One aspect of pavement management systems (PMS) that has begun to receive increased attention is that of data quality and variability. This paper provides recommendations for quantifying and controlling the variability of longitudinal profile data, and the resulting International Roughness Index (IRI) values. The recommendations were determined from an analysis of the profile data collected by the Long-Term Pavement Performance Program (LTPP). In particular, the effort was focused on developing quantifications of variability that could be useful and economical for network-level pavement management. In the first phase of the variability analysis, the effects of region and season were evaluated using the screened profile data, because the quality of data used in the analysis is crucial to the resulting statistical conclusions. The relationship between the standard deviation and the mean of repeated profile measurements was further modeled using raw profiles; these raw measurements are more relevant to the typical conditions of single-pass network-level data collection. The second phase of the variability analysis was concentrated on quantifying run-to-run variability, such as can be used for frequent checks on control sections and device acceptance, or for project-level data collection. It was found that a run-to-run d2s% varies between 6% and 8% for asphalt concrete (AC) pavements and 6% and 9% for portland cement concrete (PCC) pavements. The final phase of the variability analysis was to develop measures of the variability between routine network-level profile visits to a site, thus enabling a level of cost-effective quality control. While high visit-to-visit variabilities were observed, these values may still be useful for flagging potential data collection problems. The values also indicate the need to use long-term trends for planning, as opposed to single measurements.