I'm doing some research right now that involves heavily mining state campaign finance records in Colorado. For the most part, I'm just using the Colorado Secretary of State's search engine. It's not great -- the server is slow, the interface is awkward, candidates sometimes appear as people and other times as committees, etc. But it's functional.
Some friends have pointed me to followthemoney.org, the website of the National Institute on Money in State Politics. This is a searchable database across all states over many years. Every state has its own way of and schedule for collecting and reporting this information, so what the folks at NIMSP have done is truly amazing.
That said, there are inconsistencies. I've looked up several Colorado statehouse candidates in both the Colorado and the NIMSP database and come up with different numbers. For example, Linda Stahnke ran in the GOP primary in the 17th Colorado House district in 2004, losing to Mark Cloer. The Colorado Secretary of State reports that she raised $30,186 for that contest; NIMSP claims she raised $35,218. There are lots of small discrepancies like that.
I'm guessing the Colorado numbers are more likely to be correct, but I don't understand why these discrepancies exist. Different ways of summing the numbers? Errors? Definitely something to think about before jumping into the data.