Customer segmentation for small and medium-sized enterprises with machine learning algorithms

Abstract

The purpose of this thesis was to study how machine learning algorithms could be utilized to improve the cost efficiency of cold calling done by PT-design Oy. The aim was to find out which machine learning algorithm would best suit the segmenting of the customers of PT Design Oy. The theory chapter of the Thesis covers the main elements around a machine learning pro- ject, from marketing to artificial intelligence and a subcategory of it, machine learning. The sources include the literature of the fields, articles, encyclopedias, studies, and reports. This enables a holistic view of the topic and helps readers to understand the relevant information around the topic and to understand the applicable part of the Thesis. The data applied with machine learning algorithms were collected from the database of PT- Design Oy. First, it was analyzed, validated, enriched, and finally cleaned. After the data preparation, it was analyzed again to understand how the preparation might have affected it. Once it was ensured that the preparation had only a minor effect on the data set, applying the machine learning algorithms was started. Four algorithms were applied to the data set, which was compared and evaluated with three different mathematical calculations. As the objective was to try to deliver as good insight as possible to the CEO of PT-Design Oy, the same algorithms were also applied with mathematically sub-optimal parameters to ensure that all insight was yielded from the data set. The objective of improving the cost efficiency of cold calling was not met mainly due to the seasonal changes, the amount of collected data, and the fact that there was no labeled data to be utilized. Also, the cleaning removed a significant share of the data, as the data was in- complete in many ways. Nevertheless, the thesis was able to answer which of the algorithms was most suitable and provide significant improvement suggestions to PT-Design Oy on how to enhance the collection of one of the company's most critical assets, the data.