> they should shell out for dedicated labs

Seriously. This goes for all kinds of analytics. We've normalized surveillance, and if you dare suggest opt-in analytics rather than forced or opt-out you get hit with "but then we won't get enough data to improve our product." That's your problem. Do a UX study, interview people, make a focused effort to make good software. That's how we got the best interface design of all time. Now we're drowning in data and our software is fucking garbage that constantly frustrates and confuses its users. Clearly the issue is not a lack of usage data. It may even do you some good to fly blind if this is your alternative.

Really you should do the lab, though. Take this field seriously and cut out the live A/B testing.