It is good to start with the truth: currently making money selling desktop software is really hard and, in general, doesn't payoff the amount of work you need to be there. We have a lot of examples of companies doing herculian work beyond Sublime Text and Atom which don't have enough revenue. Part of this is related to the web and mobile trends but also the desktop software maturity. In Microsoft Windows you can find a lot of tools but very few assume that they need to pay for them, they can be very popular (look at the most downloaded desktop software) but the profits/users has a very low ratio. In general, the successful ones rely on a mature sales force or they are unicorns. I think it is a good start studying the current desktop ecosystem to see how hard it is.

Notwithstanding that, I recommend to think about a services/product business model more than in selling "millions of units" only. Although a little old now, I recommend Michael Cusumano's work on this [1]. My small company uses (complex) desktop and server software products as a marketing tool for services and support.

Now the question is what is the minimum profit you will accept to make. Is it $ 100k / year, $ 200k or $ 1mm ? From that picture you can deconstruct how many units of your app/service/support you need to sell to be there. As a thumb rule I will say that selling few units for a higher price will be much easier than selling a lot of units for a few dollars. If you want to try, I would recommend start selling to SMB companies who have a budget for this.

Regarding finding ideas, talk to people who have real software problems in their companies.

[1] http://ebusiness.mit.edu/research/papers/197_Cusumano_ProdSr...