Users form information trails as they browse the web, checkin with a geolocation, rate items, or consume media. A common problem is to predict what a user might do next for the purposes of guidance, recommendation, or prefetching. First-order and higher-order Markov chains have been widely used methods to study such sequences of data. First-order Markov chains are easy to estimate, but lack accuracy when history ma ers. Higher-order Markov chains, in contrast, have too many parameters and su er from over ing the training data. Fi ing these parameters with regularization and smoothing only o ers mild improvements. In this paper we propose the retrospective higher-order Markov process (RHOMP) as a low-parameter model for such sequences. is model is a special case of a higher-order Markov chain where the transitions depend retrospectively on a single history state instead of an arbitrary combination of history states.ere are two immediate computational advantages: the number of parameters is linear in the order of the Markov chain and the model can be t to large state spaces. Furthermore, by providing a speci c structure to the higher-order chain, RHOMPs improve the model accuracy by efciently utilizing history states without risks of over ing the data. We demonstrate how to estimate a RHOMP from data and we demonstrate the e ectiveness of our method on various real application datasets spanning geolocation data, review sequences, and business locations. e RHOMP model uniformly outperforms higher-order Markov chains, Kneser-Ney regularization, and tensor factorizations in terms of prediction accuracy.