International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 Proposal And Effectiveness Of A Highly Copelling Direct Mail Method - Establishent And Deployent Of PMOS-DM Hisatoshi Ishiguro, Aoyaa Gakuin University, Japan Kakuro Aasaka, Aoyaa Gakuin University, Japan ABSTRACT No clear processes are used at car dealers when deciding target custoers for direct ail capaigns, and individual sales representatives rely on their experience when aking such decisions. This eans that dealer strategies lose their effectiveness and dealers fail to achieve the desired increase in custoer visits. Thus, for this study, the author has established the Practical Method using Optiization and Statistics for Direct Mail (PMOS-DM) as a ethod of deciding the ost suitable target custoers for direct ail capaigns. Specifically, in order to both clarify the dealer s target custoer types and increase the nuber of custoer visits, the author applied atheatical prograing (cobinatorial optiization) using statistical science to establish a odel for deterining the ost suitable target custoers for direct ail capaigns. This odel has subsequently been applied at copany M dealers, deonstrating significant effectiveness in increasing custoer visits. Keywords: Marketing; Direct Mail; Nueric Siulation; PMOS-DM INTRODUCTION C hanges in the recent arketing environent have ade the personal relationship between businesses and their custoers even ore critical. Businesses are faced with the task of constructing sales schees that are able to flexibly and accurately grasp peculiarities and trends in custoer preferences. This study looks at the effectiveness of the direct ail advertising ethod in bringing custoers to auto dealers, based on the idea that foring personal bonds with custoers is a core coponent of successful sales. One of the unique features of direct ail is the advertiser s ability to select which custoers will receive the ailings. This point is the focus of this study, which attepts to identify an optiu decision-aking process for selecting target custoers. The way that dealers currently select which custoers will receive direct ail is by having individual salespeople choose the on the basis of personal experience and knowledge. Because there is no clear decisionaking process, the response rate is lower than expected and dealers do not achieve the targets set forth in their sales strategies. To address these issues, this study presents a ethodology for selecting direct ail recipients fro the dealer pool that will result in increased response (dealer visit) rates. The proposed echanis for selecting these optiu recipients is the PMOS-DM odel. This odel uses atheatical prograing and statistics to aid the decision-aking process and boost the effectiveness of direct ail advertising. 2012 The Clute Institute 1
International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 IMPORTANCE OF STRATEGIC ADVERTISING Though Japanese copanies are estiated to have spent a total of around six billion yen in advertising over the past four years, the results have not been significant. Businesses need arketing activities that can axiize the effects of their liited advertising budgets. The fact that the auto industry tops the list in ters of oney spent on advertising highlights the critical role that prootional efforts play at these copanies. Activities that involve sales divisions searching in isolation for short-ter strategies no longer work in today s arket. The 4P strategy (Product, Price, Place - distribution channel- and Prootion) is a typical exaple. Instead, copanies ust work with divisions both inside and outside their organizations to develop a strategic approach to arketing activities that gives the central and far-reaching role in their business operations. The key to today s arketing activities is aking full use of custoer data in a way that goes beyond traditional custoer surveys. In an age when our lives are saturated with physical obects, consuer studies ay be the the only way to identify custoers internal wants and needs. Modern advances in inforation technology and the ability to utilize POS data and other electronic transactions data enables businesses to conduct up-to-date analyses of vast aounts of custoer purchase histories using data ining tools and other techniques. It is now possible to analyze the changing and evolving behavior of custoers and gain an even greater understanding of their preferences. Accordingly, arketing tools are also undergoing significant and far-reaching changes. Market segentation based on siple, observable custoer characteristics like age, gender, business size, and business category being replaced with latent clustering or latent class analysis segentation ethods that look at the underlying desires and preferences of custoers. DIRECT MAIL Because this study targets direct ail advertising, the following provides soe background inforation on the role of direct ail in advertising and how this ethod is currently being put into practice. Role in Advertising When looking at advertising effectiveness, there are any odels available that track the stages of psychological change in custoers as they ove fro the point the first encounter an advertiseent to the point where they actually purchase the product. Today s advertising inforation processing odels are largely based the AIDA odel, first proposed by Elias St. Elo Louis in 1898. The AIDA odel breaks down the psychology of custoer purchasing behavior into four steps fro which the acrony is derived: Attention, Interest, Desire, and Action. Television coercials and other advertising and prootion activities seek to target custoer attention, which is the first stage in the AIDA theory. Newspaper inserts attept to ove custoers into the psychological stages of interest and desire. In other words, advertising edia that dealers use to increase the nuber of custoers visiting the shop (such as newspaper inserts or direct ail), are extreely iportant. Furtherore, dealers that are firly established in the counity play a critical role in foring personal relationships with custoers. Direct ail in particular is a eans of pinpointing specific consuers that the dealer wishes to provide with inforation. Because it works in a direct anner, it is considered an iportant advertising ethod for increasing the nuber of custoers who coe into the shop. In other words, direct ail is an effective eans of developing personal relationships between the dealer and its custoers. 2 2012 The Clute Institute
International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 Current Process This section describes the way that direct ail activities are actually carried out. First, dealer headquarters puts together an event and creates the associated direct ail aterials. The individual dealer shops then deterine how any of the ailings to send out. Then, dealer salespeople then use this nuber to decide who aong their custoers will receive the ailings. At this point, each eber of the sales staff akes an individual decision about who to target. The selection process is vague and inexplicit. Take for exaple the process when a vehicle odel is copletely redesigned and the new cars first go on sale. Dealers decide to send out direct ailings announcing the new vehicle, hoping to target custoers who will be receptive to the design concept. However, the salespeople often fail to consider this strategy, with veteran staff ebers sending inforation only to their best custoers and new staff ebers sending it to whoever they happen to have access to. Several probles arise fro this process. First, the direct ailings result in an extreely low response rate. Second, dealer salespeople are not choosing the sae custoers that the dealer wishes to target. And third, because the process is not well defined, the quality of work perfored by different sales personnel is highly inconsistent. Because of these factors, dealers are currently failing to ake their direct ail activities effective. Related Works A search of previous research on direct ail both in Japan and overseas found no enough exaples of studies that focused on strategies for selecting target custoers. Instead, direct ail research tended to focus on optiizing ailing frequency, the reasons custoers chose to visit dealers upon receiving direct ail, which custoers were ost likely to visit the dealer after receiving direct ail, and the design features of the actual direct ail aterials. THE PMOS-DM MODEL Because dealers are currently not aking effective use of direct ail, this paper presents the PMOS-DM odel to boost the ability of direct ail activities to bring in custoers. Research Ais The new odel uses a three-pronged approach to resolving dealers current probles with direct ail activities. The first goal is to increase the response rate, or the percentage of custoers who visit the dealer as a result of receiving direct ail. To achieve this, the PMOS-DM odel uses statistical analysis to deterine which custoers are ost likely to respond. The second is to reflect dealer ais in the recipient selection process. This is achieved by using a siulation driven by atheatical prograing to optiize the selection of target custoers. Finally, the third goal is to clarify the recipient selection process by providing dealers with a odel that outlines a specific approach. Following an explicit odel infored by statistics and atheatical prograing keeps inconsistency aong salespeople to a iniu. The three-pronged approach proposed in this study therefore provides a direct ail ethod that allows dealers to both target their desired custoer segent and boost response rates at the sae tie. The odel proposed here is the PMOS-DM (Practical Method using Optiization and Statistics for Direct Mail). In short, it uses statistics and atheatically prograing for an obective decision-aking process that does not rely on the current selection ethods used by salespeople, which are based on personal knowledge and experience and therefore vague and inexplicit. At the sae tie, the odel ais to boost the direct ail response rate in line with dealer targets. 2012 The Clute Institute 3
International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 The PMOS-DM Method Figure 1 shows the PMOS-DM direct ail ethod, which is designed to effectively bring custoers into the dealer. The odel works according to the following ethod. Step 1 of the process organizes custoer inforation for analysis. Step 2 identifies the custoer segents that are ost likely to be influenced to visit the dealer as a result of direct ail. Step 3 then runs a nuerical siulation to select direct ail recipients. Finally, Step 4 evaluates the ix of target custoers identified in Step 3. OPTIMAL SELECTION USING A MODEL FORMULA In its nuerical siulation, the PMOS-DM odel uses a atheatical forula to select target custoers for direct ail. In coing up with a forula to deterine who should be targeted by direct ail, the authors referred to the forulas shown in (1) and (2) below, which were developed by Koia et al. MIN M W 2 ( f x CR ) (1) J subect to L C f x H C (2) Custoer attributes (e.g. sex, age, age of current vehicle) Custoer nuber W Weighting for custoers with custoer attributes in direct ail target group f Indicates whether or not custoer has attribute (0 or 1) x Marks custoer for direct ailing (0 or 1) R Ideal percentage with custoer attributes in direct ail target group C Total nuber of direct ailing sent L Lower liit for the percentage of direct ailings sent to custoers with attribute H Upper liit for the percentage of direct ailings sent to custoers with attribute Figure: 1 Highly Copelling Direct Mail Method PMOS-DM 4 2012 The Clute Institute
International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 The target function of forula (1) is to iniize the gap between the ideal nuber of direct ailings sent to custoers with attribute (CR ) and the nuber actually sent to custoers with that attribute (Σf x ). In other words, the forula expresses the concept of setting a target value when sending out direct ail. Accordingly, the forula can be adapted to cases where a clear, rational target value can be set. However, the forula cannot be used when it is difficult to set a logical target value for the nuber of direct ailings to be sent and a dozen or so of the dealers that the authors studied did not set one. For those dealers, the authors set up a forula that would clarify the process that senior sales staff used to deterine who should be targeted by a given direct ail capaign. In the process of conducting interviews, the authors learned that senior sales staff uses an abstract ethod of targeting those custoers who see like they would have an easy tie coing into the dealer. The authors then constructed a akeshift definition of this group of custoers as follows. Each group of custoers defined by a given attribute (ale, feale, 20s, 30s, etc.) has different preferences that would otivate the to coe into the dealer. Each custoer s willingness to coe in can be assigned a cuulative value based on that person s attributes. Those with a high cuulative value can be considered the ones who are likely to coe into the shop. With this line of thinking, the authors developed a forula for calculating the total willingness for custoers targeted by direct ail. They then constructed a odel for optiizing those values. Finally, the authors cae up with a set of constraints in order to put liits on the nuber of ailings dealers would send, with the ai of axiizing the effectiveness of those that were sent. Model Forula This is the odel forula used in the nuerical siulation. MAX ( E ( f x )) (3) subect to x C (4) L C f x H C (5) Custoer attributes (e.g. sex, age, age of current vehicle) Custoer nuber E Effect of custoer attribute () on the likelihood that the custoer will visit the dealer f Indicates whether or not custoer has attribute (0 or 1) x Marks custoer for direct ailing (0 or 1) C Total nuber of direct ailing sent L Lower liit for the percentage of direct ailings sent to custoers with attribute H Upper liit for the percentage of direct ailings sent to custoers with attribute This atheatical forula is designed to deterine a value for the variable x. If the value is 1, ailings should be sent to the custoer nuber indicated by. If it is 0, a direct ailing should not be sent. The other variables are paraeters that ust be given values before solving the forula. C, L, and H are set at the discretion of whoever is sending out the direct ail. The value f is deterined based on the custoer inforation that the dealer has. E is deterined later via statistical analysis. 2012 The Clute Institute 5
International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 The roles of the individual forulas are as follows. The obective function in forula (3) is used to axiize the custoer response rate (the percentage of custoers that coe to the dealer as a result of the direct ail). The constraint in forula (4) deterines the nuber of direct ailings that are to be sent out. The constraint in forula (5) deterines how any direct ailings are to be sent to each custoer segent, which is how dealer ais are incorporated into the odel. The atheatical forula is designed so that the nuber of custoer attributes it handles () can be increased at will. Depending on what custoer inforation dealers have, they can liit these attributes to basic life stages or expand the to include hobbies, preferences, and other lifestyle characteristics. Recipient Selection Process This section describes the procedure for using the atheatical forula provided to select direct ail recipients. First, a response likelihood value ust be set for each custoer using the variable E. The list of custoers is then reordered with those with the highest likelihood of responding at the top. The purpose of the obective function in forula (3) is to order custoers according to their likelihood of responding (visiting the dealer as a result of direct ail). Next, this list is used to select the nuber of custoers equal to the nuber of direct ailings (the constraint) to be sent out, starting with those ost likely to respond. For exaple, if 50 direct ailings are to be sent, they would be sent to the top 50 custoers ost likely to respond to the. This is the basic principle behind the developent of the forulas. In addition, when the dealer has a specific ai in ind (e.g. sending a large nuber of direct ailings to woen), the constraint function in forula (5) can be used to incorporate that ai in the calculations. For exaple, if the dealer wanted at least 60% of the 50 ailings to go to woen, the woen custoers would be listed in order of response likelihood and the top 30 custoers would be selected to receive direct ail. The reaining 20 recipients would be selected fro the entire pool of target custoers in order of their response likelihood as well. The purpose of this function is to allow dealers to use their arketing strategies to boost response rate. PUTTING PMOS-DM TO WORK The researchers teaed up with Copany M to use the PMOS-DM to guide direct ailing efforts in conunction with an event showcasing ultiple new vehicle odels. The following sections show how the odel was applied. Organizing Custoer Inforation (Step #1) Since this study requires custoer inforation to select direct ail recipients, raw custoer data used by Copany M dealers was collected. Participating dealers had inforation on a total of 391 custoers, which included data on sex (ale/feale), age (20s, 30s, 40s, 50s, 60+), and age of current vehicle (3-5 years, 6-8 years, 9+ years). Therefore, there were a total of 391 values assigned to in the forula, and a total of 10 different values assigned to. A binary code (0 or 1) was then assigned to the collected custoer inforation in order to analyze it. This resulted in values for the f variable. Recipients of direct ailing could now be deterined based on the dealers custoer inforation. Deterining Response Likelihood (Step #2) The next step was to conduct a survey and analyze the data to deterine which custoer attributes were ost likely to lead custoers to visit a dealer as a result of receiving direct ail. The survey ethod used in this study was to ask custoers of varying attributes (sex, age, vehicle age, etc.) whether receiving direct ail had ever caused the to visit the dealer. Once the results were collected, they were quantified and subected to a Type II analysis deterine which custoers had the highest likelihood of responding to direct ail. This ade it possible to analyze custoer attributes in ters of whether or not they were likely to lead to a dealer visit in ters of an external standard. These response likelihood values were then assigned the variable E. 6 2012 The Clute Institute
International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 The forula below shows the results of this analysis. The discriinant ratio for the analysis results was 77.36%, indicating that they were fairly reliable. The linear discriinant forula produced fro the analysis results is below. y 0x 0x 0x 21 31 11 1.4x 0.9x 1.3x 22 32 12 2.1x 0.9x 23 33 3.7x 1.4 24 1.7x 25 (6) If the linear discriinant is greater than 0, the custoer is likely to visit the dealer as a result of receiving direct ail. If it is less than zero, it indicates that they are not likely to visit. Therefore, the coefficient produced by this forula indicates the response likelihood for the custoer attribute as expressed by E. The results of this analysis, which allowed us to identify which custoers were likely to visit the dealer, are suarized in Table 1. Table 1: Response Likelihood by Attribute Custoer attribute Response likelihood Men 0 Woen -1.4 22-29 years old 0 30-39 years old 0.89 40-49 years old 2.1 50-59 years old 3.7 60+ years old 1.7 Currently driving a vehicle 3 5 years old 0 Currently driving a vehicle 6 8 years old -1.3 Currently driving a vehicle 9+ years old -0.9 Selecting DM Recipients (Step #3) In the next step, direct ail recipients are selected based on a nuerical siulation. This section describes the siulation procedure. First, the custoer inforation collected in Step 1 is plugged into f, and the inforation on response likelihood for each custoer attribute is plugged into E. The nuber of direct ailings to be sent is plugged into C. The upper and lower liits for the percentage of direct ailings to go to custoers with each attribute is set at the dealer s discretion using the variables H and L. Once all the paraeters are set, the siulation is carried out. The forula given in chapter 5 is solved, and the optiu custoers to receive direct ailings are selected. During this process, forulas (3) through (5) are solved as a weighted constraint satisfaction proble. In the weighted constraint satisfaction proble, the weighted constraints are oved to the target function as in (7), where they are added as a way of iniizing the level of deviation outside of the given liits. Even if a feasible solution that satisfies the constraints does not exist, the forula allows dealers to coe as close as possible to eeting the constraints. Here, in constraining the nuber of ailings sent to custoers with the attributes defined in forula (5), it is difficult to set custoer attributes L and H, ensuring that a feasible solution is ore likely to exist. Therefore, when approaching the issue as weighted constraint satisfaction proble, it is best to find a solution that best satisfies forula (5). In other words, this allows dealers to send direct ail to those custoers ost likely to coe into the shop based on dealer strategy. 2012 The Clute Institute 7
International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 MIN ( E W W ( H ( f x ( L C f C f Evaluating the Results (Step #4) )) x x ) ) Once the recipients of direct ailings are selected based on the siulation, whoever is sending out the direct ail checks the siulation results to ake sure that they accurately reflect the dealer s arketing strategy. If the desired results are not achieved, the causes for the discrepancy are identified, the paraeters are adusted, and the siulation is run again. EFFECTIVENESS OF PMOS-DM The effectiveness of the PMOS-DM odel was assessed by coparing the response rates (percentage of direct ail recipients who visited the dealer as a result) when salespeople selected direct ail recipients based on personal knowledge and experience and when recipients were selected using the odel. Five new odels were showcased at the event held by Copany M. Four of the design concepts targeted feale buyers, and one targeted ale buyers. As a result, the dealer s arketing strategy was to target woen in particular throughout a wide range of age groups. This strategy was thus taken into account when verifying the effectiveness of the odel. These verification results are suarized in Table 2. The response rate when direct ail recipients were selected on the basis of personal knowledge and experience of the sales staff was 19%. When selection was ade using the PMOS-DM odel, the rate was 20.4%. (7) Table 2: Verification Results (All) Dealer PMOS-DM Nuber of direct ailings sent 269 269 Nuber of resulting dealer visitors 51 59 Response rate 19.0% 20.4% Table 3 shows the sae inforation for feale custoers only (those targeted in the dealer s arketing strategy). Salespeople generated a 4.2% response rate using their personal knowledge and experience, while the odel generated a 19.8% response rate, signaling a significant iproveent. The effectiveness of the odel was thereby verified in the course of this study. Table 3: Verification Results (Woen) Dealer PMOS-DM Nuber of direct ailings sent 48 61 Nuber of resulting dealer visitors 2 12 Response rate 4.2% 19.8% 8 2012 The Clute Institute
International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 CONCLUSION The present study sought to resolve probles with the effectiveness of direct ailing activities at dealers. Its purpose was to iprove the direct ail response rate by accurately reflecting dealer arketing strategy in the recipient selection process. The PMOS-DM odel was developed in order to select optiu direct ailing recipients and thus achieve these ais. Its effectiveness was then verified through application at actual dealers where an increase in response rate was deonstrated. It is hoped that custoer attributes will be studied in ore detail in the future and will no longer be liited to the three presented here (i.e., sex, age, and age of current vehicle). This added custoer inforation will result in a ore precise siulation outcoe. AUTHOR INFORMATION Hisatoshi Ishiguro is a graduate student of the College of Science and Engineering at Aoyaa Gakuin University. E-ail: c5610162@aoyaa.p Dr. Kakuro Aasaka is a Professor in the College of Science and Engineering at Aoyaa Gakuin University, Japan. He received his Ph.D. degree in Precision Mechanical and Syste Engineering, Statistics and Quality Control at Hiroshia University in 1997. Since oining Toyota Motor Corporation in 1968, He worked as a quality control consultant for any divisions, and the General Manager of the TQM Prootion Division (1998-2000). His specialty is New JIT, Science TQM, Science SQC, Nuerical Siulation (CAE) and Custoer Science. Now, He has been serving as the vice chairan of JSPM (2003-2007) and JOMSA (2008-), the director of JSQC (2001-2003). E-ail: Kakuro_aasaka@ise.aoyaa.ac.p. Corresponding author REFERENCES 1. Dentsu, Japanese total advertiseent rates, http://www.dentsu.co.p/. (in Japanese) 2. Toyo Keizai Inc., Advertiseent rates top 100, http://www.toyokeizai.net/. (in Japanese) 3. K. Aasaka, The validity of advanced TMS, A strategic developent arketing syste-toyota s scientific custoer creative odel utilizing New JIT-, The International Business & Econoics Research Journal, Vol. 6, No. 8, 2007, pp35-42. 4. K. Aasaka, Science SQC. New quality control principle, Springer, 2004 5. K. Shiizu, Theory and strategy of advertiseent, Souse publishers, 2004. (in Japanese) 6. M. Shibasaki, Y. Moriune, Attribution analysis in effect of TV coercial to use AIDA theory, Aasaka-laboratory Study Group, 2000. (in Japanese) 7. T. Shiode, M. Kawaura, Attribution analysis in effect of TV coercial to use AIDA theory (part2), Aasaka-laboratory Study Group, 2001. (in Japanese) 8. Y. Jinno, M. Tabei, Attribution analysis on flyer advertiseent for increasing the effects of drawing custoers, Aasaka-laboratory Study Group, 2001. (in Japanese) 9. J. J. Jonker, N. Piersa, R. Potharst, A decision support syste for direct ailing decision, Decision Support Systes, Vol. 42, 2006, pp915-925. 10. Y. Asahi, K. Jitaa, Y. Sugihara, T, Naatae, T. Yaaguchi, Identifying latent factors driving departent store purchases: A direct ail strategy, Operations research as a anageent science research, Vol. 49, No. 2, 2004, pp75-80. (in Japanese) 11. J. R. Bult, T. Wansbeek, Optial selection for direct ail, Marketing Science, Vol. 14, No. 4, 2005 pp378-394. 12. T. Kiura, M. Yaai, K. Aasaka, A study of Scientific Approach Method for Direct Mail, SAM-DM effectiveness of attracting custoer utilizing Advanced TMS, Proceedings of the 5 th Asian Quality Congress, Incheon, Korea, 2007, pp938-945. 13. T. Koia, T. Kiura, M. Yaai, K. Aasaka, Proposal and developent of the direct ail ethod PMOS-DM for effectively attracting custoers, International Journal of Manageent and Inforation Systes, Vol. 14, No. 5, 2010, pp15-22 2012 The Clute Institute 9
International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 NOTES 10 2012 The Clute Institute