A Model of Consumer Decision Making for a Mud Based Game

A networked, multi-player, simulation-based, interactive multi-media, educational environment (the SELL game) was implemented to teach the principles of micro-economics, retailing and targeting. To effectively teach these concepts, it is necessary that an "authentic" economic simulation be provided, in order to preserve the player's interest over a period of several weeks. This paper describes the economic model underlying the simulation -in particular stressing the elements of simplicity and plausibility that makes it succeed. Introduction SELL is a multi-user, educational, economic game implemented in LambdaMoo, which is an object-oriented MUD1. The SELL game economic model is designed to simulate consumer decision making. In SELL players run small retail stores located in Springfield, a virtual city of around 100,000 residents. Springfield is divided into eight neighborhoods, each with its own social and economic flavor. Each store is located in one of these neighborhoods and each player must be aware of the impact the local neighborhood has on their store in order to succeed. SELL is a turn-based game where each turn players set their prices, service level, inventory levels, and advertising. These decisions, for all players, are then sent to the main server. The server then computes each store's sales based upon all players decisions and the interests, wants and location of consumers of Springfield. For players of SELL there are no absolutely right strategies which will work all the time. Rather the returns of any action will depend on the actions of competitors and interests and desires of the consumers in Springfield. Since different consumers live in different neighborhoods, one action which will succeed in one area may fail in another. To be successful the player must know who are their potential customers, where they live, how far they can travel to shop, what they are interested in buying, how much service they want, how much they are willing or able to spend, and how strongly they are attracted to high quality or low cost items. Also the player must know how to reach potential customers through advertising. Educational Goals The goal of this project is to teach a wide set of skills associated with running a retail business by allowing the student to run a simulated store in a simulated economy. In order to survive in SELL's open competition the player must set a competetive price, hire the appropriate staff, and select appealing products while always keeping an eye on the competition. The player must think about who their likely customers are, what goods these customers want, and how they can use their advertising dollar most efficiently to reach these customers. The strategy of carefully selecting consumer groups is often called targeting by marketing and advertising experts, and targeting is the key teaching goal of SELL. An educational game should be both fun and informative. Players should acquire concepts and skills as a consequence of playing the game, and this learning should transfer to contexts outside the game. The challenge then is to construct a game of sufficiently interesting complexity that is true to its premise. When the player acts in the simulated environment, the environment 1MUD stands for Multi-User Domain (although originally the "D" was for Dungeon). must re-act in consistent and plausible ways. Without this veracity the game will fail the ultimate test: the players will not play it. Figure 1: the map of Springfield Game Play There are a number of choices each player must make during the game. The player must first decide where to locate. Then, each turn, the player must decide what to sell, how much to stock, who to buy from, and what price to set. They must decide what service level to offer and choose from a number of advertising options. To support the making of these decisions SELL provides a number of tools. SELL provides financial tools, which tell the player their inventory, assets, liability, expenditures, and profits. SELL also provides tool which give the user important information concerning competitors, including their inventory, prices, profits, liabilities, and staffing. Sell provides a newspaper where players can learn general information about Springfield's business environment as well as reading other players print advertising. Players have a businessto-business yellow pages which allow them to contact wholesalers. Each player has a radio with which they can "hear" other stores radio advertising. The player can also get important information by selecting questions to ask their own staff. SELL provides a map of Springfield (figure 1), showing all the neighborhoods, other player's stores, and the Springfield Bank. Double clicking on an area provides important information about the character of that area. Double clicking on another player's store takes one to that store (figure 2), where the player can ask questions of the staff, check prices and inventory, or engage in an on-line conversation with the other player directly. Double clicking on the bank takes the player there, where they can apply for a loan. A demographic information tool is provided so users can query all important information about the general types of consumers a in Springfield and their preferences. For example a player can see which are the largest consumer groups in a certain area. They can see what products these groups are most interested in, what media they are most exposed to, how much they value low prices or high service. With this SELL promotes learning about targeting. At the end of each turn the user can also see which consumers shopped at their store and what they bought. For this paper we look at two example stores: Tom's Bike located in Gentry Village (described below), and Bob's located in the neighboring Downtown area. For this example these are the only two stores in all of Springfield. Tom's sets his service to a low level (service level=1) in order to keep costs low, while Bob's offers high service (service level=3) in order to attract wealthier clients. Figure 2: a store in Springfield The Model of Consumer Decision Making Realistic simulations are computationally complex and therefore computationally expensive. This complexity is desirable if players are to believe the simulation, but undesirable if the game is to be played in real time. A balance between complexity and efficiency must be found or the game will be either too simple or too slow. The SELL game comes close to this center point; the simulation is easy to maintain, easy to expand, and efficient while still being sufficiently complex the hold the players interest. Rather than attemping to model economic behavior solely with complex mathematical formulae, SELL uses psychographic characterizations manipulated by simple arithmetic calculations combined with a detailed though simple representation of consumers and products. This simplicity facilitates the implementation of products and consumer groups, and makes it fairly easy to add and modify products. Further, this simplicity and modularity allows the rapid creation of new scenarios. Assumptions The SEll economic model assumes rational, cost-minimizing consumers. Therefore, consumers consider travel costs, search costs, service benefits, and product quality as well as price when making buying decisions. SELL assumes that advertising increases sales by reducing the search cost to consumers in finding information about desired products. In this way advertising helps consumers find the best value on goods they already want. Not all economists are so positive about advertising, many assert that advertising increases a companies sales by misguiding consumers into paying more then they need to or instilling in the consumers desires that they would not normally have (Galbraith 1961, Thomas 1981) Sell assumes that players do not have access to a budget sufficient to afford the kind of advertising that would be needed to change consumers preferences or create artificial desires (i.e. no player can "create" desire, their means are too limited). Sell models the entire consumer population by defining it in terms of cluster groups (CG). The concept of cluster groups is similar to the idea of psychographic segmentation, employed by many advertisers and marketers. Psychographic segmentation is the classification a of population into groups which share similar values, attitudes, and lifestyles (Rice, 1988; Piirto, 1990). The premise is that persons with similar values and lifestyles will have similar buying behavior. Psychographic segmentation is a growing method in marketing, for it promises insight into the emotional and lifestyle factors which motivate consumer's buying behavior. Representations To simulate consumer decision making SELL uses an explicit representation of the population of Springfield along with their values and preferences for certain products. It is the representation of the consumers which drives the model. Since the main goal of this project is to teach marketing skills it was decided to use a form of psychographics to represent the population of Springfield. This virtual city is thus represented with cluster groups. A cluster group is a coherent segment of the population where the members are alike in age, income, life narrative, interests, values, and lifestyles. As a consequence the members of a cluster group have similar consumer behavior. In SELL there are 20 distinct cluster groups. The following are three cluster groups used in the SELL consumer model, cg1 Working Class: middle aged couples with children, renters cg2 White Collar Singles: young singles, renters, upwardly mobile cg3 College students. Table 1: Sample cluster groups To model a specific neighboorhood's total population a population is given to each of the 20 cluster groups for that area. There are eight neighborhoods which comprise the city of Springfield. Each of these areas has its own particular flavor, comprised of very different populations mixes. For example there is University Heights, mostly made up of college students and professors, or Gentry Village home to most of Springfield's young professionals. The population for the three example cluster groups in Gentry Village is: cg1 Working Class 1255 persons cg2 White Collar Singles 4123 persons cg3 College Students 1185 persons Table 2: Sample cluster groups as population of Gentry Village Products Sell uses a two-level representation of products: product classes and models. SELL has 30 models in 7 product classes. A model represents a particular good for sale, while a product class is used in this implementation to describe the market for a entire class of goods. For example the bike model "Road King 1000" is a adult bicycle possessing certain features and quality level. While the product class "adult bicycles" contains data about the market for all adult bikes. All models must belong to one and only one product class. All models in the same product class compete with each other while no models in different product classes compete. This certainly is an over simplification, since almost any product can compete with any other product for the consumers dollar. For the purposes of this game though the convention made running the simulation much more efficient. Also, even though all products can possibly compete with each other in the real world, most competetion goes on between very similiar products rather then disparate products. Cluster Group Name I USB UQB UTC USC feature/ Road feature/ Race PS MIN MAX cg1 Working Class 60% $1 $80 $25 $25 $0 $0 130% $80 $400 cg2 White Collar Sngls 180% $10 $120 $10 $20 $0 $50 90% $200 $800 cg3 College Students 210% $8 $100 $30 $30 $30 $0 100% $100 $600 Table 3: The product class definition for Adult Bikes Product Class Definition The product class definition is the heart of the SELL simulation. It contains all information on consumers likes and dislikes for a given product. SELL uses the explicit data about consumer preferences stored in the product class definition to generate consumer behavior. Based on the information stored in the product class representation consumers decide how much of a product they want to buy, which stores they buy from and how much they can afford to spend. By changing the data in the product class definition one changes the consumer behavior generated by the simulation (see Table 3). A product class definition contains firstly a number for the average potential demand (APD), which is the average number of persons who will be want to buy the product over a given time. All other information is stored in a table, with all the cluster groups on one axis and all the features and consumer values on the other. Each column is a dimension of the consumers decision making. The values underneath represent the value of the cluster groups. In the product class definition each cluster group has a percentage number for their relative level of interest compared to the overall average potential demand (I). So cluster groups with high interest in a product class will have a percentage number over 100% while those with low values will be below 100%. These numbers were based loosely upon market research information obtained from Simmons demographic data (Simmons, 1993). For each cluster group a dollar amount is given for unit search cost (USC). The unit search cost is the cost for a group to gain information about a particular store per unit of distance between store and consumer. The greater the distance between store and buyer that more expensive it is for a consumer to learn about what the store is selling and at what prices. This unit search cost is used to simulate advertising. Unit transportation cost (UTC) is the unit cost to each cluster group of getting to a store and getting the product back home. Some cluster groups are more mobile then other and therefore have lower transportation costs. For example retired persons have very high transportation cost, while families with homes, who own cars, have much lower transportation costs. The product class definition also contains data about benefits that products or stores may offer. The product class definition gives values for unit service benefit (USB) and unit quality benefit (UQB). The unit service benefit is a dollar amount a cluster group is willing to pay for a unit of added service. The unit quality benefit is the dollar amount a cluster group is willing to pay for a unit of added quality. Both of these benefits will be subtracted from the price of a model in calculating the overall consumer's real cost. Also contained in the product class definition is an enumeration of all the possible features a model of this class may have. For example a bicycle may be a racing bike or a road bike. In SELL televisons have the features of being color or black and white, of being 12", 17" or 21", of being monphonnic, stereo, or sound-around. Models have the features enumerated in their product class definition. Different features are valued differently by different consumer groups. For each feature enumerated, the product class definition contains each cluster group's value for that feature in dollars. This is the amount a cluster group will pay for that feature. Some features may be so disliked by a particular cluster group that they have a negative value for them, which means a players would need to offer a lower price than for a product without the feature, in order to make them want to buy it. The product class definition also contains data about cluster groups sensitivity to price in the form of a percentage number called price sensitivity (PS). This percentage tells how sensitive a group is compared to an average price sensitivity, where 100% means that a cluster group has average price sensitivity, and a lower number means that the cluster group is less concerned about getting the best buy . Finally, product class definitions also contain two numbers for each cluster group which represent how much they can afford to spend: a maximum dollar amount (MAX) which is the price at which no one in the cluster group can afford to buy the product and a minimum (MIN) which is the price at which everyone in the cluster group can afford to buy the good. All these values will be used to calculate consumers' decision making. Models A model represents a particular item that player's sell --it is an instantiation of the abstract product class. Each model has a pointer to its parent product class definition, along with its name, feature settings, manufacturers suggested retail price (MSRP), and an iconic picture. All models in the same product class class compete against each other, but not against the models in another product class. For example, high quality adult bikes compete against more affordable bikes, while neither competes against children's bikes. Three models created for the SELL game are shown in Table 4. Notice, for example, from the feature/Race column of Table 3, that white collar singles are willing to pay an extra $50 for a racing bike like the Euro-Elite in Table 4. Model Name Road King Street Bike Super-Elite Euro-Elite Model Slogan "Your Basic Street Bike" "The King of Street Bikes" "The Best Racing Bike You Can Buy" MSRP $310 $600 $700 Quality Quality = Basic (= 0) Quality = High (=3) Quality = High (= 3) Feature Kind = road Kind = road Kind = race Table 4: Three Models of the Adult Bikes Product Class Tom's bike store, hoping to attract bargain hunters sells only the Road King Street Bike for $310, while Bob's tries to attract up-scale clients with the Super-Elite, selling for $600, and the Euro-Elite selling for $700. Advertising SELL models the effect of advertising by first determining the search cost, to a consumer group, of finding out if a particular store sold a desired item and at what price they sold it --if that store did not advertise. This is found by multiplying the cluster group's unit search cost (USC), found in the product class definition (Table 3), by the unit distance between store's nighboorhood and consumer's nighboorhood. Advertising reduces that search cost by some percentage. The amount an ad reduces this search cost is called the percentage search cost reduction (RSC). Different advertising reaches different consumers. SELL models an advertisement by giving, for each cluster group, a percentage number for the reduction in search cost. Different advertisements in different media and different sizes have different percentage numbers (Table 5) Cluster Group Name Leisure Section, Quarter Page, RSC Sports Section, Quarter Page, RSC cg1 Working Class 8% 5% cg2 White Collar Singles 15% 10% cg3 College Students 5% 20%. Table 5: Two Quarter Page Print As in the Leisure and Sports Sections Bob's takes out a quarter page advertisement in the Leisure section while Tom takes out a quarter page advertisement in the sports section. For the working class cluster group, a quarter page advertisement in the leisure section reduces search cost 8% while one in the sports section reduces search cost 5%. A quarter page ad in the sports section reduces college students search cost 20%, while one in the leisure section only 5%. The SELL game supports 8 forms of print media and 3 radio stations. Algorithm Each turn the virtual consumers decide what they want to buy, how much and from whom. The model of consumer decision making has four main steps. First, each group of customers decides how much of a particular product they want to buy; second, they decide which stores they like best; third, the real costs are calculated; and fourth,they decide how much they can afford to buy. After all the players decisions about hiring, pricing, purchases, and advertising have been submitted, the server determines the market share for each store. First the server determines the level of consumer interest in each product. Then it ranks each store's products according to over all value to each cluster group. Finally, it distributes sales based on relative value ranking and then checks how much can actually be afforded. Step 1. The level of consumer interest in a product. First, each cluster groups potential demand (PD) in each product is determined. The server obtains this by looking at the population representation and the product type definition. A particular cluster group's level of interest is calculated by taking the population in each cluster group and multiply it by percentage potential market of that group for that product. The formula for potential demand is: PD= P * I * ADM where: PD = potential demand P = population I = demand index ADM = average demand. This is repeated for each cluster groups in an area, and and then for all areas in Springfield. For example the potential weekly demand for adult bikes for three example cluster groups living in Gentry Village are shown in Table 6: Cluster Group Name Potential Demand (PD) cg1 Working Class 23 cg2 White Collar Singles 223 cg3 College Students 75 Table 6: Potential Demand for the Adult Bicycle Product Class That is, there are 223 White Collar Singles (cg2) in Gentry Village who want to buy bikes every week plus 75 college students and 23 working class families. Step 2. Real cost to cluster groups is calculated. Next each cluster group's real cost is calculated for all products for sale in a product type. Real cost is calculated with the following formula: RC = P + TC + SC (SB + FB). where RC = real cost P = price TC = transport cost SC = search cost SB = service benefit FB = feature benefit. The transport cost is calculated by: TC = 2 * UTC * D(S,C); where TC = transportation cost UTC = unit transport cost for the cluster group D(S,C) = distance from store to cluster group neighborhood. In Table 3 the UTC for white collar singles is $10 --this represents the notion that college students have less personal resources and less cars then many other cluster groups and therefore view transportation as more costly. Thus college students are much more likely to shop in their own areas. The search cost are calculated: SC = RSC *(USC * D(S,C) ^2) where SC = search cost to cluster group RSC = % reduction in search cost from ads by store USC = unit search cost to cluster group D(S,C) = distance between store and cluster group. The service benefit is calculated as