Identify what works and what doesn’t – then see if you can see why.
This includes not just the « keyword » or « intent » terms, but the nature (medium/format) and the performance over time. You should get a bunch of queries that work poorly, and a bunch of those that you do well with.
You can then start looking at variants (makes, models, locations, time periods, etc.) that you can research and look for opportunities.
I’m also a big fan of using it to see potential looming failures – pages that are dropping for terms, etc. This can indicate a shift in intent, change of meaning, or more technical issues that need to be looked at.