Open Source Analytics

Last evening I went to Web Analytics Tuesday in Stockholm to listen to Eric T. Peterson, author of Web Analytics Demystified and other books, and meet some people in the analytics parts of the Internet world. Following Eric's presentation (the "resources, analysis, testing, processes" approach to web analytics makes sense) was a panel debate with participants from some of the major suppliers of analytics tools (IndexTools, Omniture, WebTrends and consultancy Satama).

Some takeaways.

The first one has little to do with analytics, but is an observation that there is not one Internet scene in Stockholm. Rather there are several. Startup/entrepreneurs, business people, tech/developers, web analysts etcetera. At the edges there seem to be overlap between the sub-scenes, but not to a great extent.

Going back to analytics. The vendors still feel very e-commerce oriented to me. Most of them have business models where a customer's cost scale in relation to pageviews. That is ok for an e-commerce site with relatively few pageviews and a high value per pageview, but unattractive for a media or community site. If someone wonder why Google Analytics is popular with such sites, it is not only because it is free (to be precise: not free but bundled with AdWords) but because costs don't grow in-line with pageviews.

As the vendors seem to be keen to sell their solutions as services, I see an opportunity to create an open source web analytics product that is hosted by customers. For large sites it makes the solution scale better (definitely on the business side and likely also on the operations side). The business model for such a project would be like the one used MySQL, make money on relatively cheap commercial licenses and support.

The first step for such a product would be nail unique visitors, visits and pageviews tracking at scale (50+ million pageviews/month). Then add additional functionality. The product should be hosted on the customers' servers, as customers operating at large scale usually have the operations capacity to handle it and probably a traffic pattern a local vendor is not setup to handle.