This document provides a summary of a study analyzing the spatial distribution of high street retail stores selling mobile phones in London. It uses geographic information systems and spatial analysis methods to understand the business context, examine networks of stores and customers, and identify key conclusions. Maps show the density and clustering of stores across London boroughs alongside demographic and deprivation indices to understand the consumer base. Network analysis of streets and catchment areas around stores is also used to evaluate accessibility and coverage.
1. High Street Retail of
Mobile Phone
Suppliers in London
Spatial market analysis
and visualisation
December, 2019
UCL INNOVATION & ENTERPRISE
2. Contents
1. Study Framework
2. Geographic Business Context
3. Network Analysis
4. Geodemography of Consumers
5. Key Conclusions and Recommendations
3. Study Framework
Challenges
▪ Understand the geographic business context
▪ Provide insights into the current marketplace
▪ Support formulation of marketplace strategy with granular insights on the intersection
of trade areas and customers
Deliverables
• Support potential market(s) for expansion based in total addressable population and on
geo-demographic insights
• Targeting geomarketing campaigns by geographical locations
4. Methodological Approach
I. Retail morphology analysis from
competitors in the local market
with spatial extent, population
and neighbourhood typologies
II. Consumer demographics :
customer profile mapping.
Identify, describe and segment
groups of customers
III. Network analysis for human
mobility and store area coverage
IV. Geographic segmentation:
catchment estimates for location-
based targeting strategies
1. Research
I. Optimization of dealer network:
insights for a better fit of physical
store location from which to
serve and retain customers
II. Select potential catchment areas
for business expansion
III. Targeting geomarketing
campaigns
2. Location Planning Strategy
Drive strategic
decision making
and business
outcomes
5. The Spatial Scale
In this false-colour satellite image taken on 20 September
2019, wavelengths combine to make dense vegetation
appear green.
The ring of woodland and open space encircling great
London is referred as a ‘green belt’, a restraining device
set up in 1935 to keep the city’ s buildings from sprawling
into the countryside.
According to the Office for National Statistics (ONS),
London has a population of 8.9 million in 2019. The next
census in 2021, population is projected to reach 9 million.
1
4
5
2
3
5
Slough
4 Heathrow Airport
3 Richmond Park
2 City of London
1 Thames Estuary
Note:
false colour composite with spectral band combination 1-4-7.
Satellite imagery source:
U.S. Geological Survey (USGS), Earth Resources Observation and
Science (EROS) Center (2019). LANDSAT TM Collection 1 - Path: 201
Row: 24 for Scene: LC08_L1TP_201024_20190920_20190926_01_T1.
Remote-Sensing Image. Sioux Falls, SD: U.S. Geological Survey (USGS)
Earth Resources Observation and Science (EROS) Center.
6. In this false-colour satellite image wavelengths are
combined to make vegetation appear red.
Standing on the River Thames in the south-east of
England, at the head of its 50-mile (80 km) estuary leading
to the North Sea, few cities can match London’ s ethnic
diversity. According to the 2012 Census, fewer than 45% of
Londoners consider themselves ‘White British’.
London is surprisingly vast: the Isle of Dogs could hold 500
football pitches on it and the entire boundary of greater
London 220.000 pitches.
Isle of Dogs
Waterloo
Bridge Tower Bridge
4 km
Note:
false colour composite with spectral band combination 1-1-7.
Satellite imagery source:
U.S. Geological Survey (USGS), Earth Resources Observation and
Science (EROS) Center (2019). LANDSAT TM Collection 1 - Path: 201
Row: 24 for Scene: LC08_L1TP_201024_20190920_20190926_01_T1.
Remote-Sensing Image. Sioux Falls, SD: U.S. Geological Survey (USGS)
Earth Resources Observation and Science (EROS) Center.
The Spatial Scale
7. The Spatial Extent
In this study, the spatial extent comprises the
area of Greater London Authority (GLA). It
spans thirty-two boroughs plus the City of
London.
The scale is based on catchment areas
typologies employed by official institutes. The
main catchment areas are the Middle Layer
Super Output Area (MSOA), Lower Layer
Super Output Area (LSOA) and London' s
administrative units, the boroughs.
All data are officially defined and provided by
the Greater London Authority and the Office for
National Statistics of the United Kingdom
(ONS).
The London congestion charge is a fee
charged on most motor vehicles operating
within the Congestion Charge Zone (CCZ) in
Central London between 07:00 and 18:00
Mondays to Fridays.
Source:
Mayor of London in https://data.london.gov.uk/
8. Retail of Mobile Suppliers: selected players
This study aims to analyse the street-side business of mobile telephone. Networks providers
of mobile connections have also been considered. Most of the companies also operates on
the web with online stores.
To analyse the retail morphology of the industry an inventory of stores that sell mobile
phones and suppliers has been set-up and uploaded. Three categories are classified:
1. Mobile network providers
2. Third party retail
3. Mobile brands direct selling
In total 441 points of interest have been uploaded from 12 different companies.
Note: data from July 2019
Third party retail Mobile brands direct selling Mobile networks
9. Contents
1. Study Framework
2. Geographic Business Context
3. Network Analysis
4. Geodemography of Consumers
5. Key Conclusions and Recommendations
10. Spatial Pattern of Physical Stores
Spatial point pattern analysis is concerned with
analysing patterns in the spatial distribution of
points. Are the stores randomly located? Do
stores tend to cluster together? Are these
areas of high or low density?
A content generalisation (classification) is
applied to emphasise similarities among the
stores with special symbol for each store
category.
The stores are in general clustered together
and concentrated within the Congestion
Charge Zone.
It is difficult to outline any insight from this map
since there are too many data points
presented. A further refinement may increase
visualisation with different levels of abstraction
and generalisation.
11. Spatial Pattern of Physical Stores
In this map, the stores are combined into
three categories maintaining the
representative pattern in their distribution.
Central London and the area inside the
London Congestion Charge Zone have the
highest concentration of stores compared to
other zones.
Third party stores are more distributed in the
city. This can be explained by the high prices
to rent a property in the central area of
London.
To identify patterns via data visualization it is
important to use generalisation functions.
This imply some sort of simplification to
reduce the complexity of a group of features
by portraying them with a smaller number of
features.
Note: Basemap by Ordnance Survey
12. Density of Physical Stores
This map helps to visualise the density of
dots through the use of a continuous colour
gradient highlighting areas with most point
features.
The density estimation is presented by
hotspots of areas with equal density of stores.
On this map, the stores are not classified but
plotted on the study area.
We can point out a higher concentration of
stores in central, west and southwest London.
East London presents the lowest
concentration of stores.
13. Proximity Analysis
In this graphic, a line has been drawn hallway
between every store and its next nearest to
show how far Londoners have to travel for their
closet shop. What emerges is a stained-glass
window of consumer choice.
Though people don’ t always use their closest
store, they won’ t go too far out of their way to
visit a competitor. This distance varies with
transport mode. Relatively few people have cars
in more central areas. They shop locally and
often avoid carrying too much.
Each Thiessen polygon contains only a single
point input feature. Any location within a
Thiessen polygon is closer to its associated
point than to any other point input feature.
14. Spatial Clustering
This map introduces measures of clustering
in areal data, a method to analyse the spatial
heterogeneity of distribution of stores across
London.
Cluster analysis uses unsupervised machine
learning clustering algorithms which
automatically detect patterns based purely on
spatial location and the distance to a specified
number of neighbours.
It aims to quantify the autocorrelation between
each spatial unit and their neighbours. It
identifies spatial clusters of features with high
or low values.
The stores are in general clustered together
and concentrated within central London. By
means of a grid-cell, we can spot low
concentration of stores in east London and in
many parts of southern London.
Note: Density-based clustering applied.
15. Contents
1. Study Framework
2. Geographic Business Context
3. Network Analysis
4. Geodemography of Consumers
5. Key Conclusions and Recommendations
16. Street Network Analysis
Mobility analysis provides emerging
applications in marketing and geo-spatial
services.
To study the spatial allocation of stores, we
have to understand spatial coverage in a way
that supplies customers most efficiently. Also
the topology of streets provides important
insights into the micro-mobility at individual-
level.
Network-based spatial analysis of streets
contains not only the location and attributes of
roads, but also information about how roads
relate to one another and other information
that affects travel paths.
We will apply techniques of street network
analysis to demonstrate how insights can be
drawn from street network data regarding
store allocation and pedestrian circulation.
17. The Spatial Configuration of Street Networks
Road network is among the most recognisable and studied types of network.
High levels of street connectivity can attract retailers and commercial activities which are pedestrian
attractors and generators.
More retailers and commercial uses appear in the more integrated parts of the city.
Therefore, attractors may reinforce natural movement, acting as multipliers.
Different network types can be specified to clarify what is an edge in the network:
• Car: get drivable public streets (but not service roads)
• Bicycle: get all streets and paths that cyclists can use
• Pedestrian: get all streets and paths that pedestrians can use (this network type ignores one-
way directionality by always connecting adjacent nodes with reciprocal directed edges)
Directedness does not matter for pedestrian analysis as for drivable networks.
Car Bicycle Pedestrian
The street segments can be translated into a graph, in which
segments and their connections turn into nodes and links.
The basic element of topological representation
is the street segment between intersections (i.e. edges) and
the vertices, also known as nodes.
Source: Mirco Musolesi and Stephen Law.
18. Network Connectiveness
To analyse and predict pedestrian movement, a street network with
a buffer distance of 1500m from a central location has been created.
This street network represents the traveling distance along the
network from Trafalgar Square located in central London. The
network draws the streets and corners (i.e. nodes) in the spatial area
available for pedestrians.
Street network connectivity is among the significant factors affecting
pedestrian volume. Connectiveness can be studied by metrics of the
urban grid and street physical connectivity and indicates the
importance of streets and corners in a street network area.
▪ Closeness centrality: is the average length of the shortest path
between the node (represented by a street corner) and all other
nodes in the graph. The more central a node is, the closer it is to
all other nodes.
▪ Degree centrality: is defined as the number of streets
connected to a node (or street corner).
▪ Betweenness centrality: measures the number of times a node
acts as a bridge along the shortest path between any pair of
nodes.
19. Street Centrality
These maps represent the street network in Central
London with Trafalgar Square as a central point
location. The stores are represented by the overlay
of pinpoints.
Closeness Centrality Betweenness CentralityDegree Centrality
▪ Closeness centrality: is the average length of the shortest path between
the node (represented by a street corner) and all other nodes in the graph. The
more central a node is, the closer it is to all other nodes.
▪ Degree centrality: is defined as the number of streets connected to a node
(or street corner).
▪ Betweenness centrality: measures the number of times a node acts as a
bridge along the shortest path between any pair of nodes.
20. Node Centrality
Closeness Centrality Betweenness CentralityDegree Centrality
These maps represent the node network in Central
London with Trafalgar Square as a central point
location. The stores are represented by the overlay
of pinpoints.
▪ Closeness centrality: is the average length of the shortest path between
the node (represented by a street corner) and all other nodes in the graph. The
more central a node is, the closer it is to all other nodes.
▪ Degree centrality: is defined as the number of streets connected to a node
(or street corner).
▪ Betweenness centrality: measures the number of times a node acts as a
bridge along the shortest path between any pair of nodes.
21. Network Service Area Coverage
A network service area is a region that
encompasses all accessible streets within a
certain walkable catchment.
This map introduces the service area
coverage of the stores located in Central
London with catchments of 5 and 10-minutes
walking distance from each store.
Examining accessibility can help identify what
is near existing stores and give important
insights to allocation decisions.
Evaluating accessibility also helps with
geodemographic questions as the number of
people living within a catchment distance from
a store.
22. Contents
1. Study Framework
2. Geographic Business Context
3. Network Analysis
4. Geodemography of Consumers
5. Key Conclusions and Recommendations
23. 2. Super Profile
Psychographic
classification (lifestyle)
1. Core Classification
Methodological Approach
Geodemographics analysis is done with the profile of consumers based on
standard geographic areas, customer addresses or site locations. By combining
internal and external data, geodemographic characteristics are profiled, and areas
of opportunities uncovered.
Geodemographic classifications organise areas into categories sharing similarities
across multiple socioeconomic attributes.
All data are officially defined and provided or derived by official institutes:
1. Office for National Statistics of the United Kingdom (ONS)
2. Consumer Data Research Centre (CDRC)
3. GLA Intelligence and Analysis Unit
Univariate composite index that takes different input variables and reduces them
into a single has also been applied in this research.
For a better understanding of the application of geodemographic data to geospatial
analysis, a hypothetical profile matching has been drawn for some statistics.
These examples must be adapted for different requirement specifications.
Census demographic
variables
1. Core Classification
- Census demographic variables
- Social, economic and demographic features
24. Retail Location and Population Density
This composite map presents the
variable population density per LSOA
(lower Super Output Area) calculated
by the ONS.
The index of population density
combines the most recently available
small-area population density available
(ONS/NOMIS, 2011).
The map draws attention to the high
concentration of stores in areas more
densely populated.
The stores ate also located in areas
with concentration of work.
Source:
Mayor of London in https://data.london.gov.uk/
25. Retail Location and Indices of Deprivation
Today, London uses an Index of Multiple
Deprivation to identify poorer areas in the
capital.
This map combines two variables on the
same analysis: 1. the number of stores per
borough and 2. the composite index of
multiple deprivation (IMD)
IMD is a composite index composed of the
following domains: 1. employment; 2.
education, skills and training; 3. health and
disability; 4.crime; 5. barriers to housing and
services and 6. living environment. The
inputs are standardized and then combined
to create an overall score. The index depicts
the most (red) and least deprived (blue)
areas of London.
Even though the stores are not evenly
distributed across London, there is no clear
correlation between store location and
indices of deprivation.
Areas in northeast London present a low
concentration of stores.
Source:
Mayor of London in https://data.london.gov.uk/
26. Internet Use and Engagement Index
Example of attributes
Group Name Profile matching
1 Youthful Urban Fringe +
2 e-Cultural Creators
3 e-Professionals
4 e-Mainstream
5 e-Veterans
6 e-Rational Utilitarians
7 e-Withdrawn
8 Digital Seniors
9 Passive and Uncommitted Users
10 Settled Offline Communities -
The 2018 Internet User Classification (IUC) is a bespoke classification that describes how people
living in different parts of Great Britain interact with the Internet. Engagement with the Internet has
an obvious impact on consumer behaviour, such as the use of brick-and-mortar retail relative to
online shopping.
Online behaviour attributes include direct
measures such as annual transactions per
person from online retailers and others from
the British Psychological Society (BPS)
survey. For the creation of the IUC,
conventional regression-based statistical
model and machine learning have been
applied.
Source: Consumer Data Research Centre (CDRC).
https://data.cdrc.ac.uk/dataset/internet-user-classification-
2018
Example of profile matching
27. Internet Use and Engagement Index
Geographically, retail stores are mainly located
close to the city centre.
Mapping overlay with IUC shows that stores
are mainly concentrated in areas with the
presence of users with high levels of internet
engagement on social networks and good
levels of online shopping.
These areas comprises fairly young
populations of urban professionals, typically
aged between 25 and 34.
The rising smartphone penetration, particularly
amongst those aged 55 and over, provides
opportunities in higher-margin product
segments.
28. London Output Area Classification (LOAC)
The London Output Area Classification (LOAC) is a bespoke geodemographic for London. The LOAC
uses a combination of over 60 Census variables to classify every single small area in London within a
hierarchical structure.
The classification is built at the level of the Output Area / Small Area and is created entirely from 2011
Census data.
LOAC attributes include age groups, levels of unemployment, occupations and levels of education,
ethnicity and housing patterns. The clusters making up the typology are formed by 8 super-groups
and 19 groups.
• Lives in a communal establishment
• Mixed/multiple ethnic group
• Main language is not English and cannot
speak English well or at all
• Full-time student household
• Schoolchildren and full-time students: Age
16 and over
• Use of private transport
• Two or more cars or vans in household
• Age group
Source: Census Information Scheme – GLA Intelligence
https://data.london.gov.uk/dataset/london-area-classification
Example of variables
Super group name Profile matching
1 High Density and High-Rise Flats Concentrations +
2 Multi-Ethnic Suburbs
3 Settled Asians
4 Intermediate Lifestyles
5 Ageing City Fringe
6 Urban Elites
7 City Vibe
8 London Life-Cycle -
Example of profile matching
29. London Output Area Classification (LOAC)
Geographically, retail stores are mainly
located in areas comprising young
professionals working in the science,
technology, finance and insurance sectors
and living in Zone 2 of the London travel
network.
In these areas, residents are highly qualified,
employment rates are high, and employment
is concentrated in the technical, scientific,
finance, insurance and real estate industries.
Compared to the London average, few
individuals originate from the Indian sub-
continent, but mixed ethnic groups are well
represented, as are migrants from pre 2001
EU countries.
30. Contents
1. Study Framework
2. Geographic Business Context
3. Network Analysis
4. Geodemography of Consumers
5. Key Conclusions and Recommendations
31. Key Conclusions
1. Stores are in general clustered together
Two patters can be pointed out. First, stores tend to be located in areas more densely populated and with high
concentration of work. They are located in surfaces like shopping malls and high streets commercial areas. Secondly,
central and west London present the highest concentration of stores.
High-paying jobs draw workers from far way. Thousands travel into the capital each day from all directions. It is easy
to focus on London’ s retail centres as Oxford Street and forget the importance of areas that link them. Areas along
the Underground’ s Central Line may offer good opportunities for prospective new stores.
2. There is no correlation between store location, occupation and index of deprivation
There is no clear correlation between store location and indices of deprivation. Additionally, the fact that mobile
telephone turns out to be ubiquitous, every person is a potential consumer regardless of each individual occupation as
managers, mid-level employees or technical and services workers. Targeting geomarketing campaigns must address
different profiles with personalisation through local based focus.
3. Similarities attract
Londoners span the full spectrum of ethnicities, ages and occupations but socio-economic classes tend to stick
together. Ethnicity patterns emerge where communities have developed along with the amenities that serve them,
such as places for worship, specialist food shops and schools. These in turn attract more people who value them. The
resulting kaleidoscope from London Output Area Classification (LOAC) helps to target local iniciatives and tells where
to set up a new shop.
32. Further Recommendations
Build a Composite Index of geodemographic data
An exclusive analysis of categories of neighbourhood must be set up for each
company. A personal Geodemographic Composite Index is based on a
contextual mode of analysis and sees neighbourhood as bundles of variables.
The index permits to match neighbours’ s attributes with customer profile by
means of an index.
Choosing the right location
Select the areas for expansion based on profile matching analysis. The most
suitable sites must be drawn from geodemographic insights, location data
and marketplace analysis.
It is also vital to track socio-demographic characteristics of consumers,
analyse competitors and segment the market with spatial cluster areas.
Example of a hypothetical Geodemographic Composite Index with
overlay of competitor’ s physical store buffers
33. Research Author
Data Credits
Laurent L. Santos
laurent.santos.18@ucl.ac.uk
ucesljl@ucl.ac.uk
UCL Department of Civil, Environmental & Geomatic Engineering
Prof. James Cheshire, University College London
Consumer Data Research Centre
Greater London Authority
Office of National Statistics
Ordnance Survey
United States Geological Survey
Photo Credits Fred Mouniguet