[tjic]

Here at SmartFlix we're pretty analytical about it.

When we're batting the idea around, we create a page in the wiki, and the first section is "rationale / analysis". In there we put the number of customers it will attract, or the plausible increase in revenue per customer (with some error bars), then calculate a value for this (a 5% chance of an extra $1 from each of X thousand customers = $Y). We also calculate the cost (we have a nominal price of $75 per engineering hour, and half of that for a design or marketing hour). If a project doesn't pay back its investment in a year, we don't do it. If a project does look to pay back its investment, we then rank order it against other projects to figure out which to do first.