What Barrier Level Your Data Will Defend

How to read this sheet

Read the impact bound against the impact mean before anything else. Where the mean is inside the limit and the bound is outside it, the fabric is not the problem and reformulating it would be money set on fire: the problem is that four specimens cannot prove anything about a population, and the fix is more specimens. That is the column marked the spread and not the mean, and on the seeded catalogue it is three fabrics out of six. Why the bound is so far from the mean. On four specimens, at ninety-five per cent confidence that ninety-five per cent of gowns conform, the tolerance factor is five point one four. On eight it is three point one nine, on twenty it is two point four, and it never falls below one point six four however many you run. That last number is the whole reason the sheet can tell you when no sample size will do: if the limit is closer to the mean than one point six four standard deviations, the population itself is out of specification and no amount of testing will rescue it. What is being assumed. Every specimen count the sheet proposes assumes the further specimens come out like the ones already run - same mean, same spread. They will not exactly, and a fabric close to the line should be retested before the claim is moved rather than after. The viral counts assume every further challenge passes, which is a stronger assumption still and is flagged separately for that reason. What is not modelled. Nothing here values a claim you cannot make: a tender that requires level three and a fabric that defends level two is a lost tender, not a price step, and the sheet prices only the step. The prices are one level in each direction because that is the only step a converter can usually put a number on, and where a fabric is further than one level from its claim the sheet says so rather than extrapolating. Batch-to-batch drift is not modelled at all - the bound is drawn around one sample, and a fabric that has been reformulated, resourced or moved between lines since that sample was taken is being judged on evidence that no longer describes it. And the level scheme itself is entered as four numbers rather than assumed, so that a revision of the standard, a different jurisdiction or an entirely different barrier scheme is a change to the fields and not a change to the sheet.