Stores selling specialist products such as medical devices need a targeted marketing approach to ensure they are constantly encouraging new clients to purchase their products and also retaining their loyal clients. Gaining an understanding of patterns of customer behaviour is an interesting and profitable venture for specialist companies. Data visualisation provides a useful tool for generating complex and hidden insights which are not obvious through domain knowledge or by simple numerical calculations on data tables. This background motivation was the key for the framing of the research questions for this project, which could provide useful insights into customer behaviour based on customer data provided for an anonymised medical supply store chain in Toronto. The following research questions were investigated: •Over the three years, is there a seasonal trend in customers purchasing products from stores? This can be useful for stores looking to target their marketing campaigns at certain times of the year. • Over the three years, what is the spatial distribution of products purchased by customers in stores across the Toronto metro area? Are there denser clusters in some parts compared to others? • How are the clients being referred to do the store and where are clients who visit through referrals mainly located? An insight into this aspect can also help in increasing sales for the store.
The data for this project was available as csv file downloads as part of the Medical Store Geospatial Challenge from Databits, a community website setup for creating interactive data visualisations ( http://databits.io/challenges/medical-store-geospatial-challenge) These include data over three years on client purchases, referrer sources, location and coordinates of all cities and medical supplies stores. Additional csv files were created through parsing (slicing, merging, reformatting etc.) of these main data sources in Excel to produce it in the most simplistic structured format as required by the workflow stages of ‘acquiring and parsing’ specified by Fry (2008)