BLS’s estimates of quarterly labor productivity, output per hour worked, are revised because of revisions to source data. Early estimates of hours worked and output are subject to substantial revisions for a variety of reasons. The BLS productivity program produces three regularly scheduled estimates of labor productivity growth: the preliminary estimate, the first revised estimate, and the second revised estimate. We consider revisions to the preliminary and first revised estimate relative to the second revised estimate. Our goal is to develop intervals to help data users better assess the size of these revisions. Most of the revisions result from regularly scheduled updates of source data. We analyze these revisions to get a better understanding of their sources and to determine whether there are any systematic patterns that could be exploited to construct intervals. We find no evidence of trends or systematic patterns that we could exploit. Most notably, the largest revisions to current and prior quarter output coincide with the BEA’s annual revision to GDP. We then consider three alternative methodologies for constructing intervals: modified confidence intervals, model-based intervals, and percentile-based intervals. We argue that the percentile based intervals are preferable, because they are less sensitive to outliers and therefore result in narrower intervals for a given level of statistical confidence.