Calculating short-term forecasts in real time is challenging. First, analysts performing multi-period forecasts must choose whether to use a recursive model based upon one-step-ahead forecasts, or a multi-period model that is estimated with a loss function tailored to the forecast horizon. Although the recursive method provides more efficient parameter estimates than the direct method and does not require different models for different forecasting horizons, it is prone to distortion if the one-step-ahead model is incorrectly specified and usually requires separate forecasting models for the explanatory variables. Which approach is better depends on the properties of the forecast model and will ultimately be an empirical matter.Additionally, the execution of short-term forecasts in real time faces specific problems in the daily monitoring of economic developments. The first problem is that the real-time data flow of all variables involved in prediction analysis does not occur simultaneously. Although national statistical institutes publish economic data in blocks and publications follow a relatively stable calendar, most publications are asynchronous. In addition, the economic indicators of the current economic situation are available with varying delays. Typically, hard indicators referring to economic data typically show relatively long reporting delays, usually two months. Soft indicators based on opinion polls are published promptly as they are usually available at the end of the month under review. Financial data are available daily and are also available at the end of the month. In accordance with this special release procedure, the automatic prediction models should support the reading of linked data in an asynchronous manner.Failure to take this publication pattern into account would mean that users of traditional forecasting models that develop forecasts from balanced data sheets would inevitably incur one of the following two significant costs. The first one appears when predictions are made from the last available balanced panel. In this case, the forecasts lose the latest and most valuable information contained in the timely indicators at the time of the assessments. The second cost is the delay. When the analysts decide to wait until all indicators of the economic cycle are available, their conclusions refer to the past.The second problem with real-time short-term forecasting is that it is usually based upon time series data sampled at different frequencies. Many key macroeconomic indicators, which are the most important time series to be predicted, are queried with low frequency and published with considerable delay. A noteworthy example is GDP, whose data are requested quarterly and published with a delay of about one and a half months compared to the end of the reference quarter. In order to obtain early estimates of these low frequency variables, analysts often focus on higher frequency economic indicators. These indicators typically correlate well with low frequency variables, but show much more timely information at monthly, weekly, daily or even higher frequencies.