(I think Phil will like this one...)
Back when I was working at Venrock, I had the opportunity to work on a start-up focused on in-car computing. Dash had just come out, in-car GPS units hit the $100 price point and becoming widely adopted, and the partners were excited about the opportunities here and were willing to back it up with some serious funding.
One of the great things of being a part of the incubation group was that I could get involved very early on and help set the (customer-focused) vision for the company. In this particular case, I asked the broad question, "what is the underlying goal people are trying to accomplish when they get in their cars, and how do we help them achieve that goal?"
One scenario I looked at was driving directions. Several people on the team thought this problem was "already solved" and were focused on "differentiating features" like bringing social networking and real-time web into the car. But a close look at driving directions showed that the problem was far from solved.
For one, people today often use a computer to look at a map or get directions, then they have to repeat that task on their device once they get to the car. There clearly should be an easy way to sync the history of maps on the computer with the in-car device.
Another insight: shopping is a common reason to get in the car. Stores like Best Buy and BevMo already allow you to shop online and pick it up at the store. Integrating with store inventories to tell people where items are available and even allowing people to "buy it now" on the way not only helps people optimize their trip, but makes for a great business model.
But the insight I had that I love the most is with driving directions. Every single directions app out there ignores the hidden goal, which is that you first need to park the car before you go to your eventual destination. Yes, in many instances the parking and destination are the same. But in high-density areas (like San Francisco) this isn't so. What people want is a device that will get them near their destination, and then navigate them to a place where they can cheaply and safely park their car, ideally even taking care of any parking fees for you automatically. Once parked, it would then show you directions from your car to your destination also send them to your phone to help you as you walk the final leg of the journey.
Unfortunately, the company decided to go a different direction with their vision. But maybe there is another mobile computing company who will find this information useful. If so, I'd love to hear from you.