A GIS-Based Approach for Evaluating ATM Partnership Opportunities

Size: px
Start display at page:

Download "A GIS-Based Approach for Evaluating ATM Partnership Opportunities"

Transcription

1 A GIS-Based Approach for Evaluating ATM Partnership Opportunities Bilal Farhan, Ph.D Palmleaf Ct. Columbus, Ohio, USA Feb. 25,

2 Abstract To retain existing customers and acquire new ones, financial institutions continually increase the location convenience of their branches and ATMs. One way to achieve this is by expanding their ATM and / or branch networks. This is usually done in two ways: building new facilities for branches and ATMs; or acquiring potential partner s locations. The second option is preferable partly because it can give banks immediate access to more customers. However, the locations of a potential partner may not necessarily be a good geographic fit to the existing ATM network. Therefore, such partnerships should be evaluated carefully so that the right ATM partner is selected. This paper develops a GIS-based approach for evaluating ATM partnership opportunities. It is shown how GIS is instrumental throughout this research. Keywords: GIS, business, ATM, location modeling, coverage. 1. Introduction Location convenience is an important factor when customers select a financial institution (see Mylonakis et al. 1998; Driscoll 1999). A customer may find a bank convenient if it has a branch or an Automated Teller Machine (ATM) near his / her residence or workplace, say within 2 Km 1. To stay competitive, banks usually attempt to increase convenience (be as close as possible to customers) by expanding their bank and / or ATM networks. Basically, such expansions could be done in at least two ways (see Birkin et al 2002a): by building branches / installing ATMs in new locations (so called organic growth); or by acquiring an existing (e.g. competitor, partner, etc.) network. The former option is likely to be expensive and time consuming, so many banks resort to acquiring or partnering with already established ATM and / or branch networks. Although the second option is a relatively quick way to increase network coverage, an acquired branch / ATM network is unlikely to be a perfect complement to an existing 1 ATMs are common in the USA (and much of the world). Customers can use ATMs to access their bank accounts to obtain cash money or check available funds, etc. ATMs are available in banks and in facilities such as malls, shopping centers, gas stations, parking facilities, etc. 2

3 network, due to possible coverage overlap and / or lack of coverage. Therefore, such acquisition / partnership opportunities should be evaluated carefully so that opportunities resulting in minimal inefficiencies are selected. As an example, as shown in Fig. 1, Bank X should choose B as a partner over potential partner A because B has a better spatial / geographical distribution (i.e. serves more customer demand). In other words, many locations of partner A are in areas unlikely to be desirable by Bank X because they would not cover much additional demand In this research, we focus on evaluating ATM partnership opportunities, even though the developed methodology could be applied in other contexts. While evaluating ATM partnership opportunities can be thought of as a merger / acquisition, approaches that deal with optimizing merger and acquisition for banks (see, e.g., Birkin et al 2002b) are not necessarily appropriate for evaluating ATM partnerships. One reason is that such 3

4 approaches compare among potential partners and do not compare them to an ideal partner. That is, if all potential partners do not have an appropriate geographical distribution, such methods would still select a potential partner, although in such a case, the best course of action might be not to select any partner. Another issue that is not directly addressed in existing approaches is budgetary constraints. As an example, if a Bank X has 5 ATMs in an urban area (see Fig. 1), and has a limited budget that is enough to add 2 ATMs, the question is which 2 sites (out of the 3 Partner B sites) should be selected such that they give the best customer coverage. Again, existing approaches do not directly deal with this issue. To the best of our knowledge, there are no approaches that deal with the two important issues discussed above. Therefore, this paper deals with these issues by developing a GIS-based approach for evaluating ATM partnership opportunities. As mentioned before, this research focuses on ATMs, not bank branches. This paper is organized as follows. In the next section we provide background about relevant spatial optimization models. In the section that follows, the methodology is described. Then application results are detailed. Finally, conclusions are given. 2. Background Evaluation of potential ATM partners should be done according to some criteria. For example, in the case of merger and acquisition of banks, revenue, market share, cost, or profits could be used to select the best partner (see Birkin et al 2002b). In our case (i.e. ATM partnership opportunity evaluation), since location convenience is important to 4

5 keep existing customers from switching to competitors, it is important to cover as many customers as possible with ATMs. The first step is to find the number of ATMs needed (and their locations) to cover the customers in a given urban area. The location set covering problem (LSCP) detailed in Toregas et al. (1971) can be applied to achieve this (for detailed reviews of location models, see Drezner (1995) or Owen and Daskin (1998)). The objective of the LSCP is to find the minimum number of facilities (and their location) needed to cover all demand. The following notation is introduced before presenting the mathematical formulation: i = index of demand areas j = index of potential facility sites d ij = distance or travel time between demand area i and potential facility site j S = maximum acceptable service distance (or time) N i = { j d S} Decision variables: ij X j 1 if a facility is located at site j = 0 otherwise Location Set Covering Problem (LSCP) Minimize X j (1) j Subject to: j N i X 1 i, (2) j X j {0, 1} j. (3) The objective (1) minimizes the total number of required facilities (i.e. ATMs) needed to completely cover demand areas. Constraints (2) assures that every demand area i is 5

6 covered by a facility (e.g. ATM) within the coverage standard. Constraints (3) are integrality requirements. While the LSCP is useful for identifying how many facilities are needed and where they should be sited, it requires that all customer demand be covered see Constraints (2). Recall that when evaluating ATMs, there may be budget constraints that do not make it possible to cover all customer demand. The maximal covering location problem (MCLP), proposed by Church and ReVelle (1974), addresses the issue of budget constraint. The following additional notation is introduced before presenting the MCLP: a i = demand in area i p = number of facilities to be located Decision variables: Z i 1 if area i is covered = 0 otherwise Maximal Covering Location Problem (MCLP) Maximize a Z (4) i i i Subject to: Z i, (5) i X j j N i X j p, (6) j X j {0, 1} j, (7) Z i {0, 1} i. The objective (4) maximizes the demand covered. Constraints (5) are used for tracking whether demand area i is covered or not by a sited facility (e.g. ATM) within the 6

7 coverage standard. Constraint (6) takes into account budget limitations by specifying the number of allowed facilities to locate. Constraints (7) are integrality conditions. 3. Methodology In this research, it is logical to employ the MCLP to select / evaluate potential ATM partners. MCLP will be employed to address the two issues we mentioned: comparing a potential partner to an ideal partner, and considering budget constraints (i.e. number of ATMs allowed). The procedure for evaluating potential ATM partners is as follows: 1. Consider a study area (e.g. city or urban area) 2. Identify the demand to be covered in the study area 3. Identify potential ATM sites in the study area a) Identify potential sites for an ideal partner. This is important for comparison purposes with potential partners. It is an ideal case because it considers every possible location in the study area. b) Identify potential sites for the partner in consideration. Unlike the previous case, the potential sites here are the actual locations the potential partner has. Depending on the budget, part or all of these locations would be selected. 4. Given the demand, the potential ATM sites (as defined by a and b above) and the number of allowed ATMs to locate (according to budget), solve the MCLP problem to find the optimal locations for the two cases above: a and b. 5. Compare between the selected locations resulting from a and b above based on the coverage criterion (amount of customers covered by ATMs in each case), and recommend the best course of action. 7

8 4. Application results ATM network expansion in the study area. Consider the following situation for applying the above methodology. Bank X has currently 32 ATMs in the study area (see Fig. 2). Potential partner A has 26 potential ATM locations distributed throughout the study area as well. This is an opportunity for Bank X to expand its ATM network by up to 26 ATMs. To determine how well the new locations (26) could serve the existing ATM network, the MCLP was applied. The first step was to determine the criteria for coverage. We consider that we are attempting to cover as many customers as possible (i.e. customer demand). This is logical because, after all, banks attempt to maximize convenience for their customers. Identifying demand. The next step is to prepare the spatial data for implementation. The urban area is divided into grid cells. The centroids of the grid cells were then determined using an ArcView GIS functionality. The location of customer demand was represented by these centroids, whereas the amount of demand at each centorid was calculated by summing up number of Bank X customers within the grid cell (see Fig.2). This was accomplished using the spatial join capability in ArcView. The demand points were refined so that customers already covered by existing ATMs (i.e. customers within 2 miles of an ATM) were excluded. This was achieved by doing spatial queries / buffers in ArcView (see Fig. 3). 8

9 Identifying potential sites. Potential sites for new ATMs were also represented by points. For the ideal case (Case a above), these points represent every possible location 9

10 for a new ATM. These include the centroids of the grid cells plus the existing locations of potential partner A. For Case b, the potential locations are the existing locations of partner A. Determine optimal locations. Given the demand and potential sites points (see Fig. 4), the next step is to find the optimal locations for the two cases, a and b. For Case a, there were 3,757 demand points, and 3,783 potential sites (which includes Partner A locations). For Case b, the demand points are the same, but the potential sites are only the existing locations of partner A (26). Finding optimal locations was achieved by applying the MCLP. The spatial data were first prepared within ArcView. Then, Avenue, a scripting language within ArcView, was used to produce the MCLP code in text file format. SAS was then used to solve the optimization model. Results were exported from SAS for display and analysis in ArcView. 10

11 Fig. 5 shows the solution (selected sites) based on cases a and b, assuming that there is a budget for 5 additional ATMs. For Case a, this answers the following question: given 3,783 potential sites, where are the optimal five? This is an ideal case because every possible site was considered. For Case b, the solution is also shown in Fig. 5. Similarly, the selected sites shown answer the question that given 26 locations potential sites, where are the optimal / best five? To evaluate partner A, a comparison between cases a (ideal partner) and b (Partner A) is made based on coverage. The percentage coverage given that the number of allowed facilities is 5 is 74% (3,741 customers covered by Partner A / 5,035 customers covered by the ideal partner). In other words, partner A can cover 74% of what can be covered under ideal conditions. The MCLP was also applied for other budget scenarios (10, 15, 20, and 26 ATMs). Fig. 6 shows the coverage resulting from selecting 5, 10, 15, and 26 optimal sites out of the set of potential sites. In addition to addressing the above two issues, Fig. 6 could be used 11

12 to deal with other issues. For example, even if Bank X has a budget for a certain number of sites, does it really help to have all of them? Fig. 6 could be used to answer this question. Assume that we have a budget for 26 sites, are they all needed? Fig. 6 suggests that taking up to 20 sites (covering 7,171 customers) may be useful given the relative increase in coverage compared to taking 15 sites. As the number of sites increase to 26 (covering 7,248 customers), though, not much additional benefit in terms of additional coverage can be obtained. This suggests that even if there is a budget for 26 sites, only 20 sites should be taken from Partner A. Fig. 6: Comparison of coverage between "ideal" ATM sites and potential partner A 16,000 14,000 Bank X HHs covered 12,000 10,000 8,000 6,000 4,000 2,000 "Ideal" partner Partner A Selected sites In the case that there are more than one potential partner to consider, such partners can be included in the analysis in a similar fashion as above. For example, if B was another potential partner to consider along with Partner A, we could find the 5, 10, 15, 20, and 26 12

13 optimal locations from the available Partner B sites. Fig. 7 shows the coverage provided by these sites. It can be seen that Partner B would provide more coverage than Partner A, which suggests that it may be a better partner than A. Fig. 7: Comparison of coverage between "ideal" ATM sites and potential partners A & B 16,000 14,000 Bank X HHs covered 12,000 10,000 8,000 6,000 4,000 2,000 "Ideal" partner Partner A Partner B Selected sites 5. Conclusions Banks expand their ATM networks basically to keep existing customers or acquire new ones. Expanding an ATM network can be done essentially in two ways: building new locations (organic growth) or by partnering with a retailer. The first option is likely to be costly and can take a long time, so the latter may be a better option. Choosing the right partner is not an easy decision to make. This paper developed a GIS-based approach that could be used to help make such a decision. It takes into consideration issues that were 13

14 not accounted for in previous approaches: comparing potential partners to an ideal partner and budget limitations. As shown in the paper, the developed approach was used for evaluating one or more partners so that the best option is chosen. The decision that should be made is which partner to select, or not to select any partner if they are not good enough. Additionally, this approach reduces inefficiencies by determining the appropriate number of ATMs needed to cover demand. GIS was instrumental throughout this research. As was shown in this paper, GIS was used for representing spatial data (e.g. demand represented using grid cell centroids), processing spatial data (e.g. performing spatial queries and buffers to determine uncovered demand), preparing the data for the spatial optimization model (i.e. writing a computer code in Avenue to be solved by the optimization software), validation (i.e. showing the solutions of the optimization model on the map to assure that the model is applied properly) and display purposes (i.e. mapping). 14

15 References 1. Birkin, M.; Clarke, G.; and Clarke, M. (2002a), Retail Intelligence and Network Planning, England, John Wiley. 2. Birkin, M.; Clarke, G.; and Douglas, L., (2002b), Optimizing spatial mergers: commercial and regulatory perspectives, Progress in Planning 58, Church, R., and ReVelle, C. (1974) The maximal covering location problem, Papers of the Regional Science Association, 32, Drezner, Z. (1995) Facility location: a survey of applications and methods, New York: Springer. 5. Driscoll, J. (1999), ``Bank wars: episode 2. The branches strike back'', Bank Marketing, Vol. 31 No. 12, pp Mylonakis, J., Malliaris, P. and Siomkos, G. (1998), ``Marketing-driven factors influencing savers in the hellenic bank market'', Journal of Applied Business Research, Vol. 14 No. 2, pp Owen, S., and Daskin, M. (1998) Strategic facility location: a review. European Journal of Operational Research, 111, Toregas C., Swain, R., ReVelle, C., and Bergman, L. (1971) The location of emergency service facilities. Operations Research, 19,