A method of extracting essence from data on an event sequence is developed. This study aims at highlighting the hidden structure of influence from/to segments in the given data, rather than reducing the data quantity without reducing the quality of data mining results. Here, the influence from each item in a segment to the overall sequence, and the influence from the sequence to each item in a segment are computed. Then, from each segment, items of the highest influence from and to the sequence are extracted as key-items. The key items are collected and listed in the order of segments, to form the squashed data. Data mining and visualizing tools were applied to the obtai ned squashed data, and the present method is found to improve the understandability and the novelty of scenarios to be discovered from the data.