Deciding What Features to Implement: Let Features Emerge From User Behavior
Part two of the series “Deciding What Features to Implement”. In this edition I observe a powerful new class of features that emerge from the aggregated behavior of a web site’s users: emergent features.
This is part 2 of the Deciding What Features to Implement series. Part 1 is about choosing win-win features.
In part one of the “Deciding What Features to Implement” series (it became a series when I wrote this one), I suggested a super quick method of assessing the relative value of a feature you’re considering for your web site. In short, the method is choosing only those features that are a win-win: a win for both the users of a site and the site itself.
My suggestion, however, assumed that we had a nice, orderly list of features from which to choose. If it were only so easy! Designers rarely have this luxury. In some cases, the domain a site is in is so new or different that there are no existing features to copy or improve upon. In others, your audience or business model might be unique enough to warrant completely unique features. In still others, you might have exhausted the “expected” features of sites like yours and are hunting for new ones.
Unique Features from Unique Behavior
One class of features that will always be unique to your audience are the features that are built from their very behavior. Features like these are “emergent” in the sense they emerge from the collective aggregation of user behavior.
My favorite feature of this type is the list of popular bookmarks on the site del.icio.us. Oddly enough, this powerful feature has nothing to do with the reason why del.icio.us is currently so popular, namely, the folksonomy. No, this feature is simply a listing of the bookmarks that have been stored most frequently in the recent past.
The power of this feature is that it is leveraging the aggregated behavior of users to create a list of compelling content. Taking a look at the list over even a short period of time will convince you of its value. It is so good that I use it in my normal news gathering routine alongside established news outlets like the New York Times and CNN.
Works on All Kinds of Sites
What excites me about features like this is that they exist, or could exist, on nearly every single web site on the Web. Each site has its own set of users doing something particular to that site, and in many cases it would be helpful to those users to have a better idea of what everyone else is doing.
Because they are dependent on the users of a particular site, many of these features will be unique to the site itself. In some cases, it may be impossible to replicate them on another site. Even so, there are several general characteristics that most of these features will have:
Features of Emergent Features
- They’re based on actual, recordable behavior of users on that site.
- They’re often in the form of a list, and are therefore easy to form into an RSS or Atom feed
- They facilitate social proof by showing what other users are doing.
- They’re based on real, aggregated data, therefore they reflect what’s currently going on and change accordingly.
In other words, these features are simply observations given back to the people to help them make future decisions. Not a big deal, right? Well, take a gander at this list of new features cropping up on some of the latest services as well as some that have been around for a few years. These companies are relying on these sorts of features in a big way:
- Flickr’s everyone’s photos and popular tags
- Technorati’s Top 100 and Booktalk sections
- Amazon’s most popular computers and most wished-for items
- 43thing’s goals and city zeitgeists
- Cnet’s The Pulse and top movers in downloads
This list is only the tip of the iceberg: these features are everywhere, and they’re growing in number. The next time you’ve got need for something new, think about how you might implement something like these. Emergent features are nothing more than telling your users more about themselves by showing them their own behavior.
Currently working on: