States in the Regional Educational Laboratory (REL) Central region serve a largely rural population with many states enrolling fewer than 350,000 students. A common challenge identified among REL Central educators is identifying appropriate methods for analyzing data with small samples of students. In particular, members of the REL Central Educator Effectiveness Research Alliance in Colorado, Kansas, South Dakota and Wyoming are interested in understanding how the precision of student growth percentiles (SGPs), a measure of student growth in their accountability systems, varies depending on sample sizes. To support the EERA members, this study investigates the precision of SGP estimates when SGP calculations are based on small sample sizes. In small samples, very few students have exactly the same prior achievement score. In order to increase the sample size at any given level of prior achievement, some states with small student populations have considered using a coarser measure of prior achievement, such as dividing students into four achievement levels, instead of using exact achievement scores. This study investigates how categorizing students coarsely by prior achievement level before SGP analysis affects precision. Findings suggest that SGP estimates are less precise for high- and low-achieving students than for students with average achievement when the total sample size is small. Moreover, categorizing students coarsely by prior achievement before estimating the SGP model results in an increase in the precision of SGP estimates for the highest and lowest achieving students; however, this technique also reduces the similarity of students whose growth is compared. Results for different sample sizes may help states plan their strategy for SGP implementation and communication with stakeholders, such as reporting SGP bands instead of single numbers or cautioning stakeholders about making comparisons between SGPs that are similar. Data and Methodology are appended.