Tourism trips are increasingly planned and organised online. This generates some digital traces correlated with the tourist movement, and thus potentially useful to improve the accuracy and timeliness of forecasts. This hypothesis is based on the fact that before booking a trip, travellers and tourists look for information about the destination on the Internet. Therefore, we expect a significant relation between the online popularity of a destination and the real, physical visits it receives. It is therefore possible to take advantage of these digital traces with the aim to improve tourism forecasting. Previous studies have shown the capacity of online data to improve the forecasts of tourism-related variables in different regions. Assessing the potential and applicability of online behaviour data sources to support the production of official statistics is a new line in which statistical offices are working. Eurostat is performing some pilot studies on different big data sources applied to different fields, including tourism. The study we present in this paper is developed as a pilot to assess to which extent online sources, which are providers of big data, can help to predict real tourist movements. The aims of this on-going study are two-fold: Firstly, to check if online data from Google Trends and Wikipedia pageviews can help to improve the accuracy of forecasting models for tourist arrivals. Secondly, to compare the two online sources in order to assess not only which one performs better, but also to check if they are complements or, on the contrary, they are substitutes. This is fundamental for official statistics offices to decide which online sources are worth pursuing further investigation and introduction in official statistics production, e.g. to develop flash estimates and forecasting models based on these new sources of massive, timely and granular data.