Wednesday, May 28, 2014

The Three Commandments of Web Site Feature Development

Lesson #1: The 90-9-1 Rule - The rule breaks down like this: 90 percent of the readers of any blog or Web site will never leave a comment, nine percent will comment once during their entire tenure of readership, and one percent will do the vast majority of the commenting. If you think that sounds unreasonable, consider any of the radio call-in shows you may have listened to in your life. How many calls do those shows get, as a percentage of their total audience? How many times have you called into those shows? Yet there are regular callers; they're just a tiny percentage of the overall audience.

I've found the 90-9-1 rule to be strikingly true online and, if anything, optimistic in the percentage of the audience which are regular posters. And that's just for a very basic interactive activity like posting comments. The percentages drop steeply as you get into interactive functions that require more time and effort, like filling out profiles or writing reviews. I took a stab at setting interactivity expectations here, based on what we learned at TechRepublic with some flamingly unsuccessful blog, profile, and social bookmarking projects. I can say with a straight face that getting a decent user-submitted video is literally a one-in-a-million proposition.

Moral: Design for the 90 percent, not the one percent, if you want to actually see an increase in activity, visits, and traffic. Focusing on the power users will almost never move the needle, especially since your usage zealots are already doing almost everything they can or will on your site.

Lesson #2: Design for an Audience of One - Flickr and YouTube get a lot of hype for how 99.999 percent of their content was acquired for free from users, and how the users employ tagging and groups to create wonderful emergent communities, content, and traffic bursts. What people don't talk about is that those are side effects of YouTube and Flickr's business model and use cases.

The vast majority of YouTube's initial users didn't give a crap about making viral videos or monetizing video content, they just wanted an easy way to format videos and post them online. Almost all Flickr users don't care about aggregated group feeds or discovering like-minded photogs via tags, they just want an easy way to post and store pictures online. Flickr and YouTube have value to me even if I'm the only guy using them. All those group-dependent features are a result of Flickr and YouTube's scale. You can't start with those features, you tack them on once you're massive.

Moral:
Any feature spec that includes the phrase "will be useful once a bunch of people join in" will almost certainly fail because there is no value for the initial users.

Lesson #3: More of the Same is the Only Feature That Matters - Nobody bookmarks a page anymore. They either search Google directly for what they want (and don't care where it comes from) or get it sent to them in an RSS feed or social network from whichever sources they prefer. Again, your zealous users are a tiny percentage of your audience, so any efforts made to be the be-all, end-all of your audience's activity are likely to fail because most of your users aren't devoted to your site and prefer to go elsewhere.

That upvoting feature that works just like Reddit? They'll use actual Reddit instead. Your blog platform? If they wanted to blog, they'd use WordPress, Blogger, or any of the other services out there first. A user profile? Maybe you've heard of Facebook. Until you reach the stratospheric heights of traffic, there's no point in trying to create new user behaviors on your site. The best thing you can do is reinforce the existing user behaviors. If they came to you for content, the best thing you can do is show them more content. If they came here looking for help making a buying decision, help them make a buying decision.

Moral:
Any feature spec that includes the phrase "if we can just get the users to do X" will fail, because if they aren't doing X already, they aren't likely to start.

[This post was originally published on Jan. 18, 2010.]