Prairie restorations have been implemented using a variety of seeding methods and techniques. The outcomes of these prairie restorations have not been surveyed on United States Fish and Wildlife Service lands in eastern North Dakota and western Minnesota. A survey was initiated to determine the outcomes of these restorations and to provide guidance on what restoration methods and techniques produce desired results. A total of123 sites were sampled across the area in various upland positions. Data collection took place during June, July, and August in 2009 and 2010. Inforn1ation collected from the field and from restoration plans included plant community data, physical data, seed mixtures, planting methods, planting age, and invasive/undesirable species information. Plant composition sampling was conducted using ocular estimation of plant cover percent (%) within three randomly placed 2 m2 quadrats placed in a triangular fashion 12 meters apart. A non-metric multi-dimensional scaling analysis was utilized, featuring correlations of restoration characteristics based on groupings from cluster analysis and multi-response permutation procedures. Logistic regressions were also performed to determine probabilities of membership to certain groups and Akaike's Information Criterion (AIC)was used to compare among factors and models. From these analyses it was found that prairie restorations in the study area can be placed into three significantly different groups (p<0.05). One group (Group 1) consisted of younger restorations that had high variability, which are likely to diverge into one of two other groups. One of the other two groups (Group 2) consists of older restorations, lacking in diversity and high in undesirable grass species. The last group (Group 3) had a variable age, moderate to high diversity, and low invasive/undesirable species which tend to be those most desired characteristics for restorations. A high probability of membership to Group 3, greater than 8 out of ten restorations, occurred when: 1) a minimum of 9 grass species was seeded, 2) ten forb species were included in the seed mix, and 3) broadcast seeding was utilized. Dormant season was the most dependable planting season for membership in Group 3, with winter (10/21 - 4/ 14) being a hundred percent predictor. The results of this survey will guide restoration practitioners as to the probability of their planned restorations developing into the two dominant groups and the characteristics of restorations that have a high probability of meeting desired restoration conditions. Knowledge of these probabilities will assist managers in developing efficient and self-sustainable prairie restorations and can help in the planning of conservation under increasingly high costs and constraints on management.