Sequential pattern mining is an important task for Web usage mining. In this paper we generalize it to the problem of mining context based patterns, where context attributes may be introduced both for describing the complete sequence (e.g. characterizing user profiles) and for each element inside this sequence (describing circumstances for succeeding transactions). Such patterns provide information about circumstances associated with the discovered patterns what is not present in the traditional patterns. Their usefulness is illustrated by an example of analysing e-bank customer behaviour.