## Samenvatting

1. Modified Huff Model

This study proposes a modified Huff model that utilises online data to determine the probability of online patronage across a geographic area and how this can be further utilised to select optimum locations for siting a new offline store for the retailer. The traditional Huff model approach results in distance decay relationship between store attraction and distance to the offline store. However, we use the notion of a growth function in relation to distance to calibrate the model in order to account for the phenomenon whereby as the distance between the customer and the offline store of the retailer increases, the online channel becomes more attractive to the customers; hence probability of patronising the online channel increases. We aim to use publicly available data as input in addition to transaction data from the retailer to predict the probability of patronising the online channel and determine a new site to open a new offline store. The mathematical formulas for calculating the probability of online channel patronage (Pij(online)) is:푷푖푗(표푛푙푖푛푒)= 푨푖푗(표푛푙푖푛푒)푫푖푗훽∑푨푖푗(표푛푙푖푛푒)푫푖푗훽

The variable Aij(online) represents attraction between online customers in the ith grid-cell and the jth store. It has dependency on the following variables: store size (Sj) represented in cubic-metres (m3); expected number of customers (Ni) in the ith location estimated to have made an online purchase; the density of households (1km2) (Hi) with children from the aged of 0-3 years; online store channel (Oj) defined as a binary measure (1= yes and 0 = no); overall coverage of levels of internet connectivity (Ii); and finally, preference minority index (PMi). 푨푖푗(표푛푙푖푛푒)= 푨푖푗(표푛푙푖푛푒)(푺,푵,푯,푶,푰,푷푴)= 푺푗푵푖푯푖푶푗푰푖푷푴푖

The variable Dij represents the distance (km) between online customers from the ith location to the jth store. The value of the parameter β was set at 1.78 based on an optimum solution obtained.

2. Results

Results show that the modified online Huff model yielded comparable results to the actual data in terms of patronage probability and potential location for siting a new offline store outlet, thereby ensuring model robustness.

Both the expected and actual model outputs reveal the same locations as the optimum site for potential new store locations. Further, this enables the application of the model for various purposes especially in scenarios where there is limited data.

This study proposes a modified Huff model that utilises online data to determine the probability of online patronage across a geographic area and how this can be further utilised to select optimum locations for siting a new offline store for the retailer. The traditional Huff model approach results in distance decay relationship between store attraction and distance to the offline store. However, we use the notion of a growth function in relation to distance to calibrate the model in order to account for the phenomenon whereby as the distance between the customer and the offline store of the retailer increases, the online channel becomes more attractive to the customers; hence probability of patronising the online channel increases. We aim to use publicly available data as input in addition to transaction data from the retailer to predict the probability of patronising the online channel and determine a new site to open a new offline store. The mathematical formulas for calculating the probability of online channel patronage (Pij(online)) is:푷푖푗(표푛푙푖푛푒)= 푨푖푗(표푛푙푖푛푒)푫푖푗훽∑푨푖푗(표푛푙푖푛푒)푫푖푗훽

The variable Aij(online) represents attraction between online customers in the ith grid-cell and the jth store. It has dependency on the following variables: store size (Sj) represented in cubic-metres (m3); expected number of customers (Ni) in the ith location estimated to have made an online purchase; the density of households (1km2) (Hi) with children from the aged of 0-3 years; online store channel (Oj) defined as a binary measure (1= yes and 0 = no); overall coverage of levels of internet connectivity (Ii); and finally, preference minority index (PMi). 푨푖푗(표푛푙푖푛푒)= 푨푖푗(표푛푙푖푛푒)(푺,푵,푯,푶,푰,푷푴)= 푺푗푵푖푯푖푶푗푰푖푷푴푖

The variable Dij represents the distance (km) between online customers from the ith location to the jth store. The value of the parameter β was set at 1.78 based on an optimum solution obtained.

2. Results

Results show that the modified online Huff model yielded comparable results to the actual data in terms of patronage probability and potential location for siting a new offline store outlet, thereby ensuring model robustness.

Both the expected and actual model outputs reveal the same locations as the optimum site for potential new store locations. Further, this enables the application of the model for various purposes especially in scenarios where there is limited data.

Originele taal-2 | English |
---|---|

Titel | XXIV EURO Working Group on Location Analysis 2018 |

Uitgeverij | EURO Working Group on Location Analysis |

Status | Accepted/In press - 2 apr 2018 |

Evenement | XXIV EURO Working Group on Location Analysis 2018 - The International Centre for Mathematical Sciences, Edinburgh, United Kingdom Duur: 23 mei 2018 → 25 mei 2018 https://www.maths.ed.ac.uk/ewgla/ |

### Conference

Conference | XXIV EURO Working Group on Location Analysis 2018 |
---|---|

Verkorte titel | EWGLA 2018 |

Land/Regio | United Kingdom |

Stad | Edinburgh |

Periode | 23/05/18 → 25/05/18 |

Internet adres |