PACIFIC RIM PROPERTY RESEARCH JOURNAL Pacific Rim Property Research Journal, Vol 17, No 1, 2011 157 HOW ATTRACTIVE IS A SHOP: A FUZZY-EXPERT SYSTEM MATTHEW LEUNG and K.C. WONG University of Hong Kong ABSTRACT Professional planners of shopping centres know that intangible factors of a shop, including product type, quality, variety, pricing, branding, and image, are all vital in determining the pedestrian flow into the shop. These intangible factors cannot be easily quantified. Yet by experience, professional planners know how to assess them, and hence assign different shops to different strategic locations within a shopping centre, in order to optimize pedestrian flow. The objective of this study is to model experts’ professional judgment on these intangible factors. A model combining a fuzzy expert system and regression is found capable of predicting pedestrian flow at reasonable accuracies. By fine tuning the location of potential tenants, using an integrated simulation model, it is possible to optimise the shopping centre’s rental performance, even at the early stage of the shopping centre design. Keywords: Shopping centre, shop attractiveness, intangible factors, pedestrian flow, fuzzy expert system INTRODUCTION Success of any retail shop or shopping centre relies heavily on the pedestrian flow. Dawson (1983) stated that “sales are closely related to the volume of pedestrian traffic passing the shop.” Northern (1984) believed that the ultimate success of any shopping centre would be directly proportional to the number of shoppers who pass through the centre. Location is one of the significant determinants. Various neoclassical theories about location (described by Brown (1992) as traditional statistical models) have been studied, such as Central Place Theory, the Principle of Minimum Differentiation, Spatial Interaction Theory, etc. Further to these theories, regression models have been adopted to analyze the impact of various tangible factors, such as the location, physical dimensions of the shop, the walking distance between the shop and major traffic and attraction points, etc. Yet only using tangible factors is insufficient to predict the pedestrian flow drawn by a shop or at any particular point within a shopping centre (Northern 1984). He has identified various significant qualitative Pacific Rim Property Research Journal, Vol 17, No 1, 2011 158 factors, namely tenant mix, centre/ shop characteristics, brand name and prestige images, nearby competition, the ‘quality’ of retail space available, etc. Northern has classified shopping centre image as a qualitative factor; however Dennis et al. (2005) has attempted to quantify the impact of image on the performance of shopping centres. He has investigated the relationship between “image” or “attractiveness” and individual shopper behaviour. He asserted image as the complete mix of cues (e.g. sensory) which communicates with customers and influences shopping behaviour. He has consolidated the gravity, spatial interaction and Central Place approaches in his attitude-behaviour theory. The attractiveness of shopping centres has been quantified and triangulated with a branding framework. His study reported that pleasure and enjoyment, deterrence effect of travel and motivation significantly affected a shopping centre as an object of desire. The ‘pleasure’ and ‘enjoyment’ would be affected by atmospheric stimulus, image of the shopping centre environment; image of stores and products, and arousal stimulation affect. Deterrence effect would be influenced by travel distance and time, distance to a competing centre, size and attractiveness of other centres in and near the catchment area. Shopper’s motivation was significantly related to self-esteem and relatedness. Steenkamp and Wedel (1991) stated that retail image was the key parameter reflecting the total value of shopping centres. A unique and favorable image helps create a sustainable competitive advantage and establish a clear marketing position from other competitors. Five identified attributes of shopping centre image are merchandising, accessibility, service, atmospherics and entertainment (Sit et al., 2003). Besides studying shopping centres, Martineau (1958) has studied and shown that the drawing force of a store was the store personality or image – the way in which the store was defined in the shopper’s mind, partly by its functional qualities and partly by an area of psychological attributes. He stated that “whereas the retailer thinks of himself as a merchant concerned with value and quality, there is a wide range of intangibles which also play a critical role in the success or failure of his store”. Current practitioners predict the pedestrian flow mainly with their own experience and judgment. In order to pursue the best performance, as well as its consistency, of a shopping centre, a scientific management system to control the layout design and tenant mix scheme is necessary and desirable. Compared to tangible attributes, intangible attributes are more difficult to measure and study. This paper will investigate the measurement of shop attractiveness, which is one of the most significant intangible attributes, and the relationship between shop attractiveness and pedestrian flow, with the data collected in three shopping centres in Hong Kong. Pacific Rim Property Research Journal, Vol 17, No 1, 2011 159 SHOP ATTRACTIVENESS The attractiveness of a shop is defined as its ability to draw potential customers into, and around the neighborhood of, the shop. Such attractiveness is determined by both tangible (e.g. shop size, location, etc.) and intangible factors (e.g. branding, market position, etc.). Intangible factors are difficult to be quantified, as individuals have different perceptions on various attributes. A fuzzy expert system, which is a codification of the common sense, imitating how people use their imprecise information to make the right decision, is designed to predict pedestrian flow at specific points inside a shopping centre, It is a widely accepted tool for modeling nonlinear functions of arbitrary complexity and dealing with linguistic variables, which transform descriptive words or sentences into quantitative values. Wong and So (1995), Bagnoli and Smith (1997, 1998) and Ng et al. (2002) applied fuzzy logic on solving problems in the real estate and construction industries. Wong and So have constructed a fuzzy reasoning model and applied the model on contract decision making in Hong Kong. Bagnoli and Smith investigated how to apply fuzzy logic on real estate valuation. Ng et al. have demonstrated how to set up the fuzzy membership functions of procurement selection criteria. The system proposed in this paper involves three main components: fuzzification, inferencing and defuzzification. Fuzzification is the procedure to convert raw data (i.e. linguistic variables) into membership functions. The membership function is a generalization of the linguistic input functions in classical sets. It can be represented graphically of the participation magnitude of each input with an interval ranging from zero (false) to one (true). Inferencing refers to the reasoning or logic applied in the system which constitutes the rule base. Rule base specifies conclusions drawn from assertions known or assumed to be true (Jantzen 1999). With the membership functions and truth values of inputs obtained in fuzzification, the rules applied will be invoked to determine the result, which will be mapped onto a membership function and truth value controlling the output variable. In this paper, the centroid method, which determines the centre of the area of the combined membership functions and produces more representative results than other methods (Nurcahyo et al. 2003), is chosen to perform the defuzzification process. FUZZIFICATION The first step in fuzzy logic is to establish membership functions of input terms which are generalization of the indicator functions. In fuzzy logic, it represents the degree of truth as an extension of valuation. In other words, common sense is codified and reflected in the membership function. The membership functions are established with Pacific Rim Property Research Journal, Vol 17, No 1, 2011 160 the aid of a questionnaire, which collects and quantifies the views of interviewees with different classical sets. The framework of the questionnaire (as shown in Appendix A), with reference to Ng et al. (2002), is designed to collect the maximum, minimum and average values of both input and output functions. The questionnaires are distributed to ten current practitioners/ industry experts who could share their experience towards the operation and mechanism of the retail market. Quantitative data of the linguistic variables are collected to develop the membership functions. Common shapes of membership functions are triangular, trapezoids, smooth triangular and smooth trapezoid. During practical application, the number of curves and their placements are far more critical than the shape type. Three to seven curves (i.e. terms) are generally adequate to cover the universe of disclosure, representing all objects that come into consideration, of the input and output values. In order to design a system for market practitioners who do not process advance mathematical knowledge, a simple triangular function is chosen for simple graphical representation. Another advantage is that simple triangular function can be processed with some common softwares; for example, SPSS, Microsoft Excel. With reference to previous studies, eight input variables (IVs) and three output variables (OVs) were suggested for the model. Input variables, including shop youth image, quality of goods sold, pricing competitiveness, branding, variety of goods sold, trading, shop size and shop location, are the settings and characters of the shops. Output variables refer to the percentage of pedestrians, belonging to different age groups, being attracted to the shop. They consist of three age groups: OV1 (age 15- 24); OV2 (age 25-35) and OV3 (age above 35). Each variable is associated with several terms. The definition of input and output variables are shown in Appendix B. The interviewees are required to indicate their views towards the variables which will be expressed in percentage terms according to the specified formulas to elaborate the linguistic variables. Maximum, mean and minimum values of terms are recorded and will be adopted in later inferencing. A descriptive statistical table of linguistic variables together with terms is presented in Appendix C. For example, see IV15 in Appendix C. Experts consider that a shop’s youth image is said to be mature if 65% to 85% of the goods found in the shop are mature – where a mature good is defined as one that shoppers aged above 35 may find interest in. The mean of all 10 experts’ opinions is 76%. This would give an overall triangular distribution of: 65%, 76%, and 85%; against which the membership function of 0.0; 1.0; and 0.0 could be assigned respectively. Given this triangular distribution, any shop with a certain percentage of mature goods found inside the shop (ranging from Pacific Rim Property Research Journal, Vol 17, No 1, 2011 161 65 to 85%) could be assigned, by interpolation, a “degree” of membership (from 0 to 1) to the mature image. RULE BASE Rules of experience constitute the inference engine which is applied to the shops in estimating their drawing power. In this paper, the rules are categorized by trades, namely restaurant, fashion, beauty, jewelry, gift/ furniture, electrical appliance, store/ supermarket and retail services. They are then subcategorized with other aspects, such as image, quality, pricing, branding, goods variety and shop size. 30 rules (as listed in Appendix D) have been formulated and are considered to cover most, if not all, daily situations. The outcome of each rule is the experts’ rules of thumb applied in their practices. For example in Appendix D, Rule 1 – the Youngsters’ Café Rule – could be read as follows: IF the shop’s youth image is teenagers or youngster; AND goods quality is bad or just okay; and pricing competitiveness is very cheap or cheap; AND branding is unknown brand or local unpopular or local popular or international; AND goods variety is limited; AND trade is restaurant/entertainment; AND shop size is tiny or small or normal; AND location is easy to access or convenient or very convenient; then the shop’s attractiveness to teenagers (age 15-24) is very high; AND to middle age (25-45) is medium; AND to mature customers (above 45) is very low. How applicable Rule 1 is to a specific shop would depend on (a) the degree of membership of each individual IVs, as assessed in the previous section on fuzzification ; and (b) the use of the Max-Min Rule. i.e. choosing the maximum of the various degrees of membership wherever the logical syntax in-between the IVs (or groups of IVs) is OR; and minimum when it is AND. PEDESTRIAN FLOW In this paper, three regional shopping centres in Hong Kong are chosen to study, namely Festival Walk, Langham Place and Pacific Place. All of them share many characteristics in common, such as similar size and scale, managed by leasing and property management professionals, attached to railway/ subway, office buildings attached, etc. These help control the external factors of the subjects. The main objective of the fuzzy model is to evaluate the pedestrian flow at a particular point in a shopping centre with the estimated attractiveness of individual shops. Membership functions of intangible variables of individual shops are assessed and Pacific Rim Property Research Journal, Vol 17, No 1, 2011 162 then incorporated into the inference engine. An index representing the drawing power of each individual shop is derived accordingly. It is possible that more than one shop would have an impact on a particular measuring point. Hence, three principles are established to select shops with drawing power that would heavily affect the pedestrian flow at the measuring point. These three principles are used in the following order: 1. Select shops with entrances visible from the measuring point. 2. Choose those with entrances within 10m from the measuring point; 3. Select the best three shops – those with the highest drawing power to customers. A measuring point in Festival Walk is chosen to demonstrate this method. The selected point is at LG1 with a pedestrian flow count of 801 shoppers per a 10-minute interval. The nearest three shops fitting all the above mentioned principles are Page One, Festival China and Cour Carre. Page One is a book shop with a lettable area approximately 12,600 square feet. It is internationally reputable and adopts a premium pricing strategy. Combined with its wide range of book selection, its marketing strategy focuses on middle class customers with relatively strong purchasing power. Experts’ assessment extracted from the questionnaire and its membership functions are summarized in Table 1. Table 1: Assessment of membership functions of Page One Linguistic variable Experts’ assessment Input variable Membership function Image 50% IV14 0.94 Goods quality 45% IV24 0.71 Pricing 110% IV34 0.68 Brand name 60% IV44 0.67 Variety 160% IV54 0.56 Trade 30% IV62 0.94 Area 126% IV75 0.38 Location 60% IV82 0.91 Applying the Max-Min operation, the output strength is 0.38 and the combination of assessments complies with Rule 24 – Large Toy/Book/Music Shop Rule. According to the rule, three output variables are OV15, OV24 and OV34. In order to assess the implication of the three output variables, they are defuzzified with the calculation of centroids of the outputs. The defuzzification results of OV1, OV2 and OV3 are 69.7%, 66.4% and 61.5% respectively. It could be interpreted as 69.7% of shoppers aged 15 to 24, 66.4% of shoppers aged 25 to 35; and 61.5% of shoppers aged above 35 are “pulled” towards Page One respectively. According to the experts’ opinions, the Pacific Rim Property Research Journal, Vol 17, No 1, 2011 163 proportion of the three age groups in shopping centres is approximately 1:2:1. A weighted average index for the shop pedestrian is calculated as 0.66. Festive China is a local traditional Chinese restaurant, with a lettable area approximately 7,000s.f. Compared with existing nearby competitors, its pricing strategy is positioned at the middle-mass market, providing quality food and beverages at affordable price levels. Food choice is plentiful, although limited to “Chiu Chow” cuisine. The assessment of the shop variable experts’ assessment and its representative membership functions are shown in Table 2. Table 2: Assessment of membership functions of Festive China Linguistic variable Experts’ assessment Input variable Membership function Image 50% IV14 0.94 Goods quality 45% IV23 0.71 Pricing 90% IV33 0.94 Brand name 30% IV43 0.39 Variety 50% IV52 0.91 Trade 20% IV61 0.71 Area 70% IV74 0.50 Location 60% IV82 0.91 The assessments conform to Rule 2 – the Young Restaurant Rule. The output strength is 0.39. After defuzzification, the results for OV1, OV2 and OV3 are 14.3%, 66.5% and 42.0% respectively. It could be interpreted as 14.3%, 66.5% and 42.0% of shoppers aged 15 to 24, 25 to 35, and above 35 are “pulled” towards Festive China respectively. The pedestrian index at the point is 0.47. Cour Carre is a local fashion chain store with a lettable area approximately 1,630s.f. It targets at middle-income working class. Compared with other chain fashion stores, its goods quality and pricing level are reasonable, even though the choices are relatively limited. The assessment of the shop variable experts’ assessment and its representative membership functions are given in Table 3. Pacific Rim Property Research Journal, Vol 17, No 1, 2011 164 Table 3: Assessment of membership functions of Cour Carre Linguistic variable Experts’ assessment Input variable Membership function Image 30% IV13 1.0 Goods Quality 40% IV24 0.94 Pricing 85% IV33 0.63 Brand Name 35% IV43 0.77 Variety 50% IV52 0.91 Trade 50% IV63 0.83 Area 16% IV72 0.53 Location 60% IV82 0.91 The assessments comply with Rule 11 – the Reasonably Priced Small Popular Fashion Shop Rule. The output strength is 0.53. The defuzzification results of OV1, OV2 and OV3 are 14.9%, 46.0% and 42.8% respectively. It could be interpreted as 14.9%, 46.0% and 42.8% of shoppers aged 15 to 24, 25 to 35, and above 35 are “pulled” towards Cour Carre respectively. The pedestrian index generated by Cour Carre at the point is 0.37. The combined pedestrian index, through summation of three individual indices at the point, is 1.50, with the assumption that the pedestrian drawing power of individual shops is independent from each other. By applying the same methodology, the combined pedestrian indices at other testing points are estimated. VALIDITY Three shopping centres (Festival Walk, Pacific Place and Langham Place) have been selected for testing the validity. Combined Pedestrian Index at each particular point was evaluated with the similar mechanism. They are tested against the pedestrian flow at each point using regression models. It is expected that there will be a significant relationship between shop attractiveness (represented by the index) and the pedestrian flow. The results are given in Tables 4-6. Pacific Rim Property Research Journal, Vol 17, No 1, 2011 165 Table 4: Festival Walk The pedestrian flow at 54 measuring points of Festival Walk has been counted and tested against the combined pedestrian index. The model is statistically significant at the 99% confidence level (i.e. p<0.01) with adjusted R-squared of 0.62. Combined pedestrian index is identified as significant at the 0.01 level. Table 5: Langham Place The pedestrian flow at 44 measuring points of Langham Place has been counted and tested against the combined pedestrian index. The model is statistically significant at the 99% confidence level (i.e. p<0.01) with adjusted R-squared of 0.68. Combined pedestrian index is identified to be significant at the 0.01 level. Dependent Variable: Pedestrian Flow PEDP Sample Size: 53 Variable Coefficient Std. Error t-Statistic Prob. Combined Pedestrian Index 481.872 52.350 9.205 0.000 C 70.009 49.144 1.425 0.160 R-squared 0.624255 Mean dependent var 494.874 Adjusted R-squared 0.616887 S.D. dependent var 198.466 S.E. of regression 122.8426 Akaike info criterion 12.497 Sum squared resid 769605.3 Schwarz criterion 12.571 Log likelihood -329.1623 84.730 F-statistic Dependent Variable: Pedestrian Flow PEDP Sample Size: 44 Variable Coefficient Std. Error t-Statistic Prob. Combined Pedestrian Index 719.553 74.772 9.623 0.000 C -28.275 70.049 -0.404 0.689 R-squared 0.687985 Mean dependent var 578.705 Adjusted R-squared 0.680556 S.D. dependent var 357.647 S.E. of regression 202.1396 Akaike info criterion 13.500 Sum squared resid 1716137 Schwarz criterion 13.581 Log likelihood -295.004 92.609 F-statistic Pacific Rim Property Research Journal, Vol 17, No 1, 2011 166 Table 6: Pacific Place The pedestrian flow at 34 measuring points of Pacific Place has been counted and tested against the combined pedestrian index. The model is statistically significant at the 99% confidence level (i.e. p<0.01) with adjusted R-squared of 0.64. Combined pedestrian index is identified to be significant at the 0.01 level. Results of all three models have demonstrated a significant relationship between the combined pedestrian index and the pedestrian flow. By comparing the fuzzy indices assessed at different points within the same retail development, the shop attractiveness map / pattern would be estimated. A regression model, compiling all of the above data, is set up to test the predictive power of the indices on the actual pedestrian flow at particular points. The dependent variable is the pedestrian flow at particular points (PEDP), while independent variables are the Combined Pedestrian Index (INDEX) and the Centre’s Pedestrian Flow (PEDC). PEDC represents the factor of total pedestrians being drawn under the influence of the overall setting of the shopping centre, while INDEX represents the drawing power of particular points. Dependent Variable Pedestrian Flow PEDP Sample Size: 34 Variable Coefficient Std. Error t-Statistic Prob. Combined Pedestrian Index 477.332 61.693 7.737 0.000 C -153.784 59.166 -2.599 0.014 R-squared 0.651659 Mean dependent var 271.221 Adjusted R-squared 0.640774 S.D. dependent var 213.865 S.E. of regression 128.1809 Akaike info criterion 12.602 Sum squared resid 525770.6 Schwarz criterion 12.692 Log likelihood -212.2303 59.864 F-statistic Pacific Rim Property Research Journal, Vol 17, No 1, 2011 167 Table 7: Pedestrian flow forecast The model is statistically significant at the 99% confidence level (i.e. p<0.01) with adjusted R-squared of 0.65. Two independent variables are identified to be significant at the 0.01 level. The model shows significant predictive power on the actual pedestrian flow within the retail development. CONCLUSION Fuzzy logic has been widely adopted in pedestrian flow simulation models in various scenarios, such as railway stations and stadiums. The established simulation models are mainly designed for emergency purpose only, with very few applications for commercial uses. This paper is the first to adopt fuzzy logic to assess shop attractiveness within a shopping centre. The results are positive and a map of shop attractiveness would be drawn. The assessment of the performance of different tenant mix schemes would be carried out systematically beforehand and does not need to be relied solely on the experience of the leasing manager. This method is, therefore, potentially beneficial to developers of shopping centres. Although positive results have been obtained, there are several limitations: 1. The combined pedestrian index was assessed with the assumption that the pedestrian drawing power of individual shop is independent from each other. The Dependent Variable PEDP Sample Size: 131 Variable Coefficient Std. Error t-Statistic Prob. Centre's Pedestrian Flow: PEDC 0.548254 0.07314 7.495995 0.000 Combined Pedestrian Index INDEX 573.3924 41.49664 13.8178 0.000 C -401.5748 63.36572 -6.337414 0.000 R-squared 0.654326 Mean dependent var 464.9835 Adjusted R-squared 0.648924 S.D. dependent var 290.2337 S.E. of regression 171.9682 Akaike info criterion 13.1551 Sum squared resid 3785353 Schwarz criterion 13.2210 Log likelihood -858.6611 121.1453 F-statistic Pacific Rim Property Research Journal, Vol 17, No 1, 2011 168 combined drawing power might have been overestimated. Further study of integration of shop attractiveness is suggested. 2. Shop attractiveness is not the sole factor affecting the pedestrian flow of a shopping centre. A comprehensive system should also include other significant tangible factors, such as location, size, shape, layout, number of floors and number of shops, etc. To assess the overall performance of shopping centres, both tangible and intangible attributes should be taken into account. Compared to shop attractiveness, other factors are more tangible and would be included into a regression-expert system. The system will be desirable for real estate investors, developers, shop tenants and designers, which will allow them to modify the alternative design and tenant mix schemes in the early development stage, without going through the pain of the trail and error stage. REFERENCES Areni, C. and D. Kim (1995), “The Influence of In-Store Lighting on Consumers’ Examination of Merchandise in a Wine Store”, International Journal of Research in Marketing, Vol. 11(4): 117-125 Bagnoli, C. and Smith, H.C. (1997) “Fuzzy Logic: The New Paradigm for Decision Making”, Real Estate Issues, Vol. 22(2): 35-41 Bagnoli, C. and Smith, H.C. (1998), “The Theory of Fuzzy Logic and its Application to Real Estate Valuation”, Journal of Real Estate Research, Vol. 16: 169-99. Brown, S. (1992), Retail location: a micro-scale perspective, Aldershot: Avebury. Brown, S. (1992), “Tenant Mix, Tenant Placement and Shopper Behaviour in a Planned Shopping Centre”, The Service Industries Journal, Vol. 12(3): 384-402. Dawson, J.A. (1983), Shopping Centre Development, London: Longman. Dennis, C., A. Newman, and D. Marshland (2005), Objects of Desire-Consumer Behaviour in Shopping Centre Choices, New York: Palgrave MacMillian. Engel, J.F., Blackwell, R.D. and Miniard, P.W. (1986), Consumer Behaviour, 5th edn. Chicago: Dryden Press. G. W. Nurcahyo, S. Shamsuddin, R. Alias and M. Noor Md. Sap (2003), “Selection of Defuzzification Method to Obtain Crisp Value for Representing Uncertain Data in a Modified Sweep Algorithm”, JCS&T, Vol. 3(2) : 22-28. Pacific Rim Property Research Journal, Vol 17, No 1, 2011 169 Herrington, J. and L. Capella (1996), “Effects of Music in Service Environments: A Field Study”, Journal of Services Marketing, Vol. 10(2): 26-41. Jantzen, J. (1999), “Tutorial on Fuzzy Logic”, http://www.cs.baskent.edu.tr/ ~muluer/educational/ai/JJ_fuzzy_tut.pdf. Martineau, P. (1958), “The Personality of the Retail Store”, Harvard Business Review, Vol. 36: 47-55. Michon, R., Chebat, J.C. and Turley, L.W. (2005), “Mall Atmospherics: the Interaction Effects of the Mall Environment on Shopping Behavior”, Journal of Business Research, Vol. 58: 576-583. Ng, S., Luu, D.T., Chen, S.E. and Lam, K.C. (2002), “Fuzzy Membership Functions of Procurement Selection Criteria” Construction Management and Economics, Vol. 20: 285-296. Northen, R.I. (1984), Shopping Centre Development Reading: Centre for Advanced Land Use Studies, College of Estate Management. Sit, J., Merrilees, B. and Birch, D. (2003), “Entertainment-seeking Shopping Centre Patrons: the Missing Segments”, International Journal of Retail & Distribution Management, Vol. 31(2): 80-94. Smith, J. (1985), “Pedestrianisation – Shopping Streets in Scotland”, The Planner, May:12-16. Steenkamp, J. and Wedel, M. (1991), “Segmenting Retail Markets on Store Image Using a Consumer-Based Methodology”, Journal of Retailing, Vol. 67(3): 300-320. Wong, K.C. and So, T.P. (1995), “A Fuzzy Expert System for Contract Decision Making”, Construction Management and Economics, Vol. 13: 95-103. http://www.cs.baskent.edu.tr/%20~muluer/educational/ai/JJ_fuzzy_tut.pdf� http://www.cs.baskent.edu.tr/%20~muluer/educational/ai/JJ_fuzzy_tut.pdf� Pacific Rim Property Research Journal, Vol 17, No 1, 2011 170 Appendix A: Individual shop attractiveness assessment Individual Shop Attractiveness Assessment Q1. What are the best figures to describe the following five catalogues of shop youth image? Please indicate your responses in the following boxes. ** Teenagers % Youngster % Young % Adult % Mature % Q2. What are the best figures to describe the following six catalogues of goods quality? Please indicate your responses in the following boxes. ** Quality is defined as [Material quality + workmanship quality + design concept quality] x 100% Material quality = Workmanship quality = where Design concept quality = Bad % Just Okay % Reasonable % Above average % Good % Prestige % Q3. Please indicate your responses in the following boxes. ** Very cheap % Cheap % Okay % Reasonable % Expensive % Luxury % amount of money paid for extra material for better quality than the norm / total material price for the norm x 100% amount of money paid for extra design concept effort for better quality than the norm / total price for design concept effort for the norm x 100% Pricing competitiveness is defined as [Normal anticipated pricing in the shop / average market price expectation of the same type of goods] x 100% Image is defined as the % of mature goods (shopper's age > 35 may be interested on that goods) found in the shop What are the best figures to describe the following six catalogues of shop pricing competitiveness? amount of money paid for extra workmanship for better quality than the norm / total workmanship price for the norm x 100% Pacific Rim Property Research Journal, Vol 17, No 1, 2011 171 Q4. What are the best figures to describe the following five catalogues of shop branding? Please indicate your responses in the following boxes. ** Rarely heard % Local % Local popular % International % Prestige % Q5. What are the best figures to describe the following four catalogues of goods variety in the shop? Please indicate your responses in the following boxes. ** Limited choices % Average % Plenty % Abundant % Q6. What are the best figures to describe the attractiveness of the following trades? Please indicate your responses in the following boxes. ** % % % % Supermarket/store/retail services (Trade 5) % Q7. What are the best figures to describe the following five catalogues of shop size? Please indicate your responses in the following boxes. ** Size is defined as [Shop size in shopper impression / 10,000sf] Tiny % Small % Normal % Large % Very large % Q7. Please indicate your responses in the following boxes. ** Very inconvenient % Inconvenient % Normal % Convenient % Very inconvenient % Location is defined as [Travel time from the main entrance to the shop / travel time from the main entrance to the most distant shop] Restaurant/entertainment (Trade 1) Book/music/furniture/toy (Trade 2) Fashion/ sports (Trade 3) Beauty/ Jewelry /AV shops (Trade 4) Attractiveness of trade is defined as the [No. of visitor of the shop/ Total no. of visitors of shopping centre] x 100% What are the best figures to describe the following five catalogues of shop location in term of convenience? Branding is defined as [Proportion of shoppers knowing the brand name / Proportion of shoppers knowing the most famous brand name] x 100% Variety is defined as [Normal anticipation of the choices of goods in the shop / Normal anticipation of choices of goods in typical shops] x 100% Pacific Rim Property Research Journal, Vol 17, No 1, 2011 172 Appendix B: Input and Output Variables of the Fuzzy Expert System Input Variable Description Term IV1 – Shop youth image Shoppers’ general impression on the youth image of the shop Represented formula: % of mature goods (i.e. age of shopper > 35 may be interested) found in the shop IV11 – teenagers IV12 – youngster IV13 – young IV14 – adult IV15 – mature IV2 – Goods quality Shoppers’ general impression on the quality of the goods sold - Represented formula1 (Material quality + workmanship quality + design concept quality), where: : Material quality = (amount paid for extra materials for better quality than the norm/ total material price for the norm) x 100% Workmanship quality = (amount paid for extra workmanship for better quality than the norm/ total workmanship price for the norm) x 100% Design concept quality = (amount paid for extra design concept quality for better quality than the norm/ total price for design concept effort for the norm) x 100% IV21 – bad IV22 – Just okay IV23 – reasonable IV24 – above average IV25 – good IV26 – prestige IV3 – Pricing competitiven ess Shoppers’ general impression on pricing competitiveness of goods sold Represented formula: (Normal anticipated pricing in the shop / average market price expectation of the same type goods) x 100% IV31 – very cheap IV32 – cheap IV33 – okay IV34 – reasonable IV35 – expensive IV36 – luxury IV4 – Branding General impression from shoppers on the brand name of the shop Represented formula: (Proportion of shoppers knowing the brand name / proportion of shoppers knowing the most famous brand IV41 – rarely heard IV42 – local unpopular IV43 – local popular 1 With reference to the methodology suggested by Ng et al. (2002) Pacific Rim Property Research Journal, Vol 17, No 1, 2011 173 name) x 100% IV44 – international IV45 – prestige IV5 – Goods Variety General impression from the shoppers on the variety of the goods sold in the shop Represented formula: (Normal anticipation of the choice of goods of the shop/ Normal anticipation of choice of goods in typical shop) x 100% IV51 – limited IV52 – average IV53 – plenty IV54 – abundant IV6 – Trading The drawing power on shoppers by the nature of the shop trade trade 1 – restaurant / entertainment; trade 2 –books, gift, music and similar goods; trade 3 –fashion, sports and similar goods; trade 4 –beauty cosmetic, jewelry, audio & video, & other luxuries trade 5 – supermarket Represented formula: (No. of visitor of the shop/ Total no. of visitors of shopping centre) x 100% IV61 – trade 1 IV62 – trade 2 IV63 – trade 3 IV64 – trade 4 IV65 – trade 5 IV7 – Shop size Shoppers’ impression on shop size Represented formula: (Shop size in shopper impression / 10,000sf) x 100% IV71 – tiny IV72 – small IV73 – normal IV74 – large IV75 – very large IV8 – Location Shopper impression on the convenience of shop location Represented formula: (Travel time from main entrance/ Travel time to visit the most distant shop) x 100% IV81 – very inconvenient IV82 – inconvenient IV83 – normal IV84 – convenient IV85 – very convenient Output Variable Description Term OV1 How many shoppers in the age group between 15 and 24 are attracted to the shop Represented formulas: Number of shoppers (for the age group between 15 and 24) visiting the shop/Total number of shoppers visiting the shopping centre OV11 – very low OV12 – low OV13 – medium OV14 – High OV15 – Very high OV2 How many shoppers in the age group between 25 and 34 are attracted to the shop Represented formulas: Number of shoppers (for the age group between 25 and 34) visiting the shop/ Total number of shoppers visiting the shopping centre OV21 – very low OV22 – low OV23 – medium OV24 – High OV25 – Very high Pacific Rim Property Research Journal, Vol 17, No 1, 2011 174 OV3 How many shoppers in the age group above 35 are attracted to the shop Represented formulas: Number of shoppers (for the age group above 35) visiting the shop/ Total number of shoppers visiting the shopping centre OV31 – very low OV32 – low OV33 – medium OV34 – High OV35 – Very high Pacific Rim Property Research Journal, Vol 17, No 1, 2011 175 Appendix C: Statistical summary of expert opinion on intangible variables Intangible Variable Term Minimum Mean Maximu m 1. Shop youth image IV11 – teenagers 0% 5% 15% IV12 – youngster 5% 12% 25% IV13 – young 15% 30% 40% IV14 – adult 30% 49% 65% IV15 – mature 65% 76% 85% 2. Goods quality IV21 – bad 0% 6% 15% IV22 – Just okay 5% 15% 25% IV23 – reasonable 20% 32% 40% IV24– above average 25% 41% 55% IV25 – good 40% 53% 65% IV26 – prestige 60% 77% 90% 3. Pricing titi IV31 – very cheap 20% 36% 50% IV32 – cheap 50% 69% 80% IV33 – okay 75% 91% 110% IV34 – reasonable 80% 103% 125% IV35 – expensive 125% 140% 160% IV36 – luxury 150% 170% 200% 4. Branding IV41 – unknown brand 0% 3% 10% IV42 – local unpopular 5% 11% 30% IV43 – local popular 25% 38% 55% IV44 – international 45% 55% 70% IV45 – prestige 65% 81% 95% 5. Goods Variety IV51 – limited 15% 35% 50% IV52 – average 30% 52% 80% IV53 – plenty 50% 75% 120% IV54 – abundant 120% 144% 180% 6. Trading IV61 – trade 1 5% 16% 30% IV62 – trade 2 15% 29% 45% IV63 – trade 3 35% 47% 65% IV64 – trade 4 50% 68% 80% IV65 – trade 5 50% 77% 100% 7. Shop size IV71 – tiny 1% 4% 10% IV72 – small 5% 13% 20% IV73 – normal 15% 26% 40% IV74 – large 35% 60% 80% IV75 – very large 70% 86% 150% Pacific Rim Property Research Journal, Vol 17, No 1, 2011 176 8. Location IV81 – very inconvenient 65% 84% 100% IV82 – inconvenient 50% 61% 75% IV83 – normal 28% 36% 50% IV84 – convenient 10% 17% 32% IV85 – very convenient 5% 9% 15% Intangible Variable Term Minimum Mean Maximum 1. Output Variable I OV11 – very low 5% 14% 25% (Age group 1) OV12 – low 10% 24% 40% OV13 – medium 30% 44% 55% OV14 – high 45% 60% 75% OV15 – very high 50% 74% 90% 2. Output Variable II OV21 – very low 10% 18% 25% (Age group 2) OV22 – low 15% 25% 35% OV23 – medium 30% 49% 60% OV24 –high 50% 67% 85% OV25 –very high 60% 73% 80% 3. Output Variable III OV31 – very low 5% 14% 20% (Age group 3) OV32 – low 15% 23% 35% OV33 – medium 30% 43% 55% OV34 –high 50% 61% 75% OV35 –very high 50% 71% 85% Pacific Rim Property Research Journal, Vol 17, No 1, 2011 177 Appendix D: Rule base – 30 rules