People say things like "Why do you want to build and maintain two codebases when you can maintain one?" but it's interesting that we're so bad at predicting or even understanding overhead. This applies to other domains — we don't know how much overhead our cars have.

From a project management point of view, native is the least risky. Yes, it takes more people, more salary cost but you know it will work. The last thing I want to find out after six months or an year that our strategy isn't working and we need to rewrite things. Risk is worse than financial cost or time. That is, I'd rather spend a quarter longer or X rupees more if I know I'll get something working at the end of it.