Which is better for a startup company: Web Analytics or End-User Experience Monitoring?

As analytics tools have become more developed and pre-packaged (e.g. Google Analytics, WebTrends, etc), it's easy for companies to focus on web analytics - traffic patterns, traffic sources, navigation path, visitor segmentation, etc. Conducting a baseline analysis is critical. But how do you improve from there? Let's say you find people are not clicking on a link off the home page that you would like them to find. What should you do? For the sake of speed and efficiency, you can go straight to testing a new design implementation without speaking to a single user. With enough resources you may even be able to do some simple A/B testing of new designs along the way. This is the beauty of the web world and you should take advantage to act quickly.

But let’s say you’re launching new functionality that will represent an investment of hundreds if not thousands of staff hours. You can’t afford to get it wrong. Bringing in users, watching them interact with your site, presenting them with prototypes, being able to ask questions, etc. is invaluable to a good product development process.

Perhaps I’m old school but live usability testing is a lost art. It’s easy to identify problems when users raise their hand to tell you about a problem (e.g. user reviews in the iTunes store for a specific iPhone application). The more pernicious problems are ones where users don’t or can’t tell you what’s wrong. Even more likely, they’re satisfied with the experience but are not taking an action YOU want them to take. There are still companies who are religious about usability testing, and more broadly, interacting with their customers. A good example is Intuit’s Quicken and QuickBooks groups. It’s not just watching their customers use their product but it’s also trying to understand their needs in the first place. This drives product innovation.

Over time analytics and user testing should be part of a virtuous cycle. The following is a sample cycle for a released web site or online product -

Step 1 Analyze current state (web analytics) and identify issues

Step 2 Triage simple issues that are low cost to fix; test design tweaks using a/b approaches; iterate quickly

Step 3 For more complex issues, conduct user testing (online or live user monitoring, user interviews and surveys, etc.). Introduce prototypes to address issue

Step 4. Evolve prototypes and retest with users

Step 5 Implement changes

Step 6 Start the process all over again

All of this said, it seems like there are better ways to get real-time user feedback. I wonder what others consider to be best practices here. A lot of this will depend on the type of product you have as well I would imagine.