Businesses selling clothing and other fashion items face obstacles in trying to forecast how much product to order: tastes and styles change quickly, while manufacturing clothing takes a significant amount of time. Researchers Youran Fu and Marshall Fisher have found that combining sellers’ own data with information gathered from social media can dramatically improve the accuracy of such forecasts—by 24 to 57 percent in the cases they directly studied. Better predictions mean demand is easier to meet without retailers becoming overstocked.