Why does the author assume normal distribution (bell curve) for apps' revenues?
He doesn't. He tries to estimate the median precisely because he doesn't actually know the distribution although in several places he specifically mentions that it's likely not 'normal'. He's going for the median because it's more informative than the average, given the uncertainty about the distribution. It's a little moot, anyway, since the average figures are pretty lousy and given the most likely distribution, the median just makes it worse.
He doesn't. He tries to estimate the median precisely because he doesn't actually know the distribution although in several places he specifically mentions that it's likely not 'normal'. He's going for the median because it's more informative than the average, given the uncertainty about the distribution. It's a little moot, anyway, since the average figures are pretty lousy and given the most likely distribution, the median just makes it worse.