Personalization is a common technique used in Web search engines to improve the effectiveness of retrieval. While personalizing some queries yields significant improvements in user experience by providing a ranking in line with the user preferences, it fails to improve or even degrades the effectiveness for less ambiguous queries. A potential personalization metric could improve search engines by selectively applying personalization. One such measure, click entropy uses the query history and the clicked documents for the query, which might be sparse for some queries. In this article, the topic entropy measure is improved by integrating the user distribution into the metric, robust to the sparsity problem. Furthermore, a topic model-based ranking for the personalization method is proposed using grouped user profiles. Experiments reveal that the proposed potential prediction method correlates with human query ambiguity judgments and the group profile-based ranking method improves the mean reciprocal rank by 8%.