[drwh0]

but anyone in useability will tell you that you ruin the experiment if the user knows that someone/software is looking "over their shoulder". any software installed or testing environment introduced will produce different results in users. this is why larger sites settle with coarser-grained analytics derived from tracking cookies and js beacons (which can measure abandonement). indeed cookies+beacons are a superior technique because you can randomize or exclude parts of your test body at will, or even other metrics (geolocation etc) to add/remove people to tests at will. indeed, everyone reading this has unwittingly been in a test for a major website at some point without knowing it (which means your data is actually valid!)