Omniture: On-Demand Web Analytics


I spent some time with Omniture this afternoon. I knew these guys pretty well when they were MyComputer.com. Omniture (nee' MyComputer.com) has always been in the ASP space (now called "on-demand applications") and they were one of the early innovators. I remember seeing Josh James at least a dozen times on the plane between SLC and Silicon Valley in the old days. I didn't really believe that the ASP model would work for analytics then, but now its the only model I can imagine using. I'm not alone in that--many Fortune 500 companies use Omniture and an on-demand application model for analytics.

The biggest reason this model is a winner in analytics is what's called "global roll-ups." By that, I mean that Omniture and other on-demand analytics tools allow Web sites on multiple hosts and with different domains to be treated as a group and analyzed as a group. At iMall, we spent a lot of time building tools to create analytics from dozens of hosts and domains. If I were making that decision now, I'd just use Omniture or something like it.

A recent Jupiter Research study asked CIOs what web site technology they were going to deploy in the next year. Web analytics was first with almost 60%. Search technology was second with 52%. This isn't too surprising--people need to measure things to justify ROI.

I've used Webstat.com to track visitors to this blog and my UtahPolitics.org blog for over a year now. The idea is the same, but Omniture is much more sophisticated. They're customer list reads like a who's who of online companies. The sophistication is based on custom variables that are put on pages so that Omniture can segment traffic according to things that aren't typically part of a standard Web page request.

Omniture's reporting it really impressive. When you see how they generate fallout reports on custom tracks through a site, you'll understand why they have the customer list they do. This kind of information is pure gold and not something you get from run-of-the-mill Web analytics tools.

Understanding behavior relies on not just seeing what visitors do, but also understanding the changes in behavior in the context of what is happening on the site. Say, you're drop-off rate decreases. You'd like to know why. That implies that the Web analytics ought to be tied into the content management system in some way, or be able to track changes on the site to show behavior changes in light of site changes.

Omniture doesn't really solve this problem (one of the downsides of on-demand models), but they have a tool called A-B comparison. The tool lets you pick a date range and show metrics from those dates side by side. To see the behavior in light of page changes, you'd have to encode those changes somehow in the page variables. This may not be manageable for random page changes, but its very manageable in a situation where you can randomly show different users different pages and use A-B comparison to measure the response to those pages.

I was surprised to find out that Omniture doesn't yet have a way to feed real time analytics information out to their customers. Right now, its pretty much their dashboard or nothing. I'd like to be able to take their data in real time and combine it with data from my business in something like Iteration. That's the future of business intelligence and a critical feature for any company hoping to sell me analytics of any kind.

I suspect that Omniture is trying to figure out how to provide the data and still provide sufficient value that they don't get disintermediated. The answer is to not provide raw data. Instead, provide report building tools that allow me to create custom data feeds that merge, filter, and sort the raw data into custom reports that come across as XML instead of HTML.

I haven't really followed Omniture for a couple of years, so its was good to catch up with the company and their technology. Another positive data point in the on-demand application space.