evidence boundary |
| 0 | No demonstrated directional difference under the declared convention | Separate unknown from confirmed similarity in notes |
| − | Worse than the baseline on this criterion | Redesign, investigate, or reject if it violates a must |
Connections
Related methods and phase use
Placed under Improve because of the outcome it supports. Use the method in other phases when its purpose and assumptions fit the work; the phase list is a guide to selection, not a compulsory sequence.
Sources
Professional foundations and further reading
The explanation and fictional teaching examples are written for LeanTrainingLab. Linked primary sources provide professional foundations; example values and local conventions are stated in the guide. The accepted toolset draws on Six Sigma+Lean Toolset, second edition (Lunau, editor; Springer, 2013). No book text or examples are reproduced here.