When developing a web app, your app's capabilities are far more important than it's features. Aren't they the same thing? Not in my mind: a capability is something you can do with the product while a feature is something a product can do.
Capabilities are awesome. Features for features sake suck.
The first iPhone was a hit because even though it lacked in the feature department (No 3G, MMS, etc) it let people do things with their phone they couldn't or wouldn't have before. Same with the first iPod. In what has to be one of the funniest-in-retrospect single line reviews of the first iPod, alpha-geek CmdrTaco, the creator of Slashdot wrote:
No wireless. Less space than a nomad. Lame.
He couldn't have been more wrong. We as geeks tend to focus on the features because we think they are important to us, but the truth is that we are users too; what we can achieve with something is orders of magnitude more important than what the product itself can do.
The good news is that there seems to be a strong trend in software development to focus less on features first and rather on user stories.
On of the nice things about user story driven design is that you say "A user should be able to do YYY" and then figure out the simplest way to make that happen.
Feature driven design might say "we should have a web 2.0 interface" and so you spend two weeks creating the perfect developer framework for whatever web 2.0 means - and then the minute you try to add in an additional capability into the site you realize your framework doesn't support what you want to do. It's true at launch time you will get to check the checkbox "Web 2.0" on the back of the metaphorical box but all you really ended up with is code that will need to be supported but isn't at all essential to your users.
Microsoft word is the starkest example of this I've seen. The hell that the Microsoft programmers must go through supporting all the features that no one uses must be a very dark place indeed. I need my word processor to have a very limited set of capabilities:
After that anything else that gets added in is a feature I probably won't use. OpenOffice and Google Docs are slowly supplanting Word because they make the user experience better than Word (I'm counting not having to shell out $500 for useless features as part of that user experience) The first version I used of OpenOffice was awesome because it had, right there in the File menu a "Export as PDF". This was a feature - but it made the sharing of documents so much easier that it greatly added to the "sharing" capability - I could email a document and be sure it would look that same way on the person on the other end of tube's computer.
The poster boys for the "less features" community are 37Signals' applications - BaseCamp, HighRise and Backpack in each case they dramatically underfeatured any of the other project management / CRM / intranet suites out there but had all the capabilities that people needed. They also added the capability of "usable by moms and/or CEO's" because their products were so simple and built-on standard web user interface concepts. That's a capability that you don't ever get by adding features - it's something you only get by having solid UX coherency and a lack of confusing and overwhelming features.
If capabilities are so great, then why do we as programmers tend to spend our times focused on features?
Reason one: features are fun. I love coding in new whiz-bang features and I'd bet other developers are the same way. It means I get to walk over to another developer on the project and say "Did you see the sweet new realtime Ajax notification panel I put in?" and they say "Cool" and I feel like I earned my nerd points for the day. Unfortunately if that notification panel doesn't do anything besides add in some eye-candy it's not a win. It's going to be a weight on the codebase that will need to be carried along without adding in real value.
Reason two: far to often project descriptions are in terms of features rather than user stories. Features are easier to describe, more concrete and more estimatable than user stories. This is something that's easily solved though by making sure every feature relates back to a user story. If each feature you're adding in relates back to a user story (e.g. "Sally B. User should be able to login even if she forgot her password" = "Password Reset Feature") then the creepy features won't get in there.
For a great analysis of user stories and how use them to drive your development process I recommend "Head first software Development" from the wildly entertaining O'Reilly "Head First" serious (yes I just called a set of computer books wildly entertaining - once you've read one of the series you'll understand) Despite the title, it's by no means a basic book on programming but actually a great read on the higher level processes involved in making great software.
Remember, focus on capabilities driven by user stories, letting the features generate from the stories and everything will be ok. By the way, did I show you the cool new totally user-driven HTML5 websockets event mash-up stuff I'm working on?
....and follow @cykod on twitter