> The fact is, if it hasn’t been observed by then it’s probably not happening at all. That's a big assumption. The metrics we're shown might not have enough r

> The fact is, if it hasn’t been observed by then it’s probably not happening at all. That's a big assumption. The metrics we're shown might not have enough resolution to show the effects of 10-12 days of scandal. The fallout may be a much more gradual process than you assume. For instance: take this scenario: someone decides to #deletefacebook, even though they use it daily as a communications tool. What do you think they'll do? You seem to assume they're ragequit immediately, but I think it's far more likely that they'll dial back their usage, explore replacements for certain features, exchange alternate contact info with FB-only friends, etc. Only much later will they actually delete or show a significant usage decrease. That scenario isn't hypothetical, it's exactly what I'm doing. I've made the decision to disengage from Facebook, but I still show up as an active user because I'm not done with the process. If you're going to measure a phenomenon with data, you have to understand what you're measuring and have an instrument that will detect that. Otherwise, it's just sciency voodoo.