Contact centers sit on multitude of conversational data that contains helpful information which can assist businesses to deliver better outcomes like improving customer experience. However, finding such information manually is hard. Towards this end, we propose CauSE, a causal search engine for understanding contact center conversations that assist in finding relevant answers to a question. Using topic modelling, the engine identifies themes within conversational contexts to help reason for the given question. To address the challenge of multiple topics in a single context, we divide the context into Elementary Discourse Units (EDUs) and perform topic modelling on EDUs to better identify coherent themes as topics. Subsequently, we employ a novel contrastive ranking algorithm to surface meaningful topics, and LLM-prompting to obtain descriptions for the topics. Our evaluations of the resultant topics and proof of value exercises demonstrate the strength of the proposed engine.