Boxing customers into age and income classifications to predict their purchase behavior is an imperfect science, to say the very least. But in India, even more so, given the frustrating diversity that retailers encounter in tastes within a demographic, locational or income category. At Flipkart, a team of data scientists has been trying to make patterns out of the chaotic jumble of data points dropping into their system every day. The result: a shelf full of labels that flip in and out of screens, based on purchase behaviour rather than geography or income. Or in terms of the prevalent terminology,