When building a commercial web application, what are the factors that go into the decision of a revenue model?
How many users are necessary for an advertising-supported application? What demographic factors of the user base makes them more attractive to advertisers?
Conversely, when deciding on a subscription price, what factors go into setting a price.
This seems like a decision we would want to get correct!
It's a difficult question with no easy answer - if you run an ad supported site then you will need a high volume of traffic to make money, and your system will need to scale more (so will cost more). And your revenue will be unpredictable. Unless you're going to be the next google, and you're probably not, then this is a fairly dicey approach IMO.
On the other hand, a subscription model means you just need to focus on growing your subscription base, and only need to scale to accomodate that base. So both your income and scaling needs are more predictable.
Though, you may need to offer a free version and generally tier your prices to let people try it out and to generate some buzz, so you may need to scale quickly anyway.
You need to start with some realistic (pessimistic) estimates of how many users you will have, what they might be willing to pay, and how much it will cost you to serve them adequately (including your own costs and profit).
On the general topic, this talk from DHH (author of rails) is pretty interesting, and makes a good case for a modest subscription model.
"How many users are necessary for an advertising-supported application?"
R / (N * C * I)
where
R = Desired revenue, N = Number of ads a user sees, C = Click through rate, I = Income per click through
"What demographic factors of the user base makes them more attractive to advertisers?"
Obviously their disposable income and their willingness to spend it on the advertised products.
"Conversely, when deciding on a subscription price, what factors go into setting a price."
Again the desired revenue, and the expected number of users willing to pay the price. Note that pure subscription models are pretty much dead all over the web. The only thing that seems to work is to offer the main service for free and then offer improvements to paying customers. An example are games where you can play for free, but only subscribed users can use certain items or character classes, which makes the game more interesting for them (without making it impossible for non-subscribers to compete).
I've worked with several companies that tried the subscription model (and one I refused to work for because they really had no specific business model). The nice thing about it is the revenue is yours to make or not. With banner ads, part of your success is dependent on the quality of the ads, and the products and services behind them. There's also risk of arguments with advertisers on click rates, impressions, and how your random ad picker isn't showing their ad enough ;)
The subscription model also lets you fine tune your value to the consumer vs your revenue. If you offer a base level of services for free and charge for extras (LiveJournal, FaceBook, LinkedIn,...), you can move existing services or add new services to either group as you see the demand for them, and have different paid tiers. You can even offer features on the free side for a limited time to feel out the demand, then figure out whether to add them into a paid subscription level, leave it free, or ditch it altogether.
It does mean, though, that you need to watch the usage of the different features, and your ratio of free to paid users. The company that I refered to earlier was a start-up, but they had no idea what features they were going to charge for, how much they were going to charge, or a projection of the number of free or paid subscribers. They had no idea what they were doing, so I ran away screaming.
One downside of the subscription model is there's more work involved in collecting the revenue from hundreds or thousands of users than from tens or dozens of advertisers. Once you got everything set up, though, it should be pretty automatic except for when credit cards fail, etc. but you can just shut off the paid services for that account until it gets worked out.
Best of luck in your venture.
You could also go for the hybrid solution where you initialize the system with some advertisements but give your users the option to pay for a subscription without the ads. If you go for advertisements, you might want to start of with Google AdSense or something similar since they take care of alot of stuff for you and often provide you with the means to better target your ads.
You might then wish to create your own targeting system...
As for the demographic info attractive to advertisers, it depends. If you can convince them that you have an AI targeting system, it may help. But AdSense already seems to provide such a thing, but it is limited in power.
Like dj_segfault was saying, if you need to find out any info, like the optimal subsription price, try some out, log the results, and infer the optimal cost. You could even automate this process. Although it might be difficult to implement, it would optimize your decision process.
The beauty with a login website, is that you have all this data about your users. If you implement the right tools, you can optimize your advertisement management system, your subscription system, etc.
Yet all this is very dangerous for it may get out of hand if you are not careful; it is a double edged sword.
While revenue is important, think about your users before anything. If you can, create, establish and maintain the website for them, not for you or for your shareholders; exist for them more than for your own self.
Do not maintain a web site in order to make money; make money in order to maintain a website.
Hope this helps.