Why should a business add burden of supporting additional users unless free users bring something to table in some tangible form like free press?

I am really curious.

If you can't figure out a good reason to support free users too, then don't. Not every app is suitable for a freemium style of business.

The big advantage of freemium is lower friction. The people who buy your product don't typically need it advertised or explained to them. They've either already got experience with it due to being free users or they will set aside some time to trial you product for free for a while. Beyond that a free product can build up a strong reputation that carries over to the paid versions.

Billions upon billions of dollars are spent every year by companies trying to create familiarity and trust in their brands. Being able to do that with a free version of your actual product is just as worthy of the expense and in many ways a far more honest form of advertising.

It's essentially a marketing expense.

I do agree with this. Especially in this case if you outsource some of your marketing "expenses" to social platforms ("Like buttons", "tweet this", etc)

Aaron Levie - Why Freemium Works

http://www.youtube.com/watch?v=7hpi5nUhssY

http://techcrunch.com/2009/07/12/guest-post-is-free-the-futu...

If you're having trouble getting enough traffic to run statistically significant experiments, then offering some kind of free tier can help. Of course, these free users will behave differently, but you can at least tell if users will like it even when there's no payment decision necessary.

Network effects, in the case of LiveJournal and GitHub.

That's like a woman in a burka.

No, not irrelevant, unless "free" users also cost you nothing.

They don't have to cost you nothing, just little enough that you're net ahead. And since startups that use the freemium model invariably manage to make free accounts cheap, I assumed I didn't have to mention that explicitly.

But isn't the percentage of users who pay what determines whether you're net ahead, and not the total number of paying users [1]?

If Dropbox has ten million users paying them $10 / month, but they have to support another billion free users for twenty cents per month each, they're screwed.

1. Assuming you're making enough to cover fixed expenses

Not in practice. Usually startups arrange that the incremental cost of one additional free user decreases as the number of free users increase.

profit = paying users * price per user - total users * expense per user - overhead expenses

Business on the internet seems to be much more about driving up total users while driving down expense per user, rather than driving up paying users / total users.

That's a good point, but it's a startup point, right? At scale, the percentage of the addressable market that is sufficiently served by your free option does matter, doesn't it?

I get that this is a startup article, and my thought reading this article was "well, that's conversion math for you." But you're making a broader point.

You can actually induce your own pressure on the free users by changing your product so that a very valuable piece is now in the paid section. Sometimes you can create your own demand. Have a look at avvo.com. They do it well.

"What matters is the total number of people paying you, not what percentage of your users that fraction represents."

Surely the total revenue from paying customers, relative to the cost of providing service to all users, is what matters^? If so, then this is ultimately a function of the percentage of paying users to free users (assuming cost of servicing paying and free customers are largely similar).

The opening statement of the post, that people who want something for free won't think to pay for it, is something of a non-statement. That free users cost (as the post goes on to argue) is pretty much a given, but it still matters.

^Where (I assume) what matters is a profitable business.

Not necessarily true. The largest costs (especially at small scale) are fixed - the vps/dedicated box/rack/EC2 instance.

Freemium models tend to scale price with bandwidth/storage usage, which would be the largest variable costs. Is serving a million users with barely any data really that hard once you have the infrastructure to serve a few thousand with lots of data?

I do not agree with this.

The percentage of users that convert should be a closely watched metric that is very important to any freemium business.

If you think I disagree with that, you misunderstood what I wrote.

It is probably a case of us arriving to the same result by taking a different path. As I go back and reread your post with a broader viewpoint thats what I'm thinking?

It's possible that you could have a company that has 1% paying users that's MORE profitable than a company with 50% paying users. The size of the profit is more important than ratio of paying to non-paying, generally speaking. (And of course, in either case you pay real close attention to the conversion rate and try to get it up higher.)

The total number of paying customers is what determines your revenue. But even then, it's important to track free->paid conversion percentages, e.g. sudden drop offs.

So what you are saying is having free users is better than no users (even though it is an expense) because a greater proportion will convert (hopefully) eventually?

Write it off as a marketing expense and hope it attracts more paid users?

Paul, the notes from the developer is very much relevant and reflect on the model we expect. The values in the freemium model has depreciated. If only you can derive any significant value over time. Dropbox, evernote has already given out the rewards system (towards all). But the one we would like to model is Rovio mobiles Angry birds.

Expecting to get paid on a depreciating model is lunatic. Without any metric over such - this model works, but expecting to get paid on a depreciated value, never does.

Can freemium work without hitting a home-run ala Dropbox?

Define 'work'.

Figures from several freemium businesses that I am familiar with suggest that this model works on the 'lifestyle' end as well as on the 'more money than you could reasonably spend in a lifetime' ends of running a business, and I presume on many points in between. The trick is not to lose too much money on your free users and to make them pay in some other way, for instance by selling advertising or by engaging them to promote the product to others.

Yeah--check out the background of Bingo Card Creator. There's lots of HN discussion on it, too

http://www.kalzumeus.com/2011/12/21/bingo-card-creator-etc-y...

Great point. I wonder how the Freemium model works for something like Splunk. Do free users constitute a completely different universe or there is significant conversion of free to paying users ?

Well given that free Splunk won't index any useful amount of data, and the pricing is so high that it can only work for corporations, I can only see Splunk working as a license-enforcing trial.