Product matching aims at identifying different product offers referring to the same real-world product. Product offers are provided by different merchants and describe products using textual attributes such as offer title and description. String similarity measures therefore play an important role for matching corresponding product offers. In this paper, we propose an adaptive string similarity measure that automatically adjusts the relevance of terms for the product matching. This adaptation is done step-by-step during the match process and does not require training data. We demonstrate that this approach improves the match quality in comparison to the
generic TFIDF string similarity measure.