The viral case is when the value you can offer to paying customers actually increases as more free users join the system. In this case, the free users are the resource and as long as the revenue you can extract from their added value exceeds the cost to support them, you can do freemium profitably.

The second case is non-viral and is where most people are. Here, the basic premise is that the free users may A) become paying customers or B) attract paying customers through word-of-mouth/scale halo effect.

In both cases, you need to estimate the lifetime cost of a free user. Even if it's only $0.25 per year, it will be something and is important. You will also need to estimate the lifetime profit of a customer (lifetime value minus lifetime cost). Don't forget user/customer support costs and overhead.

For case 2A, you then calculate the conversion rate for free users. If the lifetime profit of a customer exceeds the lifetime cost of a free user divided by the conversion rate, you win. If not, free users are a drag on profitability and not worth it.

For case 2B, you again calculate the lifetime profit of a customer and compare vs the lifetime cost of free users divided by the expected referral rate free users generate. If the profit exceeds the referral cost, you win. Otherwise, again, free users are not worth it.

Obviously, the above is very rough overview, and in real life there are many additional complexities. There are various ways these cases may mix, the functions may not be linear, etc. But in the end, it's just math and you can guestimate if freemium is economical for your business.

This was point A.

An often overlooked (and super important) effect of offering a free plan is that it will decrease the number of paid users signing up to your service initially. In other words: you will make less money upfront.

If the free plan is available that's what most people will go for. It seems obvious but it's easily overlooked when trying to calculate whether or not it's worth offering a free plan. If you factor in the initial negative impact of going free as a cost then you're better prepared. Of course, the reason why I wrote the article was that most people (including myself at the time) simply add a free plan because it "works" for [random web 2.0 startup] so they figure it'll work for them too.

For example, you could argue that Google Analytics is a Freemium version of Google Urchin. That costs $9995: http://www.google.com/urchin/pricing.html

(Of course, Google has a different business model for Analytics - basically they want to commoditfy the complements to their advertising business. But my point here is that freemium can work well in enterprise software with less than 100K active users if the pricing is correct)

http://jacquesmattheij.com/Handbook+Freemium <- shameless plug.

I'd argue the opposite. Get in there and start charging right away. Get as many customers as you can before some buffoon comes in and gives it away for free. If you're lucky, you'll establish that the market price is > 0, and future competitors will charge in a likewise manner.

You missed what I consider the most important question of freemium -- does having a large user base, in and of itself, offer value to the users. I come from the dating site world. Let me tell you, on any social site, if you have a million free users and 10,000 paying users, you're fine; if you have 10,000 paying users and no free users, you're dead: the paying users have nobody to talk to.

In one case we could assume no competition or weak competition in the product's market.

Over time, one can imagine a case including competition that disrupts a once-increasingly-valuable product to suddenly drop users.