People using economic time series would like them to be available as soon as possible after the end of the reference period. However there can be difficulties in getting all the responses required to produce a series of acceptable quality in a timely manner. The earlier the time series is released the more likely there will be tardy respondents, thus the series will have to be estimated without their responses. As the quarterly gross domestic product (QGDP) is the aggregation of a large number of economic time series the difficulties are compounded. An adequate preliminary estimate of QGDP may be made by using models parsimonious in the number of time series involved. Graphical models assist us in obtaining such parsimonious models by identifying the relevant components in a saturated structural VAR enabling us to eliminate unnecessary delays. Even if an earlier release is not possible we could target work to improve the timeliness of series identified in the parsimonious models.