The best web sites feel smooth and predictable, yet powerful and deep. They achieve this by presenting a limited set of well-chosen interactions that support users throughout the experience. This lowers learning curves to near zero and allows users to concentrate on the content, which is what they’re really after.

As web sites evolve and offer users new interactions, it’s important to take stock of how they behave, and whether those behaviors are consistent. It may be easy to achieve consistent interaction design for a new site, but when more established web sites introduce modern interactions it is frequently in an ad hoc manner, and users may be presented with a variety of methods—new and old—of accomplishing similar tasks. The resulting site can feel jagged and unpredictable, with a steep learning curve and low ease of use.

To correct this problem, designers may establish design patterns to describe the appropriate interactions a site should offer to support users’ objectives. But before that can happen, it’s important to understand the current types of interactions being offered.

Using a recent audit of the interaction design of eBay.com as an example—an audit of the site’s “feel”—we’ll discuss how collecting and cataloging the variety of interactions can be an important precursor to establishing interaction design patterns for a highly trafficked web site.

We’ll walk through the process used to quantify and describe interactions on the eBay site, starting with an audit of a representative sampling of user flows. Data gathered included variability in system response to user interactions with similar controls, multiple paths to accomplish similar tasks, and different representations of similar data objects. We’ll also share a round-up of key findings and discuss future extensions of the audit, including objective and subjective metrics for characterizing the feel of a user experience.