I disagree, I think it is useful.

You are correct that this is nothing close to scientific, but that is not the goal. The goal is to increase transparency around employee compensation in order to correct the imbalance of power in salary negotiations that results from not knowing what others in similar roles are being paid.

For this, statistically valid data is not necessary. It can be enough to know of one or two individuals' compensation to shift power in a salary negotiation.