We propose a new way to perform adaptive information system design An abstract how an appropriate actual information about how an appropriate actual information structure should be whenever a query is processed the abstract information structure is modified to reflect the effect of the processed query after many queries are processed the abstract structure is examined if it reveals that the actual information structure is inappropriate then the actual structure will be modified; otherwise no change is made. this one-query-at-a-time approach is applied to two problems namely the record clustering problem. results demonstrate that our approach yields good performance is conceptually very simple is intuitively natural and has a reasonable running time reasons for using our approach versus previous methods are also sketched.