There isn't really a universal standard on this. In many ways, it depends what sorts of behaviours are particularly important to your organization, and the kinds of interaction you anticipate.
When others do this, they typically (though not exclusively) look at a few common things:
- They check how much time was spent on a certain page or view
- They count the success rate for a particular task or form
- They count how often a particular feature or page is discovered
- They see how often a view or process was cancelled or exited
But these stats alone aren't always easy to interpret out of context. If a user doesn't visit a page or view very much, is that because they haven't discovered it? Or are they aware it exists, but just don't care? If a user spends a long time looking at data, is that because they're engaged with the content, or because they're struggling to locate what they want? Analytics can tell you plenty about how people click. But it can't tell you why.
That's not to say analytics is useless. It can expose issues and risk for further investigation. But they must be read with a certain healthy scepticism. And they should never be used as a substitute for genuine user testing.