=Paper=
{{Paper
|id=Vol-2323/SKI-Canada-2019-7-3-3
|storemode=property
|title=Using Multi-Criteria Analysis to Estimate Retail Site Attractiveness for the Huff Model
|pdfUrl=https://ceur-ws.org/Vol-2323/SKI-Canada-2019-7-3-3.pdf
|volume=Vol-2323
|authors=Claus Rinner,Renacin Matadeen,Stephen Swales
}}
==Using Multi-Criteria Analysis to Estimate Retail Site Attractiveness for the Huff Model==
https://ceur-ws.org/Vol-2323/SKI-Canada-2019-7-3-3.pdf
Spatial Knowledge and Information Canada, 2019, 7(3), 3
Using Multi-Criteria Analysis to Estimate
Retail Site Attractiveness for the Huff Model
CLAUS RINNER RENACIN MATADEEN STEPHEN SWALES
Department of Geography Department of Geography Department of Geography
Ryerson University Ryerson University Ryerson University
crinner@ryerson.ca renacin.matadeen@ryerson.ca sswales@geography.ryerson.ca
offerings in a transparent economy, location
ABSTRACT plays a crucial role in reducing costs
through reducing travel distance.
The Huff model is an advanced trade area Geographic analysis techniques such as
delineation technique widely used in retail Thiessen polygons, the Huff model, and
site selection. It is based on site customer spotting were developed as
attractiveness and distance, and models normative representations of consumer
probabilities of customers patronizing behaviour (Swales 2008). In a business
specific retail locations. In this research, we context, Thiessen (or Voronoi) polygons
examine the suitability of multi-criteria separate a region with supply points (such
analysis (MCA) to estimate the as retailers or social service centres) into
attractiveness parameter in the Huff model. catchment areas, in which each demand
In a case study of shopping centres in point is assigned to the closest facility. The
Toronto, we illustrate that MCA provides a Huff model allows for more complex gravity
suitable framework for retail site modeling of attractiveness of supply
attractiveness. However, variations in the locations and competition between them by
MCA technique have noticeable impacts on assigning demand to supply points on the
the Huff model results. basis of probability (Swales 2008). Finally,
customer spotting draws on ever growing
1. Introduction affinity program datasets to facilitate
Retail geography, business geomatics, and visualisation, analysis and location
location intelligence are some of the strategies related to patronage of facilities
concepts that capture the idea that the and services (Swales 2008).
location of private sector enterprises has The Huff model (Huff 1964, Huff & Black
significant implications on their success 1997) is widely used in the retail business
with respect to production, sales, and work sector. It combines principles of distance
force (e.g. Jones & Simmons 1993). In a decay with site attractiveness. To model
political climate of accountability and retail site attractiveness for potential
efficiency in using tax revenues, government customers most accurately, analysts often
and not-for-profit organizations use composite metrics (e.g. Lin et al. 2016,
increasingly make use of the same Jia et al. 2017).
operational and strategic decision support Multi-criteria analysis (MCA) is a normative
tools to determine service site locations and modeling approach that integrates decision-
resources offered to their “clients”. maker preferences (e.g. criterion weights) in
The theoretical framework for retail the evaluation of decision alternatives such
geography includes rational choice theory – as retail sites. To support site selection, each
the assumption that consumers will make site is assessed by a composite score created
cost-effective choices with respect to their from multiple criteria (site attributes).
use of retail outlets and services (Swales In this research, we examine the conceptual
2008). Given similar product and service and practical fit of MCA techniques with the
2 Estimating Retail Site Attractiveness
creation of composite attractiveness
metrics. We estimate site attractiveness for
the Huff model using MCA and explore the
impact of MCA parameters on the outcome.
2. Methods and Data
2.1 Techniques and Technology
One of the most commonly used MCA
techniques is the weighted linear
combination (WLC), a weighted sum of
criterion values (Malczewski 2000). To
transform the criteria into a common value
range such as 0...1 for comparison and
combination, a rescaling procedure is
applied (Malczewski & Rinner 2015).
The normative aspect of MCA is most
obvious in the use of importance weights.
These are set by the decision-maker and
may include an element of subjectivity along
with expert knowledge. Weights are
commonly defined as fractions or
percentages that add up to 1.0 or 100%.
The Huff model calculates a quotient of the
attractiveness of a destination site over the
distance to a source point or area. Distances
Fig.1: User interface of the MCA component of the Huff model
may be measured as straight lines or via a plugin for QGIS
transportation network, and the distance
decay effect can be modelled linearly or 2.2 Case Study
exponentially. In retail, the source locations In a case study for the 16 largest indoor
often are residential areas representing shopping centres in the Toronto Census
potential customers. The destination sites Metropolitan Area, we experimented with
are retail locations, the attractiveness of two common multi-criteria rescaling
which is often characterized by a techniques: maximum-score and score-
combination of size and quality indicators range transformation. For criteria that are
such as floor space and type of goods to be maximized, maximum-score
offered. The attractiveness-distance transformation divides each criterion value
quotient is normalized by the sum of all by the largest value. For minimization
quotients to represent the probability of criteria, the quotient of the smallest value
customers from the source to shop at the divided by the criterion value at hand is
destination over all other destinations. For subtracted from 1.0 to form the rescaled
each source, the sum of probabilities is 1.0. criterion value. For maximization criteria,
We implemented the Huff model with MCA- score-range transformation assigns the
based attractiveness estimation in a plugin smallest value to 0.0, the largest to 1.0, and
for the open-source QGIS 3.2 package using all other values in proportion. For
Python 3.6 along with QT5 for the minimization criteria, the assignment is
development of the user interface shown in reversed.
Fig. 1. An earlier version of the underlying The data for the case study included a
scripts is available at https://github.com/ hexagonal tessellation of the Toronto
ryersongeo/qgis_location_analytics. Census Metropolitan Area to represent
source areas (without further attributes)
and a point shapefile of shopping centres
Estimating Retail Site Attractiveness 3
(Fig. 2) with a sample of nine attractiveness composite attractiveness metrics change
indicators sorted from most to least under the score-range procedure.
important:
Average Google Places score
Gross leasable area
Number of anchor stores
Number of fashion stores
Number of technology stores
Number of food stores
Total number of stores
Number of parking spaces
Adjacent housing density
Weights between 25% and 2.5% were
assigned manually for the purpose of the
experiment. All criteria were considered as
to be maximized, i.e. larger values represent
greater utility for potential customers.
Primary and secondary trade areas were
defined by the commonly used probability
thresholds of 60% and 30% (Jones &
Simmons 1993; Swales 2008) and mapped Fig. 3: Trade areas after maximum-score transformation of
attractiveness criteria
for the 16 shopping centres.
Fig. 2: Hexagonal tessellation of Toronto Census Metropolitan
Area and major shopping centre locations
3. Results Fig. 4: Trade areas after score-range transformation of
attractiveness criteria
In comparing the results of maximum-score
transformation (Fig. 3) and score-range This can be explained by the fact that the
transformation (Fig. 4), some differences maximum-score method is “anchored” at
are visually noticeable. In particular, some the maximum value, which is transformed
of the larger trade areas grow further while to 1.0, while the minimum value only gets
some of the smaller trade areas all but down to 0.0 if the original indicator value
disappear under the score-range procedure. also was 0.0; other minimum values are
Since distances do not change, the transformed to rescaled values between 0.0
and 1.0, thus larger minima than under the
4 Estimating Retail Site Attractiveness
score-range procedure, which always uses Acknowledgements
the full value range (Young et al. 2010). The Partial funding from the J.W. McConnell
score-range procedure therefore results in Foundation through the RECODE at
more extreme probabilities and trade areas, Ryerson University program and from
while the maximum-score transformation SSHRC’s Geothink Partnership Grant is
produces more equally distributed gratefully acknowledged.
probabilities (see also Fig. 5).
References
- Huff DL (1964) Defining and Estimating a
Trading Area. Journal of Marketing
23(3): 34-38
- Huff DL, Black WC (1997) The Huff Model
in Retrospect. Applied Geographic
Studies 1(2): 83-93
- Jia P, Wang F, Xierali IM (2017) Using a
Huff-Based Model to Delineate Hospital
Fig. 5: Average probabilities per centre under maximum-score Service Areas. The Professional
(top) and score-range (bottom) transformations Geographer 69(4): 522-530
- Jones K, Simmons J (1993) Location,
4. Conclusion location, location: analyzing the retail
In summary, MCA methods were shown to environment. Scarborough, Ont: Nelson
be applicable to the estimation of site Canada
attractiveness for the Huff model in retail - Lin T, Xia J, Robinson TP, Olaru D, Smith
geography. MCA provides a structured, B, Taplin J, Cao B (2016) Enhanced Huff
framework for the combination of multiple, Model for Estimating Park and Ride
commensurate criteria. This approach (PnR) Catchment Areas in Perth, WA.
integrates decision-maker preferences in a Journal of Transport Geography 54:
way that may be considered normative or 336-348
subjective. - Malczewski J (2000) On the Use of
We are in the process of expanding the Weighted Linear Combination Method in
existing QGIS plugin with additional retail GIS: Common and Best Practice
geography functionality, including other Approaches. Transactions in GIS 4(1): 5-
trade area delineation techniques and 22
alternative distance calculation. Our goal is - Malczewski J, Rinner C (2015)
to create a location analytics toolkit that Multicriteria Decision Analysis in
makes market research accessible to smaller Geographic Information. New York:
companies, non-profits, academic Springer
researchers, and the general public. - Swales S (2008), “Trade Area Analysis” in
It would also be of interest to examine the Swales S (ed.) Marketing Geography
normative modeling framework of MCA in (3rd. ed.), Boston: Pearson Custom
conjunction with crowdsourcing and Publishing
volunteered geographic information, which - Young J, Rinner C, Patychuk D (2010) The
include elements of subjectivity. In addition, Effect of Standardization in Multicriteria
the methods described could benefit from Decision Analysis on Health Policy
integrating open data and using geospatial Outcomes. In G Phillips-Wren, LC Jain,
Web technology in an online modeling and K Nakamatsu, RJ Howlett (eds)
decision support framework. Advances in Intelligent Decision
Technologies (Proceedings of the Second
KES International Symposium IDT
2010). Springer, Berlin/Heidelberg,
Germany, pp. 299-307
Estimating Retail Site Attractiveness 5