Ask a management team for its decision criteria after it has seen the options, and you get a list that fits the favorite. Criteria written after the options are a justification. Criteria written before are a decision. The plant with the laser blank question therefore closed the cost estimates in a folder and wrote its criteria first.
A criterion must be able to separate two options
Many criteria on a first list cannot do that.
- “Future proof.” Every option scores well and no evidence settles it. Better: capacity covers the volume plus 15 percent at two shifts.
- “Low risk.” Mixes volume risk, technical risk and supplier risk. Better: cost per blank at 80 percent of volume.
The test is simple: can two people with the same facts arrive at the same score? If not, the criterion measures a preference.
Must and want
A must excludes an option completely. It has no weight and no score, only a yes or a no that evidence can answer. A want separates the options that remain. It is weighted and scored from 0 to 10 against a written definition.
The plant wrote four musts:
- M1: capacity for 420,000 blanks plus 15 percent, secured in writing.
- M2: fits into the 1,150 square meters of free hall space, with a clear height of 9.5 meters.
- M3: thick plate up to 20 millimeters available, cut in house or purchased.
- M4: runs on the existing two-shift model, without a new shift agreement.
Three to five musts are a good number. Too many leave no options at all. The test for each: would you reject the otherwise best option because of this one sentence? If the honest answer is no, it is a want with a high weight.
Weights from a paired comparison
Asking a group for percentages gives numbers that add up to 140. Asking for one comparison at a time works better. Every criterion is compared with every other: 2 points if the first matters more, 1 point each if they are equal, 0 if the second matters more. Six criteria give 15 pairs and 30 points. The weight of a criterion is its points divided by 30, rounded to whole percent.
| Want criterion | Points | Weight |
|---|---|---|
| Total cost over five years | 9 | 30 % |
| Delivery reliability and lead time | 7 | 23 % |
| Flexibility for new parts and small batches | 5 | 17 % |
| Capital tied up and risk if volume falls | 4 | 13 % |
| Quality and process control | 3 | 10 % |
| Implementation effort and people | 2 | 7 % |
Each participant fills the matrix alone. Then the group reconciles only the pairs where the answers differ. In the plant, that was mainly one pair: cost against delivery reliability. After 20 minutes the group settled on equal, which is why cost has 9 points and not 10. A seventh criterion on the first list found no defender in any comparison and was dropped.
Anchors for every want
A weight is useless if a 7 means something different to each person. So every want gets a measurable definition. For delivery reliability: working days from call-off to blanks at the welding cell, and the share of late deliveries. For capital and volume risk: investment plus stock, and the cost per blank at 80 percent of volume.
The same idea of weighting appears in value analysis, where it ranks the functions of a product. Here it ranks what the organization wants from a decision.
Three rules for the session
- Weights are agreed before any option is scored. Afterward they are a way to make the favorite win.
- No prices in the room until the weights are fixed. The first number spoken would set the frame.
With four musts and six weighted wants, the plant knew what a good answer would look like. It did not yet have enough answers to choose from. The next part is about finding them.