Joel has said in the past that nothing has increased sales more than adding new features. I've always wondered how he reached that conclusion and what data he used to arrive at that. (it doesn't necessarily follow that if subsequent releases generated more sales and had more features, then it's because of those features that sales were made).
Generally I've found that simplicity may not create marketing buzz for your product but will result in more people actually using it (sadly people using != more sales all the time). That's why people still use excel instead of the enterprise crm/erp system that their COO bought.
The comparison is between improving the product and traditional marketing. We found, over and over, that improving FogBugz (sometimes by adding new features, but not always) was much better at increasing sales.
Recently, that's begun to change somewhat, mostly due to patio11 coming around periodically to tell us everything we're doing poorly in our marketing.
From the end user's perspective, features often reduce complexity, sometimes dramatically so. Before "mail merge" was a common MS Word feature, collating form letters was a tedious job.
Good engineering and design integrates features into the application in a way that avoids complexity.
This was implied by Duff's use of the word "necessary" [sic]. I believe his point was that it's possible to have features and complexity.
So although you're right that providing both simplicity and feature-richness is not free, the companies that can afford both (assuming the features address actual problems, or even just pain points, that customers have) are going to fare better than those that don't.
Generally I've found that simplicity may not create marketing buzz for your product but will result in more people actually using it (sadly people using != more sales all the time). That's why people still use excel instead of the enterprise crm/erp system that their COO bought.