Why no number is given
A bowtie has no scale table, no score formula and no result band, and numeric effectiveness thresholds are absent from the source we measured. The phrase "at least three barriers" is repeated often but it is not a measured threshold — and Optifora does not write it as though it were.
Tests instead of a number
- Every barrier must pass the validity test.
- Barriers must be independent — one failure must not take two down.
- Where possible they should be of different types; the same type shares the same weakness.
- Having at least one barrier that does not depend on a person is preferable.
The trap in counting
| Appears as | Really is |
|---|---|
| Five barriers, all fed from one panel | One barrier |
| Four barriers, three depending on one person's attention | Two barriers |
| Two barriers, one passive hardware and one automatic interlock | Two barriers |
If a numeric answer is required
"How many layers are enough" does have a numeric answer in layer of protection analysis: the required risk reduction factor is the initiating event frequency divided by the tolerable frequency, and the number of layers follows from it. That arithmetic does not exist in a bowtie.
Few but real
Two real barriers beat six boxes that fail the test. The job of a bowtie is not to make people count boxes but to show which box actually stops something.