Omniture stumbles

Chatter is building on the interwebs about Omniture’s recent (and ongoing) latency woes. Looks like both SiteCatalyst and Discover are days behind in processing data (according to messages on Twitter, up to around 5 – 7 days in some cases). And it looks like the situation is still getting worse, rather than better.

I have no insight into the cause of Omniture’s difficulties, or how widespread they are. It may be that they’re related to the December release of SiteCatalyst 14.3, which seems to contain a number of new features which are fairly broad in scope, and which may have had an impact on the platform’s ETL stability. Behind the scenes, Omniture may have made some changes to start integrating HBX’s feature set (especially its Active Segmentation) into SiteCatalyst as a prelude to a final migration push for the remaining HBX customers. Omniture’s certainly not saying – they’ve been conspicuously silent since the start of these problems.

Whatever the cause, I can certainly empathize with this kind of situation – we had all sorts of difficulty dealing with latency issues in my WebAbacus days. And we can be confident that Omniture will (eventually) fix these problems, and will probably not lose very many customers as a result (though, in the teeth of a recession, it can’t be great for attracting new customers).

But do these problems tells us something more about Omniture’s (or any other web analytics company’s) ability to run a viable business? Infrastructure costs are a big part of a web analytics firm’s cost base (at least, those with a hosted offering, which is all of them). And unfortunately, these costs don’t really scale linearly with the charging method that most Enterprise vendors use – charging by page views captured. Factors like the amount a tool is used, and the complexity of the reports that are being called upon, have a big impact on the load placed on a web analytics system, and the resulting infrastructure cost. It’s tricky for a vendor to recoup this cost without seeming avaricious.

As Omniture’s business grows, it has a constant need to invest in its infrastructure to keep pace with the demand for its services. But as the economy has worsened, it must be terribly tempting to see if a little more juice can be squeezed out of the existing kit, especially with its 2008 earnings due later this month. This will be as true for any other vendor (such as Webtrends or Coremetrics) as it is for Omniture, and these remarks shouldn’t be seen as a pop at our friends in Orem. But the nub is, can Enterprise web analytics pay the bills for its own infrastructure cost? Or will all web analytics ultimately need to be subsidized by something else (such as, oh, I don’t know, advertising)?

Your thoughts, please.