I think a lot of people could do better by looking at the range of customers for their product, a-priori deciding that some segment of that range should pay money and that the other segments can be used for harvesting attention (or links, or what have you), and then crafting their pricing chart to match.
One teacher responding to my customer survey recently wrote the comment "15 cards is too few for a teacher to use the free plan. It should really be 25 to be appropriate for a classroom." Indeed, it is almost like I planned it that way...
This will probably have better results than the typical "Plan A gets you 5 units of goodness, Plan B gets you 10, Plan C gets you twenty!", especially as customers from all of the plans tend to waaaay underuse their allocations anyhow. (Related thought: your plan names and the story they tell are probably as important as the numbers under them, if not more so.)
Frankly, your short comment was more enlightening than the article itself. It's so easy to say that with the right features, upgrade prompts, and pricing, things will work out -- much more difficult to actually do.
Offering a freemium model based on something like a simple metric (such as the 2GB example for dropbox) can work - for a while - but it will also become an Achilles heel.
Microsoft (hotmail) & Yahoo for the longest time offered < 10MB of storage, enter Gmail with 1 GB instantly changing the game.
So if you decide to make your free to paid conversion dependent on a single metric better have a battle plan ready in case someone decides to start giving away that which you sell.
Better than just a single metric is a combination of a lot of items, things that people will not feel the need for at first, but over time they'll wish to have those features. That's good, especially if the need arises at a point in time when the cost of switching becomes higher than the cost of paying for the extended features. Make sure that it is really value for the money they pay, not just some pacifier or 'feel good' feature (you're supporting the site, or something like that doesn't cut it, there has to be a balance).
If you combine freemium with ad support then make sure that your paying users never ever see another ad, and put a 'get rid of this ad' link near the ads, leading to a page that explains the benefits of signing up (other than removing the ads).
Essential features should be functional in the free package, but stuff that is hard on the server or that requires a lot of capital outlay can justifiably made part of the paid offering. Users will understand that, especially if you make some effort at explanation.
One teacher responding to my customer survey recently wrote the comment "15 cards is too few for a teacher to use the free plan. It should really be 25 to be appropriate for a classroom." Indeed, it is almost like I planned it that way...
This will probably have better results than the typical "Plan A gets you 5 units of goodness, Plan B gets you 10, Plan C gets you twenty!", especially as customers from all of the plans tend to waaaay underuse their allocations anyhow. (Related thought: your plan names and the story they tell are probably as important as the numbers under them, if not more so.)