Recommendations regarding P and K fertilization, as well as critical soil and tissue concentrations for optimal growth of alfalfa (Medicago sativa L.) differ throughout the literature. Cluster analysis is an innovative method for this statistical analysis since inherent plot-scale variation in soil P and K levels can confound analysis of treatment effects using ANOVA and regression. The objectives of this study were to investigate how long-term P and K fertilization affect soil test P and K levels, tissue P and K levels and alfalfa growth, and how this information can be used to predict alfalfa agronomic performance. A factorial combination of four P treatments (0, 25, 50, and 75 kg P per ha) and five K treatments (0, 100, 200, 300, and 400 kg K per ha) was replicated four times. Beginning in 1998, herbage samples were obtained after each forage harvest in May, June, July, and September. Soil samples were collected in April and after each harvest. Tissue and soil samples were analyzed for P and K concentrations. Cluster and variance reduction analysis applied to mass/shoot and shoots/m2 identified 6 clusters with large differences in yield, yield components, and long-term persistence. Results indicate that balanced fertility is important for maximizing yield and persistence of alfalfa. Highest fertilizer application rate did not always result in highest yield. Although low fertilizer rates can result in high yielding stands for the first four years, moderate to high fertilizer rates are required to maintain high yields beyond four production years. Ongoing work will identify cluster attributes such as fertility treatment, herbage P and K, and soil test values that impact performance. The information received will assist us in optimizing P and K fertilization strategies for long-term productivity of alfalfa.