1. At least pack in a game worth wanting. Nintendo Land is OK, but it obviously didn't have the appeal of Wii Sports. They should have tried some sort of spiritual sequel to Wii Sports because when people hear that name, they can recognize it. If they hear about some successor to it with integral Gamepad experiences, then the system would have captured some of the Wii owners.
2. The Name. Wii U isn't too bad of a name, but not being clear about whether it's simply an addon for Wii for the common consumer makes them confused. They should have gone with something like Super Wii, or Wii 2, something that makes it sound like an upgrade from Wii. "Wii" is a brand name, but adding a "U" after it won't make people think "successor". What is this "U"?
3. The Specs. After all is set and done, the MCM is obviously one of the more expensive parts of the system, and it's very well designed for what it can do. If the Gamepad (according to Miyamoto) easily adds ~100 US dollars to the price, then the system itself is likely around 200-ish to produce (since it's sold at a loss). Probably 100+ dollars for the CPU and GPU MCM, and that seemed to be wasted on low tdp than high performance. For 33watts, the system is impressive, but I'm sure making it at least 10-15W more would have allowed for higher clocks, and perhaps more overall horsepower. They can keep the Broadway architecture for Espresso if they like (they probably already planned to continue on with the PowerPC line, and I wouldn't be surprised if their next console has something like 3 enhanced Espresso CPUs (9 cores)), but they should have enhanced it further with beefier SIMD, and further increased performance per core. The upgrades they made from Broadway aren't too significant. Making it a quad core wouldn't have been bad either, it's already a very small chip. The Memory of the CPU is fine, 3MB is not a bad thing, it's just 1MB less than PS4 / X1 CPUs, but it's organized very oddly. The GPU should have been more than what it is now, more than 2x the power of Xenos. An HD 4770 in terms of overall horsepower would have easily ran 7th generation games in 1080p with perhaps some graphical enhancements and better IQ. Not that Latte can't run 7th gen games at 1080p natively (I believe it has enough power to do that, but it'll need to be close to maxed out for that, and that's not happening anytime soon), but it should have had enough raw power to brute force 1080p, not reach that until devs finally max out the GPU. Or maybe not even at the 960 GFLOPS a 4770 has, a little less, maybe 700 - 800 GFLOPS (since they would like a low-ish wattage) would have been a bit more respectable, and would make ports from next-gen easier. At least 1 or 2 more GB of RAM would have been fine, DDR3 is very cheap right now, and I'm sure it wouldn't have cost them all that much more of a loss.
A Quad Core beefier Espresso, 700 - 800 GFLOPS Latte, 2 - 3GB RAM for games, 1GB for OS, plus Gamepad, plus at least 64GB Flash memory (flash memory is faster than HDDs, but isn't cheap in large quantities, so not too much!!), 40 - 50 Watts, slightly larger console, plus Nintendo Land, plus several key Nintendo games available for purchase, plus 3rd party games from 7th generation being "definitive" versions for Wii U, and the $349.99 price tag would seem a lot more acceptable. Without the gamepad, selling it for $249.99 would definitely be worth those specs. I think Nintendo wasted too much effort on reaching low wattages, I'm almost positive that they can sell a console similarly to how it's priced now with better specs.
