High, medium and low cannot be summed, compared or defended in a budget meeting. A rating that decides where money goes needs two numbers: a probability per year and an impact in EUR and in days. Both are estimates, and both can be built from facts.
Probability
Five classes, agreed once, in writing, before the first risk is rated. Class 1 is below 2 percent per year, class 2 is 2 to 10 percent, class 3 above 10 to 33 percent, class 4 above 33 to 100 percent, and class 5 is anything that happens several times a year.
The useful translation is frequency. Class 3 means once in three to ten years. In the example plant, a site for conveyor components, the machining center MC-200 had failed for more than three days twice in five years, which is class 4 at 40 percent.
Impact
Impact classes run from below EUR 25,000 to above EUR 1,500,000, with a matching band of days: below one day, one to three, three to ten, ten to thirty, above thirty.
The number is built, not guessed. One lost production day on the housing line is 400 housings at a contribution margin of EUR 85, so EUR 34,000. Eighteen days are EUR 612,000, the spindle repair adds EUR 8,000, and nothing is lost in penalties because the finished goods stock covers the first three weeks. Impact EUR 620,000, class 4. Write that calculation next to the number. A number nobody can check will be argued away in the first budget meeting.
Score and expected value
The score is the probability class times the impact class, from 1 to 25, and it routes the risk: 1 to 4 accept, 5 to 9 monitor, 10 to 14 treat with an owner and a date, 15 to 25 decide at plant management within four weeks. The thresholds are agreed before the rating and are not moved to make a result look better.
The expected value is the probability times the impact in EUR. It is the only field that can be summed, so it gives the order of work.
| Risk | P per year | Impact EUR | Score | Expected value EUR |
|---|---|---|---|---|
| MC-200 out for more than three days | 40 % | 620,000 | 16 | 248,000 |
| ERP and MES gone for twelve days | 12 % | 980,000 | 12 | 117,600 |
| Single source for the drive motor stops | 8 % | 1,450,000 | 8 | 116,000 |
| Energy cost above the planned figure | 35 % | 320,000 | 12 | 112,000 |
| Field action on a safety relevant weld | 6 % | 1,250,000 | 8 | 75,000 |
| Fire in the paint and drying line | 1.5 % | 4,200,000 | 5 | 63,000 |
All 24 risks of the example plant carry an expected loss of EUR 1,000,000 per year, which is 1.3 percent of revenue. The matrix shows one risk in the red field, three in orange, fourteen in yellow and six in green.
The trap
Ranked by expected value, the fire comes last in that table, behind an energy bill and a staffing peak. Ranked by what it would do, it is the only event in the register that could end the plant: ninety days without the paint line costs more than the site earns in a year.
So the example plant keeps a second list. Every risk in impact class 5 goes to plant management by its impact alone, whatever its expected value, and the question there is not what it costs on average but whether the plant could carry it once.