The official statistics community is reacting to the challenges and opportunities offered by big data. At European level the heads of the national statistical institutes and of Eurostat agreed on a memorandum of understanding to tackle together big data. One of the big data sources available to official statistics is the electronic traces left by users while they use web services. Many of these services provide data based on these traces, either at real time or with very small time lags. As many human activities measured by official statistics are closely related to people’s behaviour online, this data on people’s web activity offers the potential to produce predictions for socio-economic indicators with the purpose to increase the timeliness of the statistics. Many experiments have being carried out recently and there is evidence that these predictions can be made. However, having good prediction models is not enough to produce official statistics figures. If we want to assess the possibility of using this type of big data source then we need to reflect on the transparency, continuity, quality and potential to be integrated with official statistics traditional methods, as well study in more detail the relation between web activity and the phenomena we want to predict.