US graduate programmes are starting to formalize expectations for the skills and competencies that PhD students should have by the end of their studies, finds a report from the US Council of Graduate Schools (CGS) in Washington DC. In a 2016 survey of its 241 member institutions, the CGS found that 65% of those responding reported that all or most of their doctoral programmes had developed formal ways to assess whether students are learning specific skills that are relevant to the workplace. The US academic community has long been considering how to address the fact that holders of science PhDs typically have not learned what they need for non-academic careers (see Nature 543, 277; 2017). Employers outside academia want candidates with transferable skills, including experience in data science and big data; science policy; governance, risk and compliance; and time, project and budget management. The report recommends that universities work with employers to find out what they look for in job candidates. Universities in Australia, Canada and Europe have developed similar graduate-programme assessment metrics.