Apple is living proof. Some large chunks of the iPhone are actually done by half a dozen engineers. When I was a PM at Microsoft, our whole team was something like 40 people on data synchronization.
Final Cut Pro was around 15, which is still incredibly small given how big our application was.
Apple lets go of a number of things that Microsoft chases such as a religious concern for backwards compatibility, globalized features (not just localization but geo-specific features), and deep product extensibility with corresponding dev support. No judgment on those decisions but I'd argue that the size of MS teams isn't directly reflected in the product or features most folks consider as users of Microsoft's products and I wouldn't criticize MS based on those team sizes (as the author of this article does). Instead, I'd criticize the PMs or other managers inability to bring focus to those teams to clearly tackle the right problems at the right level of investment at the right time. It takes a lot of discipline to say "no."
Then take into account all the people who do QA, all the management, marketing, business, finance, and other teams.
And you're actually left with a pretty lean engineering force. Apple engineers more stuff in house than a typical company like Dell, so that plays into the head count as well
- First, there's too much ego associated with features. If a feature is someone's 'baby', it is just very hard to get them to accept that that feature needs to go. Even with all the data in the world, it becomes confrontational and ugly very easily.
- Second, companies rarely tend to reward removing things. It is hard to have a performance review or a stack ranking/calibration meeting where your major accomplishment was removing features. This is obviously a broken performance appraisal model/culture but just happens to be prevailing culture in several large organizations.
- Third, a lot of products and features don't lend themselves to measurement very well. Without measurements and data, deciding what to remove comes down to subjective opinions. Contrast this with new features where you often have lots of data (emails from users, tweets, market research, etc).