ATM
GIS-Based Approach for Delineating ATM Trade / Market Areas
Bilal Farhan Email: [email protected]
Outline Background: ATM trade area delineation Issues with existing approaches Objective Methodology Results Conclusions
Introduction ATM
growth opportunities
New
sites shouldn’t be too close to existing network (cannibalization)
GIS
is used to help with ATM site selection
ATM Growth Opportunity
ATM Opportunity Evaluation
Trade area delineation Trade area delineation for ATMs important Trade areas are geographic areas where most of the customers originate to use an ATM Several ways to delineate trade areas in GIS
Customer based Non-customer based
Customer-based trade areas Based
on customer data, trade areas can be delineated Several ways to do it in GIS
Non-customer based trade area
Non-customer based trade areas
Rings Thiessen polygons
Assumption that trade areas are represented by 2-mi simple rings
Outline Background: ATM trade area delineation Issues with existing approaches Objective Methodology Results Conclusions
Issues with existing approach
Assumption that households are willing to travel up to 2 miles to reach an ATM
Assumption that likelihood of ATM usage is the same within trade area
Trade area extent is arbitrary
Distance Decay not considered
Therefore, a new approach is needed to better describe the trade area
Outline Background: ATM trade area delineation Issues with existing approaches Objective Methodology Results Conclusions
Objective Develop
an approach for delineating trade/market areas for ATMs Estimate
the extent of trade area Examine impact of distance
Outline Background: ATM trade area delineation Issues with existing approaches Objective
Methodology
Results Conclusions
Estimating the extent of trade area Simple plotting of households doesn’t necessarily indicate extent of trade area Although customer destination (ATM location) is known, customer origin is not necessarily known.
Origin of customers could be place of work
Therefore, need a better method to estimate where they are coming from
Attraction-constrained gravity model i = index for origin (point, area) j =index for destination (ATM) Tij = no. of transactions done at j by customer at i Aj = sum of transactions at j Vi = measure of propulsiveness for i µ = propulsiveness parameter ß = distance decay parameter Dij = distance between i and j µ
Tij =
A j *Vi * exp(− βDij )
∑V
i
j
µ
* exp(− βDij )
i=1
i=2
ATM
j
Calibration of the gravity model Calibration:
estimation of 2 parameters– distance decay (β) and propulsiveness (µ) parameters Linearization
of gravity model Regression is used to estimate parameters
Linearized gravity model 1 1 LnTij − ∑ LnTij = ∑ α h f h ( X ijh ) − ∑ f h ( X ijh ) n j n j h
Y Dependent variable
X 1, X 2 , … Independent variables
Fotheringham and O'Kelly (1989).
Variables needed for regression Household activity matrix
1
Origins (grid cells)
1
Origins (grid cells)
Vi :
1 2 3 4 5 6 7 8 : : : :
Distance Matrix
Tij
Destination (ATMs) : 2 3 4 5
1 2 3 4 5 6 7 8 : : : :
2
Origin (grid cell) 1 2 3 4 5 6 7 8 : : : :
Propulsiveness 32 43 56 : : : : : : : : :
Dij Destination (ATMs) : 3 4 5
:
Y = Tij X1 = Dij X2 = Vi
Tij: Transactions between origins and destinations Dij: Distance between origins (grid cells) and destinations (ATMs) Vi: Propulsiveness of origins
Preparing regression variables
Identify study area
Area was chosen based on ATMs whose customers are likely to use them near their residence
Identify competitor ATMs and households in the study area Quantify regression variables
Tij Dij Vi
Study Area
Competitor ATMs and households in the study area
Tij
Spatial joining
Dij: distance between each origin and all ATMs
Outline Background: ATM trade area delineation Issues with existing approaches Objective Methodology
Results
Conclusions
Regression variables: Tij, Dij, Vi Origin (cell) 69775897 69775897 69775898 69775898 69775898 69775898 69775898 69775899 69775899 69775900 69775900 69775900 69775900 69775900 69775900 69775900
Destination (ATM) Tij BX3510 1 BX3471 1 00003527 1 P61695 2 TCS04732 1 BX3510 2 BX3471 8 BX3510 2 BX3471 3 80620046 1 A510259 1 P61695 1 TCS04732 1 F002767 1 F002768 1 BX3510 2
Dij Vi (deposit) 2.8 122865 2.26 122865 2.09 152253 0.26 152253 1.54 152253 2.95 152253 2.2 152253 3.11 505329 2.17 505329 0.6 454886 0.55 454886 0.75 454886 1.11 454886 2.3 454886 4.54 454886 3.28 454886
LnTij - Avg LnTij Dij - Avg Dij Ln Vi - Avg Ln Vi -0.58 0.70 -0.05 -0.58 0.16 -0.05 -0.58 -0.01 0.25 0.11 -1.84 0.25 -0.58 -0.56 0.25 0.11 0.85 0.25 1.50 0.10 0.25 0.11 1.01 0.22 0.51 0.07 0.22 -0.58 -1.50 0.22 -0.58 -1.55 0.22 -0.58 -1.35 0.22 -0.58 -0.99 0.22 -0.58 0.20 0.22 -0.58 2.44 0.22 0.11 1.18 0.22
Distance Decay Parameter (β) = 1.47
ATM Trade Area (β=1.47)
ATM Trade Area (β = 1)
Role of GIS Spatial
data organization Data processing Calculations Visualization
Conclusions Trade area delineation is important to evaluate ATM growth opportunities Simple approaches may not be adequate The suggested approach delineates trade areas based on destination information GIS is instrumental for ATM trade area delineation
Bilal Farhan Email: [email protected]
Outline Background: ATM trade area delineation Issues with existing approaches Objective Methodology Results Conclusions
Introduction ATM
growth opportunities
New
sites shouldn’t be too close to existing network (cannibalization)
GIS
is used to help with ATM site selection
ATM Growth Opportunity
ATM Opportunity Evaluation
Trade area delineation Trade area delineation for ATMs important Trade areas are geographic areas where most of the customers originate to use an ATM Several ways to delineate trade areas in GIS
Customer based Non-customer based
Customer-based trade areas Based
on customer data, trade areas can be delineated Several ways to do it in GIS
Non-customer based trade area
Non-customer based trade areas
Rings Thiessen polygons
Assumption that trade areas are represented by 2-mi simple rings
Outline Background: ATM trade area delineation Issues with existing approaches Objective Methodology Results Conclusions
Issues with existing approach
Assumption that households are willing to travel up to 2 miles to reach an ATM
Assumption that likelihood of ATM usage is the same within trade area
Trade area extent is arbitrary
Distance Decay not considered
Therefore, a new approach is needed to better describe the trade area
Outline Background: ATM trade area delineation Issues with existing approaches Objective Methodology Results Conclusions
Objective Develop
an approach for delineating trade/market areas for ATMs Estimate
the extent of trade area Examine impact of distance
Outline Background: ATM trade area delineation Issues with existing approaches Objective
Methodology
Results Conclusions
Estimating the extent of trade area Simple plotting of households doesn’t necessarily indicate extent of trade area Although customer destination (ATM location) is known, customer origin is not necessarily known.
Origin of customers could be place of work
Therefore, need a better method to estimate where they are coming from
Attraction-constrained gravity model i = index for origin (point, area) j =index for destination (ATM) Tij = no. of transactions done at j by customer at i Aj = sum of transactions at j Vi = measure of propulsiveness for i µ = propulsiveness parameter ß = distance decay parameter Dij = distance between i and j µ
Tij =
A j *Vi * exp(− βDij )
∑V
i
j
µ
* exp(− βDij )
i=1
i=2
ATM
j
Calibration of the gravity model Calibration:
estimation of 2 parameters– distance decay (β) and propulsiveness (µ) parameters Linearization
of gravity model Regression is used to estimate parameters
Linearized gravity model 1 1 LnTij − ∑ LnTij = ∑ α h f h ( X ijh ) − ∑ f h ( X ijh ) n j n j h
Y Dependent variable
X 1, X 2 , … Independent variables
Fotheringham and O'Kelly (1989).
Variables needed for regression Household activity matrix
1
Origins (grid cells)
1
Origins (grid cells)
Vi :
1 2 3 4 5 6 7 8 : : : :
Distance Matrix
Tij
Destination (ATMs) : 2 3 4 5
1 2 3 4 5 6 7 8 : : : :
2
Origin (grid cell) 1 2 3 4 5 6 7 8 : : : :
Propulsiveness 32 43 56 : : : : : : : : :
Dij Destination (ATMs) : 3 4 5
:
Y = Tij X1 = Dij X2 = Vi
Tij: Transactions between origins and destinations Dij: Distance between origins (grid cells) and destinations (ATMs) Vi: Propulsiveness of origins
Preparing regression variables
Identify study area
Area was chosen based on ATMs whose customers are likely to use them near their residence
Identify competitor ATMs and households in the study area Quantify regression variables
Tij Dij Vi
Study Area
Competitor ATMs and households in the study area
Tij
Spatial joining
Dij: distance between each origin and all ATMs
Outline Background: ATM trade area delineation Issues with existing approaches Objective Methodology
Results
Conclusions
Regression variables: Tij, Dij, Vi Origin (cell) 69775897 69775897 69775898 69775898 69775898 69775898 69775898 69775899 69775899 69775900 69775900 69775900 69775900 69775900 69775900 69775900
Destination (ATM) Tij BX3510 1 BX3471 1 00003527 1 P61695 2 TCS04732 1 BX3510 2 BX3471 8 BX3510 2 BX3471 3 80620046 1 A510259 1 P61695 1 TCS04732 1 F002767 1 F002768 1 BX3510 2
Dij Vi (deposit) 2.8 122865 2.26 122865 2.09 152253 0.26 152253 1.54 152253 2.95 152253 2.2 152253 3.11 505329 2.17 505329 0.6 454886 0.55 454886 0.75 454886 1.11 454886 2.3 454886 4.54 454886 3.28 454886
LnTij - Avg LnTij Dij - Avg Dij Ln Vi - Avg Ln Vi -0.58 0.70 -0.05 -0.58 0.16 -0.05 -0.58 -0.01 0.25 0.11 -1.84 0.25 -0.58 -0.56 0.25 0.11 0.85 0.25 1.50 0.10 0.25 0.11 1.01 0.22 0.51 0.07 0.22 -0.58 -1.50 0.22 -0.58 -1.55 0.22 -0.58 -1.35 0.22 -0.58 -0.99 0.22 -0.58 0.20 0.22 -0.58 2.44 0.22 0.11 1.18 0.22
Distance Decay Parameter (β) = 1.47
ATM Trade Area (β=1.47)
ATM Trade Area (β = 1)
Role of GIS Spatial
data organization Data processing Calculations Visualization
Conclusions Trade area delineation is important to evaluate ATM growth opportunities Simple approaches may not be adequate The suggested approach delineates trade areas based on destination information GIS is instrumental for ATM trade area delineation
