Nice example, but a major pain point I see is dev time. There is going to be significant amount of time being spent making the analytics platform, and then scaling it. It might be better spent on the product itself until it takes off.
Really? Mixpanel[1] is really nice for doing it, doing it realtime, doing it across platforms (not just web), and adding the cohort analysis[2] (which yields the most important non-vanity metrics).<p>[1] <a href="http://mixpanel.com/" rel="nofollow">http://mixpanel.com/</a><p>[2] <a href="http://news.ycombinator.com/item?id=2986186" rel="nofollow">http://news.ycombinator.com/item?id=2986186</a>
I fully agree. We track millions of events per day and I use hive and hadoop to do my analysis. Not as user friendly but putting the data into a pivot table helps alot.
The thing about analytics is that it's BIG data. A single user event can cascade into multiple datapoints that you either have to keep normalized (like in the SQL example) or denormalize into something more manageable. The latter of which is really the only way to handle TBs of data, and it's not trivial. It's a fun job, but it's full time.
Nice to see an article on this subject. I'm of the thought that basic analytics are an extension of an app's/site's dashboard. You do have a dashboard don't you?