Can you detect any patterns? Thinks that make sense, things that don't make sense yet?
Do "converting terms" come into different pages than the usual suspects?
Can you see a lot of "typein" conversions?
Having looked at logs for dozens of clients, these are they type of patterns I look for.
Have someone who's been around look at the logs and reports. Often patterns emerge over time.
Also, for lack of better terms, there is what I refer to as the small number problem. If you have 10,000 real visitors per month with a 3% conversion rate, that means 300 customers. Many of those will are likely to be typeins from visits before so they are mysteries.
So if that is half, that leaves 150 - this will be split over 10 - 20 terms most likely. Subsequently 7 - 15 conversions per term can start to look significant. Sites w fewer visitors or different metrics show thin enough numbers so you can't really see the patterns, or be able to see causality.
Plus, there are seasonal and other complicating factors that can throw things off.
Do you use PPC conversion tools? This can often provide instant insight, but again, you've got to spread things out a lot and let things run for a while before you get enough data.
This would be a good thread to dig into ROTs (Rules of Thumb) that people use to determine causality.
What are your ROTs?