If you are spending money on measurable advertising (ie. CPC) and you know that the return on investment is worthwhile (ie. the revenue generated by converted users is greater than the amount spent on advertising) then how do you decide how much to spend?
The context is a SaaS application where the actual revenue is realised over a few years. Do you spend as much as you can within the confines of your available funds, or is there another way you decide this?
You are deciding on an investment that will bring a return in time. One way to approach this problem is by using discounted cash flow analysis.
If you invest 100 dollars in advertising, and if you have your estimates of the return in time, you can estimate the CDF of that return (for a given reference discount rate, for example based on the rate at which you could borrow money). If the CDF is higher than $100, your investment would be worthwhile.
That said, you also have to keep in mind that with advertising at some point there is a diminishing return. So for example it could be that spending $100,000 your return for ad dollar is less than at $10,000. Theoretically, you could invest up to the amount where your CDF per additional dollar becomes 1. In practice, it is impossible to know where that limit is beforehand, but you can make some estimates.
For CPC advertising (eg Google Adwords), moreover, there is a "physical" limit to how much you can spend for a given CPC, based on how many searches and visits are made for your keywords, how much competition there is, and so on. This means that your cost of acquisition get higher the more you want to spend - you can estimate that using adwords tools (traffic estimator etc.). With these it is pretty easy to estimate when your additional CDF minus the cost of acquisition becomes negative, and so when to stop.
Last thing you should keep in mind is your alternative uses for your money. For example, improving your website or service could bring a higher return per visitor. You should balance that with what you spend in ads.
If the conversion per unit capital invested were linear, regardless of amount, then you'd spend all that you can. However, it's not linear by any stretch of the imagination, at some point you will saturate and not generate any new conversions for the unit capital spent. So, you might approach the problem by asking how large your target market is, and then how much of that market is exposed to your product with a measured block of advertising. After all, once someone has seen your message 2-3 times, they're either going to investigate further or ignore it.
Once you've mapped exposure to your spend, you can then try to match that against what you think your growth rate should be. Put simply, if you had the ability to drive all of your potential market to your application all at once, would you? No, you'd ramp it up so that you can test all of the scaling assumptions on real, live users. So based on your technical load testing so far, you know how large the initial user pool should be. Use that number to back into the exposure level you need (based on the hypothetical conversion rate above) and finally back to how much you should spend. This number should be way below the "spend all the money" scenario. That should be your initial quarter budget for advertising.
The next point is very important: MEASURE. You come into this with the assertion that the revenue generated by converted users is greater than the amount spent on advertising. However, you don't just spend on advertising right? Theres the rest of the business to pay for as well. Use the initial spend to measure your actual yield for dollar spent, and then get your marginal costs for adding each additional user and do a sanity check on the pricing and cost structure. If scaling is in your favor, that marginal number will get smaller as you grow your user base, but in the nasty case where it doesn't, you've got a heads-up before you grow a huge user base that is costing you money to support.
Once that initial phase is done, then you can base your spend rates on measured yields and desired growth rates figured from your technical and strategic plan.