bri_stats said:One practical note: write down what you did and when, before you need it.
This is where I part company with the consensus forming above. The objection nobody has made: this only works if the measurement is doing what we assume it is doing, and that assumption has not been tested here.