Great article by Dan Martell about how to use user activity streams and cohort metrics to learn from your app usage and build a tighter feedback loop. The article is full of concrete examples.
The main context within which very personal metrics (the user activity stream, principally) is useful is when you don't have enough data to conduct high-level tests or analysis of traffic, for example, at the very beginning of a startup. But even later on, it can be useful to drill into a few random users and figure out what they did and whether it uncovers an issue in your product.
Of course, this type of spying on users should only be used where there are no confidentiality issues, and with appropriate anonymisation where relevant. For a twitter helper application, though, it's very reasonable.
Finally, don't forget that metrics are only valuable if they are actionable and interpreted correctly.
