This is the piece that named the problem you are asking about: the difference between numbers that make you feel good and numbers you can actually act on. Eric Ries (the Lean Startup guy) wrote it as a guest post, and it is still the clearest short read on why a traffic spike or an upvote count tells you almost nothing. Treat it as a starting point for deciding which one or two numbers your launch should live or die by.
A metric is only useful if a change in it tells you what to do next. Total hits and signup counts almost never pass that test.
Cohort analysis (following a group of users through registration, trial, and purchase over time) shows whether your launch actually changed behaviour, or just briefly inflated the top of the funnel.
Look at per-customer and per-segment numbers, not one big aggregate, because a healthy total can hide the churn and drop-off that decide whether a launch worked.
