Earlier quoted context omitted.
2749÷400000 = 0.0068725 So, 0.7% which seems like a very, very small sample. Did they spend time checking 400k tracks to find these?
If it were a proper random sample then it would be a decent-sized sample that should permit the calculation of a useful confidence interval on a non-negligible proportion. The cool thing about random samples is that the size you need is generally not dependent on the size of the population you're sampling from. If it's not a random sample then I think we probably shouldn't allow them to get away with using the word "…
How could they be held accountable for that?