Value-added ‘Progress’ measures are to be introduced for all English schools in 2016 as ‘headline’ measures of school performance. This move comes despite research highlighting high levels of instability in value-added measures and concerns about the omission of contextual variables in the planned measure. This paper studies the impact of disregarding contextual factors, the stability of school scores across time and the consistency of value-added performance for different cohorts within schools at a given point in time. The first two analyses replicate and extend previous studies using current data, confirming concerns about intake biases and showing that both secondary and primary level VA measures exhibit worrying levels of instability. The third analysis goes further by examining whether instability across time is likely to stem from differences between cohorts and whether measures based on a single cohort reflect school performance more generally. Combined, these analyses suggest a general problem of imprecision within value-added estimates and that current policy use of value-added is unjustified. Published school performance measures are likely to be profoundly misleading, in particular for those unfamiliar with the level of uncertainty in the estimates. The paper closes by considering whether value-added measures could and should be used by policy-makers as measures of school performance.