Another story: https://en.wikipedia.org/wiki/Annie_Dookhan https://en.wikipedia.org/wiki/Annie_Dookhan I don't get why they don't continuously send 10-20% of "

Another story: https://en.wikipedia.org/wiki/Annie_Dookhan https://en.wikipedia.org/wiki/Annie_Dookhan I don't get why they don't continuously send 10-20% of "control samples" (either with known result, or randomly sampled from e.g. the last 3 months → to check that the results are still the same) through the labs, thus having a continuous estimate of data quality - and they can also score "accuracy vs. technician" etc., to spot these problems. (of course this won't work when the entire system is corrupt -- but often it's just a malice/incompetence of a few persons in the lab)