With the increasing popularity of Machine Learning (ML) in many application fields such as health care, environmental protection and criminal investigation, more and more application analysts are keen on incorporating ML to their existing analytical approaches in order to improve the efficiency and effectiveness [1]. Given an application problem, often a tailor-made ML software product is developed to help addressing the existing challenge in data processing and knowledge extracting. Due to the fact that most of the application analysts are not ML experts, the challenge is to design the tool in such a way that the analyst can easily understand the underline principle behind the algorithms and feel comfortable to use it. In this abstract we introduce our journey through the design of a visual analytics tool as part of the EU-funded project “Visual Analytics for Sense-making and Criminal Intelligence Analysis (VALCRI)”, with the aim of making ML more approachable by domain analysts.