For a specified probability model, the Cramér–Rao lower bound sets the smallest variance an unbiased estimator can theoretically achieve. It expresses a limit on precision: an estimator that reaches the bound is efficient. Fisher information links the bound to how strongly the likelihood changes as the unknown parameter changes. section 3.2.2 section 3.2.2
Comparing estimator precision
Two estimators may both converge toward the true parameter while producing estimates with different spread. The one with lower variance is more precise; the bound identifies the theoretical floor for unbiased estimators under the model. section 3.2.2
Maximum likelihood estimators often attain or approach this limit asymptotically as sample size grows. The bound is conditional on the probability model and unbiasedness; asymptotic efficiency describes performance approached with more data. section 3.2.2

