A couple interesting tidbits in this talk.  First, around 26:00, Hal Varian talks about how much Google runs experiments, saying, "Any time you access Google, you probably are in a dozen or more experiments."  Second, around 34:00, Hal Varian starts a discussion of data silos by pointing out that Google avoids them like the plague, even converting data from startups quickly into Google's infrastructure, so that people can analyze each other's data easily and move teams without any transition costs (same tools they're used to work in the new team as the old).

Expanding on the first point a bit, Google and Amazon have been doing experiments for a long time (and Amazon even longer than Google), but other companies might have something to learn from this.  Once you embrace ubiquitous A/B testing, it changes how you do everything.  Deployment of new software, for example, is no longer testing for weeks, then throwing the software over the wall and praying nothing breaks.  New software development becomes testing, pushing out to a few customers, rolling back, changing, pushing out to more customers, pushing out to even more customers, and entirely out the door.  The entire way you develop software, the entire way you develop product, changes when you are A/B testing everything constantly and continuously.
Shared publiclyView activity