<p>In this work, we present a methodology that measures the customer service expertise and performance in video by applying several rules and metrics on knowledge graphs (KGs). In our approach, the KGs represent human behavior performed in the video through conversations and actions which are described from the knowledge base (KB). The definition of rules, baselines, and metrics are written by specific notations (Allen's for representing time interval relations and RCC8 for defining location relations). The methodology is composed of four stages: in 1) "behavior pattern definition" the rules, baselines, and metrics/scores are defined for assessment of behaviors in customer service. The 2) "knowledge graph constructions" process video files, extracts, and represent human activities and interactions with objects in video. During the 3) "knowledge graph retrieval" the user behavior is retrieved from a knowledge graph by means of SPARQL queries. Finally, in 4) "knowledge graph analysis" the rules and scores are applied. In order to measure the expertise following certain rules, the methodology implements inferences, queries, filters, and temporal processing on the knowledge graphs. The purpose of this step is to measure expertise in customer service. Consecutively, the user performance in the video is compared with other baselines (user expert and average). As a case study, the work was applied to elderly care customer service using public videos from the elderly behavior library.</p>