Financial data analysis is becoming more vital as the data collected from daily operations are exponentially increasing in the presence of data mining. Many financial analysts hope to extract useful knowledge from the database in the decision-making process to achieve a competitive priority. In business, companies' key performance indicators are usually characterised in a high dimensional space and companies can be categorised using them. Many literatures proposed different approaches to classify companies using different dataset. Although a plethora of multivariate analysis has been available, the computational requirements with highly complex datasets are challenging the current clustering algorithms. This paper proposes a new strategy by integrating dimensional compression techniques with clustering. The former projects the data in a few major variability dimensions; the latter further clusters the projection into groups. An application case study is provided for illustration and verification. This procedure shows a promising potential in wide variety of business applications.