I'm a fan of coming up with a concept, and then building it in the most effective way possible. The concept itself need not be optimized per se. So if you're not playing a half-elf daggermaster with twin strike, you're not necessarily "doing it wrong".
The stats really don't have to line up all that well either. One primary or secondary stat that lines up is often good enough. If none line up, that can be a bit more problematic, but there might be another way to realize the concept. For instance you want to play a halfling warrior from the wild plains who lugs around the biggest sword his people have ever seen a halfling carry, and carves big chunks out of his enemies. Well, sounds like you might want to be a barbarian or great weapon fighter, but a quick examination might show that you would likely do better as an assassin, rogue, or monk.
Saying "My concept is a gnome earth warden with high dexterity and a club and that's what I'm going to play" is not a character concept or roleplaying decision. It is an obstinate decision to be a frustrating party member.
I think 4th edition is designed so we can enjoy party synergy to its fullest in combat. So a character that doesn't play well can be prone to frequent death, or worse, hinder other players' (and DM's) fun factor.
So while optimization to the degree of abuse should not be the goal, optimization within the concept is likely to enhance the quality of time spent during combat. And there is nothing wrong with optimizing outside combat either. A party coordinating to cover every skill, with few overlaps on key skills is always a good idea. A general group focus on a skill such as Stealth can also be beneficial. A character who is more at home in the wilderness, a character who is more at home in the bustle of a city, a character who is more in touch with the deities and magic, and a character who understands the difficulties of lords and feels at home in the presence of nobles can make a nice, well rounded party outside of combat.
A ranger might find himself training in Thievery at 2nd or 4th level, if there is no one else in the party to take on that role. Many optimizers would say this is not an optimal choice, you should pick up greatbow, weapon focus, expertise, lethal hunter, etc. And many "role players" would say Thievery is not in my character's concept, I never stole anything, I won't pick it up. My personal view, figure out a way to make it work for your concept. Perhaps you've had enough of all the gauntlets you had to run through and traps that kept frying you, and during your nature and dungeoneering training, you pay a little extra attention to how to disarm traps and the like without triggering them. In similar vein, if a greatweapon fighter realizes he is the only one in a party without stealth, he might decide to drop plate mail, and train stealth. These are not so much conceptual choices (or even necessarily optimized choices for the sand box), but choices that will enhance a party's efficiency in a given campaign.
I think concept goes beyond any game mechanic. If you see your sorcerer as a quiet wallflower, you don't give him an 8 charisma. The system does not support this. You give him a 16-18 as expected, maybe train in Intimidate, and explain your charisma as a fearsome aura, exuding from him as he walks into the interrogation room and stands in the back corner with flickers of dark energy around him, while the sly halfling does all the talking.
Yet another example I have heard is someone who wants to play a brutish human rogue, who was once a bandit, likes to use axes and wears chainmail. Again, a rogue with handaxe proficiency that never gets sneak attack, is not what the rogue is meant to be. Just pick up maybe two hand axes, and play a tempest fighter with a background that gives you a suitable skill training to be a bandit, and you're set.
Concept != game mechanics. The two live in alternate universes, where the player and DM's job is to link those universes. With this understanding, every character can be optimized to do something useful for the party while mostly remaining within the character concept.