Pugh matrix seeding

Score options against a baseline as better, same or worse, then build new options by combining the parts that scored better.

Use when

You have several options, all imperfect, and a clear incumbent to compare against.
The options have separable parts that could be recombined.
You want new options rather than a winner from the existing set.

Do not use when

The options are indivisible — you cannot take half of a job offer.
There is no sensible baseline, which makes the plus-and-minus scoring arbitrary.

Inputs required

Procedure

01
Pick the datum
Choose one option as the baseline. The incumbent or the default is usually right, because the question is what to do instead of it.
02
Score relatively
For each other option and each criterion, mark better, same or worse than the datum. Do not use numbers. The coarseness is deliberate and it keeps the scoring honest.
03
Count, but do not trust the count
Sum the pluses and minuses per option. Treat the total as a prompt, not a result — the criteria are unweighted at this stage.
04
Seed new options
This is the point of the method. Take the pluses from different columns and ask whether a single option could hold them all. That hybrid is usually not on the original list.
05
Re-run with the hybrids
Add the new options and score again against the same datum. Two rounds is normally enough.

Characteristic failure mode

Stopping at the count. Used as a scoring method it is a weak weighted matrix; used as a generator it produces options nothing else would have found.

Worked example

Choosing among four flats, with the current one as the datum.

01Flat B: better on space, worse on commute, same on cost.
02Flat C: better on commute, worse on space.
03Neither dominates. The pluses are on different criteria.
04Seed: is there a flat with B“s space at C”s commute? That search had not been run, because the original four came from one website.

Result

A fifth option found by combining two columns, better than any of the original four.

Where to go next