"Anything given for free should only give a taste, at a clearly limited scale. Not all features should be given out in the free version."
Why? What justification does this have? Especially when network effects come in to play (and if network effects don't come into play, you're getting the modern web startup wrong), the value of adding more free users by making them happy is probably greater than the value of frustrating them with limited features until they buy a paid version.
Having only a percent or two of your users be paid is fine, if all of the free users increase the value of your site enough that it's worth it for the paid users, and the paid users can cover the costs. Being too limiting on free users is more likely to frustrate them and make them leave than make them convert to paid users.
I've seen plenty of services do just fine, in which the free users get all of the same features, and even generous amounts of storage space (or other appropriate limits), with the paid version only being necessary for extremely heavy users.
Anyhow, these principles seem to be plucked out of the air, completely untested, and without even a vague critical analysis (like, maybe punching some numbers into a few different models and seeing what you get). Why should we care?
No, these principles are an excellent distillation of lots of stuff I've seen elsewhere and gleaned from talking with founders and CEOs.
(DISCLAIMER: if you're assuming a scale business like Facebook or Twitter, these are indeed terrible ideas. This is for business that make money by charging their users.)
1) "Anything given for free should only give a taste, at a clearly limited scale. Not all features should be given out in the free version." - this is important because you don't want to cannibalize your paid users. If someone likes your product enough to pay for it but everything they need is in the free plan, they're not going to give you any money. The reason airlines make coach seats so cramped isn't to make more $$ off of coach tickets, it's to ensure that highly profitable first and business class customers pay a lot of money to avoid uncomfortable coach seats.
2) "There should be a 'recommended' plan." This makes buying easier. Don't buy the el cheapo plan and regret it, don't get ripped off by the big one. Happy medium, no thought, bingo.
3) "There should be a 'cheap' plan for the skeptics, that includes all of the essential features but at a limited scale." Some people just don't need what the recommended plan includes. For Wufoo, the middle plan includes multiple users, unlimited forms, and payment integration. I just needed a fourth form so I took the $10 plan. I love Wufoo so much I had no problems paying that.
4) "there should be a 'hail mary' plan" - see #1. Wufoo and 37s have both said they make most of their money from the high end plans, even though there are very few of them. If those customers could pay 75% less and still do what they wanted, it would kill the business.
5) "customers should be able to take any of the paid experiences for a trial run." - this is good marketing and accepted practice.
6) "There can be 'filler plans'" - see #3.
Another point from my experience. Unfuddle offers source hosting + issue tracking + wiki, etc. Here are their plans: http://unfuddle.com/about/tour/plans . The micro plan would have been fine but I wanted SSL since it's for my business. It's important enough that the extra $15/mo was not a problem, even though it's a blatant pricing tier move.
If you're really creating value (as opposed to your users creating the value like in Wikipedia, HN, etc) then people will pay for it. Unless you give it away for free.
They're not principle. They're at best, guidelines. Things that have worked for some people, but might not work for others. It's also important to realize that completely different approaches have also been successful.
Take your example of unfuddle.com for example. Their model seems to be working for them, but it's quite a bit different than the pricing model used by GitHub (as another example), which also seems to be working. Github gives out a heck of a lot more than a taste for free, but have still managed to find a pretty effective approach to market segmentation.
I think it's also pretty important that in most cases, the "taste" you give out for free, is a big enough taste to be enough for the customer to want to bother tasting. Otherwise you're not really using a freemium model, and you really aren't offering much more to the customer than you would if you offered them a slick powerpoint and a demo site.
Github hosts a lot of your code for free as long as you're ok with everyone seeing it. If you want to put your company's source on it, you need to pay for private. It's an even clearer pricing tier segmentation than Unfuddle's.
You're spot on pchristensen, the assumed goal for these principles is to build a $$ generating business (i.e. not the VC-backed long-term+scale model).
"Anything given for free should only give a taste, at a clearly limited scale. Not all features should be given out in the free version."
Why? What justification does this have? Especially when network effects come in to play (and if network effects don't come into play, you're getting the modern web startup wrong), the value of adding more free users by making them happy is probably greater than the value of frustrating them with limited features until they buy a paid version.
Having only a percent or two of your users be paid is fine, if all of the free users increase the value of your site enough that it's worth it for the paid users, and the paid users can cover the costs. Being too limiting on free users is more likely to frustrate them and make them leave than make them convert to paid users.
I've seen plenty of services do just fine, in which the free users get all of the same features, and even generous amounts of storage space (or other appropriate limits), with the paid version only being necessary for extremely heavy users.
Anyhow, these principles seem to be plucked out of the air, completely untested, and without even a vague critical analysis (like, maybe punching some numbers into a few different models and seeing what you get). Why should we care?