In 2015 the pre-election polls showed their worst performance since 1992. In this year’s lecture Professor Patrick Sturgis of the University of Southampton, who is leading the BPC/MRS Inquiry into the failure of the polls, takes stock of the wider implications of the polling miss for survey research. Roger Jowell used to say that pre-election polls are the only survey estimates to be regularly and rather brutally measured against reality and many researchers are likely to have felt a degree of sympathy for the pollsters on May 8th. As the most visible and high-profile part of the survey research industry, the failure of the polls is likely to adversely affect public confidence in social and market research more generally. Following the failure of the polls in 1992, several commentators advocated a shift from quota to random probability designs. However, because even the best-designed and executed probability surveys now routinely achieve low (and declining) response rates, others question whether random surveys are worth the considerable financial outlay required to implement them. Have we reached a critical stage, or even a crisis, for traditional survey methods? Is it the time to consider radical alternatives? Or is the focus on headline response rates an essentially superstitious activity, distracting us from a real understanding of bias in relation to survey findings?