Residential Status and Consumer Behavior in Helsinki

This map image is part of a larger series of study maps created to give an overview of the basic structure of grocery shops and to visualize outcomes of the different models used to analyze the interaction between shops and customers in the Helsinki metropolitan area.

Defining differentiated residential areas or area profiles finds distinctive customer segments for marketing and adjusts the supply to correspond to relevant demand. Target marketing can achieve better cost-efficiency, and the retail trade can increase efficiency in logistics and space management. This study sought to define differentiated residential areas indicated by a grid based on multivariate analysis and to identify the retail system by turnover and floor area. The interaction between shops and customers is analyzed by a multilinear model, which tries to give an overall picture of the performance of the retail system giving an individual attractiveness factor for supply units and a friction factor for customers.

A probabilistic model describes the behavior of the customers. The collection of all factor values defines probabilities for customers to visit any of the shops, and the model can be customized so that the likelihood of the observed pattern of shop visits is maximized by these probability values. The present model is fitted to a complete georeferenced customer--shop matrix, connecting all customers to all shops covered in the study. No functional form of the dependence of shopping probability on distance is postulated. Instead, the dependence is obtained from the model in numerical form. Any influence by a preconception of the researchers is kept at a minimum, and the data itself is allowed to determine the form of the distance law. Separate linear regression models have tested the resulting parameters. It showed that the shop-dependent parameter could accurately predict the turnover of a shop without any other information. If a new independent variable, the floor area, is added, the degree of explanation perked up from 61 to 82 percent.