We develop an on-line monitoring procedure to detect a change in a large approximate factor model. Our statistics are based on a well-known property of the (r + 1)-th eigenvalue of the sample covariance matrix of the data (having defined r as the number of common factors): whilst under the null the (r + 1)-th eigenvalue is bounded, under the alternative of a change (either in the loadings, or in the number of factors itself) it becomes spiked. Given that the sample eigenvalue cannot be estimated consistently under the null, we regularise the problem by randomising the test statistic in conjunction with sample conditioning, obtaining a sequence of i.i.d., asymptotically chi-square statistics which are then employed to build the monitoring scheme. Numerical evidence shows that our procedure works very well in finite samples, with a very small probability of false detections and tight detection times in presence of a genuine change-point. Indeed, Stock and Watson (2002) and Bates, Plagborg-Møller, Stock and Watson (2013) argue that, at least in the presence of "small" breaks and a constant number of factors, inference on the factor space is not hampered, thus making the change-point problem less compelling than in other contexts. Nevertheless, stylised facts show that in many applications the assumptions of a negligible break size and a stable number of factors are not, in general, correct. Most importantly, it has been argued that, in presence of a crisis, co-movements become stronger, which may suggest that the economy is driven by a different number of factors than in quieter periods -see e.g. Stock and Watson (2009), Cheng, Liao and Schorfheide (2016) and Li, Todorov, Tauchen and Lin (2017). In such cases, the impact of a change-point is bound to invalidate standard inference and subsequent applications such as forecasting. Recently, the literature has proposed a series of tests for the in-sample detection of breaks in factor structures: examples include the works by Breitung and Eickmeier (2011), Chen, Dolado and Gonzalo (Sequential detection of breaks in (1) is important for at least four reasons. First, the general motivation put forward by Chu, Stinchcombe and White (1996) holds true in the context of factor models also: it is important to verify whether a model, which has been valid thus far, is still capable of adequately approximate the behaviour of new data. Second, the aforementioned (substantial) empirical evidence that factor structures do tend to change over time, especially in presence of a crisis, illustrates the importance of a timely detection of such changes. Third, inference on factor models can be severely marred by the presence of a break (see the comments in Baltagi et al., 2017), which again shows the importance of detecting a break in real time, rather than realising this a posteriori after inference has been carried out and employed, e.g. for the purpose of forecasting. Finally, in the context economics and finance, data are collected and made available automatically, so that the cost of...