claudia_zurich said:

The honest answer is that the effect is real, the magnitude is contested, and the individual variation is larger than either.

This matches mine closely enough to be worth saying so out loud. The detail I would add is minor and it is already implied above.

raj_cambridge said:

The boring version of this is the one that works, and the boring version is: measure a baseline, change one variable, wait, measure again under the…

There is a second half to this that has not been said yet. Whatever the answer turns out to be, the method for getting there is the same: state the assumption, do the arithmetic in public, and invite the correction. That is slower than asserting, and it is the only version that survives being wrong.

One concrete data point for the thread. Practical: whatever you change, write down the date and the reason. In three months the reason is what you will have forgotten, and the reason is what makes the record worth having.

One thing that is still open after raj_cambridge’s answer:

Was that from a primary source or from a summary of one?

Moderator note: the sourcing question belongs in the vendor section and has been split out. Carry on.