New inferential procedures for the geometric distribution, based on δ-records, are developed. Maximum likelihood and Bayesian approaches for parameter estimation and prediction of future records are considered. The performance of the estimators is compared with those based solely on...