This is the "year of innovation". (What year isn't?) So it's kind of important to be doing innovation right - to keep the competition at bay, stay on a healthy growth path, or maybe just keep your job. There are two widely recognized approaches. You can start with an idea, leading to ideas-driven innovation (IDI). Or you can start with an understanding of unmet customer needs, leading to needs-driven innovation (NDI). In my opinion, you need to be using both approaches.
The merits of NDI are well established, and many companies claim to do this well. Based upon my own experience in product definition/development/commercialization (good grief, almost 35 years!), I disagree. Most companies do it poorly. It's hard work, and it requires that a very deliberate and precise process be followed for real success. (Real success = Every new-product introduction is a winner + No dogs survive the process.) Our own recipe for a successful Needs-Driven Innovation approach can be found here.
Successful NDI is much harder that most people think. You often need to use multiple research techniques, e.g. in-depth interviews, observation, surveys, focus groups. You need to properly frame your inquiry to get at real user needs. ("How would you change our existing product?" will not work.) You need to devise metrics for levels of need satisfaction. You dare not avoid economics. You need to separate product-feature innovations from the often more compelling business-model innovations (think Netflix). Difficult stuff, but well worth the trouble.
All well and good, but we cannot ignore the importance of ideas-driven innovation. Some of the world's great problem solutions have come about by IDI. At minimum we need fundamental research and scientific breakthroughs to arrive at product definitions that satisfy some of the most important, and most enduring, unmet societal needs. Nevertheless, consider just one of IDI's caveats: You must avoid either immediately falling in love with any idea or dismissing it out of hand before subjecting it to full analysis and potential adjustment.
My suggestion: Use both methods, and do them both right. You'll be proud of your results - and you're more likely to keep your job!