Allow me to disagree with you, Arlekin, on planning localization close to the release. I've worked on the localization side of things for EA, ZeniMax, Sony, Nintendo, Square Enix and others (from small 3DS games in the Nintendo store to big AAA mmorpgs) and sure it's the one thing with QA that is done last, even if QA is a permanent ongoing process that is more important than localization and needs to start way earlier, but when started too late can end in disaster, if not handled by people who know what they're doing. It can even delay world launches for AAA titles.

Localization needs to be a part of the initial dev roadmap for it to be done at least properly, if not well. Otherwise if it's planned half-assed along the way while the game is in heavy development, like someone coming along in a build validation meeting and saying "hey guys, got a great idea! why not localize everything within the next 3 weeks and test it with robots for mistakes?" (it happened, not these sentences exactly word for word but it happened), then the game will only be thought for the English language, the engine will not be able to display special characters, the text boxes will be static and made to fit the English length, there will be no option to change the language settings, no thought about displaying subtitles, etc etc etc... Usually, localized content starts being integrated as soon as the game is usable. Said content is normally tested ahead outside of the game for translation mistakes, consistency, grammar issues, spelling mistakes and whatnot.

All in all, what I'm saying is that the planning needs to be done way ahead of the doing. It's just the point of view from a localization veteran though.