UNIVERSITY OF ILLINOIS LIBRARY AT URBANA-CHAMPAIGN BOOKSTACKS CENTRAL CIRCULATION BOOKSTACKS The person charging this material is re- sponsible for its renewal or its return to the library from which it was borrowed on or before the Latest Date stamped below. You may be charged a minimum fee of $75.00 for each lost book. Theft, mutilatloiv and underlining of books are reasons for disciplinary action and may result In dismissal from the University. TO RENEW CALL TELEPHONE CENTER, 333-8400 UNIVERSITY OF ILLINOIS LIBRARY AT URBANA-CHAMPAIGN JUL 2 8 1998 When renewing by phone, write new due date below previous due date. L162 Digitized by the Internet Archive in 2011 with funding from University of Illinois Urbana-Champaign http://www.archive.org/details/diffstratanalyti1251sudh BEBR FACULTY WORKING PAPER NO. 1251 DIFFSTRAT: An Analytical Procedure for Generating Optimal New Product Concepts for a Differentiated-type Strategy D. Sudharshan Jerrold H. May Thomas Gruca College of Commerce and Business Administration Bureau of Economic and Business Research University of Illinois, Urbana-Champaign BEBR FACULTY WORKING PAPER NO. 1251 College of Commerce and Business Administration University of Illinois at Urbana-Champaign May 1986 DIFFSTRAT: An Analytical Procedure for Generating Optimal New Product Concepts for a Dif ferentiated-type Strategy D. Sudharshan, Assistant Professor Department of Business Administration Jerrold H. May University of Pittsburgh Thomas Gruca, Assistant Department of Business Administration ABSTRACT Firms are likely to consider employing multiple products in markets where competition and heterogeneity of consumer preferences suggests the use of market segmentation. The problem of identifying the optimal multiple new product concepts for such a market may be considered in either a sequential (one product at a time) or a simul- taneous fashion. We show that the "optimal" new product concepts generated sequentially are different from, and generally inferior to, those generated simultaneously. In this paper we present a new proce- dure, DIFFSTRAT, which solves this problem for a wide range of con- sumer preference models. This procedure utilizes a fundamental insight called the augmented space approach to solve this problem. INTRODUCTION Having decided upon the market it is going to enter, a firm has three broad market coverage strategies to choose from: (i) undifferen- tiated marketing, (ii) differentiated marketing, and (iii) concentrated marketing (Kotler, 1984, p. 267). Under the first strategy, the firm would ignore market segment differences and offer just one product to the entire market. In contrast, under concentrated marketing, the firm would choose one and only one segment of the market and would provide one market offering specifically aimed at this segment. Although "some brands have skillfully built up reputations of being suitable for a wide variety of people," Gardner and Levy (1955) state that: "In most areas audience groupings will differ, if only because there are deviants who refuse to consume the same way other people do.... It is not easy for a brand to appeal to stable lower-middle-class people and at the same time to be interesting to sophisticated, intellectual upper-middle-class buyers.... It is rarely possible for a product or brand to be all things to all people." In this type of market, the recommendation is to consider multiple product offerings, with each offering being aimed exclusively at one or a few segments. Such a strategy would be of the differentiated marketing type. The market place is replete with examples of firms following the differentiated marketing strategy. An example from Kotler (1984) is Edison Brothers in the retail shoe business. It operates nine hundred shoe stores that fall into four chain categories each targeted toward a different market segment. They are as follows: -2- Segment Chain name Higher-priced shoes Chandler's, Moderate-priced shoes Baker's, Budget shoes Burt's, and Stylized shoes Uild Pair. General Motors, Ford, Chrysler, Toyota, Nissan, etc., in the car industry all seem to follow a similar strategy. Procter and Gamble is a well known user of such strategies in various markets. We believe that the availability of sophisticated consumer need- based methods for generating optimal new product concepts would be useful to a manager pursuing a differentiated marketing strategy. Such methods would consider the problem of offering multiple products to a market. Each product would be targeted at either a segment or a small number of segments. In this paper we report on a procedure (DIFFSTRAT — for differentiated strategy) that has been developed exclusively for this purpose. The problem is formulated as being one of determining that combination of new product concepts in an at- tribute space that would make for an optimal differentiated strategy. Realizing that a firm using DIFFSTRAT might already have some existing products in the same market, we also provide for the following: i) consideration of cannibalization — by considering the net incremental demand of the new and existing set of products over the existing set, ii) repositioning of existing products along with new product introductions to achieve the highest incremental demand, -3- and iii) the possibility of deleting some or all existing products of the firm in considering new products for this market. Interest in analytical models/procedures for generating optimal new product concepts based on an analysis of consumer needs has been growing. This is evidenced by the recent literature reporting procedures/algorithms for new product concept generation (viz. Shocker and Srinivasan (1974, 1979); Albers and Brockhoff (1977, 1980), Albers (1976, 1979, 1982); Green, Carroll and Goldberg (1981); Zufryden (1979); Gavish, Horsky and Srikanth (1983); May, Shocker and Sudharshan (1982, 1983); Sudharshan, May and Shocker (1986); and Sudharshan (1982)). Albers' (1976) procedure SILOP was specifically developed for the single choice ideal point model (i.e., each segment is associated with an ideal point model of preference in an attribute space, and that segment's entire sales potential is assigned to the product closest (in terms of Euclidean distance) to its estimated ideal point). Pessemier, Burger, Teach, and Tigert (1971); Shocker and Srinivasan (1974, 1979); and May, Shocker and Sudharshan (1982) argued that the probabilistic choice models (each segment's share is allocated probabilistically to several products, not just one) are more realistic (i) in the case of frequently purchased consumer non- durables, and (ii) when preference models are specified at the con- sumer segment level. In the context of single optimal new product concept generation problems, there is evidence (May, Shocker, Sudharshan (1982), -4- Sudharshan (1982, 1984), Sudharshan, May and Shocker (1986)) that the use of procedures which assume single choice consumer models may be inefficient in market environments in which probabilistic choice models are better predictors of aggregate choice. Therefore, there is a need to develop a procedure permitting probabilistic consumer choice models for the problem of generating simultaneously optimal multiple new product concepts. Winter (1979) reports a procedure for determining the optimal com- bination of marketing mixes and segments, based on cost-benefit con- siderations. However, his procedure assumes that there is a predeter- mined set of marketing mixes to choose from. If we can a priori specify a set of products and then want to decide on the product- segment combinations that are optimal, this procedure would be appro- priate. If, on the other hand, our task is to find an optimal set of product concepts for a given market, then we need a procedure for searching through the product-attribute space, rather than one for evaluating a finite set of alternatives. Management may not be able to launch multiple new product concepts simultaneously. Financial and managerial resources may not be ade- quate. But such constraints do not rule out multiple new product entry phased in over time. We would argue, that in certain cases, management would be better off choosing that combination of new products that, if introduced, would provide the highest incremental revenue over existing products. In particular, if the rest of the -5- market structure is expected to be stable and/or market signals may be used to reserve product positions, tben such a consideration is perhaps warranted. The alternative is, of course, to choose the best single product now; then, at a later date, consider the best new product in the market which now has one more existing brand (the previously introduced new one), and so on. By choosing the optimal multiple new products simultaneously, management chooses the set that is expected to be the best performer and will be better able to plan new product entries. If they decide to introduce new products sequentially, knowledge of impending entries will enable (i) planned development of a new products program for an entire market, rather than on a new product by new product basis, (ii) sharper positioning of the product being introduced currently, (iii) reservation of positions for impending entries by signalling both competitors and distribution channels, if management so desires , (iv) gradual repositioning of existing product(s) as suggested by the analysis. The following example provides a comparison of the positions and expected performances between simultaneously and sequentially gener- ated multiple new products. In Exhibit 1, we consider five segments, SI, S2, S3, S4 , and S5 , with sales potentials of 80, 20, 40, 60, and 30 units respectively. Assume an ideal-point model (Srinivasan and Shocker (1973), Shocker and Srinivasan (1974, 1979); May, Shocker and Sudharshan (1982)), -6- T * ! * - t NP2 K/ PA ' 1 k O NP2. = &o ts/P3 -13o N Pk z100 /AftiS bwta Exhibit 1: Simple Example Showing Differences Between Simultaneous Consideration and Sequential Consideration -7- i.e., a consumer segment's preference for a product is inversely pro- portional to that product's weighted Euclidean distance (in perceptual space) from the segment's ideal point. Both the weights for each segment and the ideal points are estimated using a procedure such as LINMAP (Srinivasan and Shocker (1973)). VJe further assume that a segment will choose the product that is closest to its ideal point, i.e., the single choice model. It should also be noted that in the probabi- listic choice case the differences between simultaneous and sequential solutions are expected to be greater. We assume a decision space of two product attributes for this example. The contours drawn around the five segment's ideal points (in Exhibit 1) represent, respectively, their individual closest iso- preference region boundaries. The differential importances attached to the two attributes by a segment leads to the elliptically shaped isopref erence contours. If the weights are equal, the contours would be circles in a space of two attributes. The closest contour for any segment is generated by describing a locus for it with its ideal point as the center such that the closest existing product to it lies on this locus. Thus, for any new product to "capture" a segment it has to be positioned inside that segment's closest isopref erence contour. The unbroken (solid) contours represent the closest isopref erence boundary for the respective segments in Exhibit 1. Consider the sequential selection and introduction of two optimal new products in the market of Exhibit 1. The first new product (NP1) would be posi- tioned at NP1, capturing a total sales potential of 140 units -8- (S1+S2+S3). No other position in this market would provide a larger potential. Once NP1 has been introduced, the closest isopref erence boundaries shift inwards for SI, S2, and S3, as represented by the broken contours. The best new product position, given the position of NP1, will be anywhere within the closest isopref erence contour of S4 (choose NP2 without loss of generality) shown by the hatched area in Exhibit 1. The sales potential for NP2 is 60 units. The total sales potential from the two new products is 200 units. Now, consider simultaneous entry. The best simultaneous two new products solution will have a sales potential of 230 units — an optimal solution would be NP3 and NP4 in Exhibit 1. Simultaneous introduction is never inferior to. sequential, and, as we have just shown, may be superior. Further advantages of considering simultaneous entry would emerge if one takes into account the synergy that might be generated by two new products. (For some analytical formulations for incorporating synergy effects into the "profit" function see Sudharshan and Kumar (1984).) Note that a bump and shift type solution approach will not necessarily generate the simultaneous optimal solution. First generate the first new product position as the single best new product position. Next, generate the second position as the best single posi- tion in the market which now includes the first generated new product. Next, assume that the market consists of all existing new products and the second new product, but not the first. Generate a new product for -9- this market. A comparison of the two incumbent positions and the first two positions is made to determine if the new set of two pro- ducts is better than the earlier set. This procedure is repeated until no further improvement in solutions is better. In the example considered, NP1 and NP2 would be the first two pro- ducts generated. Given NP2 in the market, but not NP1, the best posi- tion still is NP1, and the procedure would stop, giving the same non-optimal solution in the sequential consideration method. An exception, in our example, would arise if the second position chosen happened to be NP4 — which is possible, but not highly probable. (The best (as per Sudharshan, May and Shocker (1986)) available procedures for this class of problems — the methods of Gavish, Horsky and Srikanth (1983), and Albers and Brockhoff (1980), would lead to a more interior solution for the NP2 position. ) As the number of products and the number of dimensions to be considered for entry increases the itera- tive procedure is less likely to produce the same solution as the simultaneous procedure. PROBLEM FORMULATION Following Shocker and Srinivasan (1974, 1979), products are con- ceptualized as bundles of benefits and costs. A product-market consists of those products judged by potential customers to be appro- priate for some generic purpose. The competing alternatives and ideal products are represented as point locations in a perceptual space -10- spanned by attribute dimensions determinant of brand preference/choice in that market. Preference behavior is modelled as a linear combina- tion of the different product attribute discrepancies (see Shocker and Srinivasan (1979) for a review of the logical and empirical justifi- cation for multi-attribute models generally). Following Pessemier, et al . (1971), choice is modelled probabilistically from among the k-closer competitors, where k can vary between 1 and the number of available brands. Following the notation in Shocker and Srinivasan (1974) and Sudharshan, May, and Shocker (1986), let n be the number of existing D brands in the product market, ri be the number of market segments, n be the number of determinant product attributes, n^ be the number of new products to be introduced, Y. = {y. } = the modal perception of the j product on the p dimension. W = fw. } = the attribute weights for the i segment, i l ip I, = jl. } = the ideal point for the i market segment. It is assumed finite, but need not lie in the region where feasible products might be located, d. . = the weighted Euclidean distance from the j product to the k segment's ideal point. S. = the i segment's demand. ir . . = the share of the i segment's demand allocated to the j product alternative, ir.. = f(d..) and and -11- £ it . . = 1 for each i = 1, 2, ..., n before entry n B +n N E tt . . = 1 for each i = l,2,...,n after entry. Following Rachem and Simon (1981) and Shocker and Srinivasan (1974), several forms for tt . . (i.e., decision rules) can be consid- ered : Case 1. Every available alternative could have some non-zero like- n lihood of purchase, e.g., tt . . = a./d.. where a. = 1/ E (1/d..) and b i.l i i.l i . , 1J .1=1 is a parameter which varies with the product class (Pessemier, et al . 1971). Whether or not a segment actually purchases a brand, there is the potential to do so. As a model of segment behavior, it is more credible than as a model of individual behavior, where individuals often are observed to restrict their purchases to many fewer than all available brands (Urban (1975); Silk and Urban (1978)). Case 2 . Individuals are assumed more likely to become familiar with products which come reasonably close to meeting their objectives, due to self-interest (Aaker and Myers (1974)). A parameter k (possibly k. which varies with each individual), restricts choice to the k "closer" b (k) (k) alternatives, tt . . = a./d.. for d.. < d. , where d. is the distance 1.1 i 1.1 i.l i i from the i segment's ideal point to its k t closer product, and tt . . = otherwise, i.l Case 3 . Individuals purchase only their most preferred brand, i.e., k = 1. This was referred to earlier as the single choice model. -12- SILOP, due to Albers (1976) provides a solution procedure only for this case. Assume that the firm's single objective is to maximize total incremental demand, or preference share, from the introduction of new products. This means that we must account for any demand for the new products which is cannibalized from the firm's existing brands. Let ¥ . = the set of k closer products before introduction. X. = the i firm's self (existing and new) products before introduction. ■n . . = product likelihoods of purchase before (after) intro- duction of new products, x = (x } = the n product's position, n L p . L = an arbitrarily large number, and ¥.*, X-*» an ^ "..* are the after entry equivalents to ¥., X-» and it . . respectively. Then, as in Albers (1979) and Gavish, Horsky and Srikanth (1983) for the single new product case, we can identify the optimal new positions by solving the mixed integer nonlinear programming problem Maximize E ( E u.tt*. - E ir..)S. i=l je X * J 1£X ± ■ subject to d< k > (1 - u.) < [ E A (I.. - x ) 2 w..] 1/2 < d (k) + LCI - u.) i i n n ij i i p=l for all x E R. and i e n , where u is zero or one depending on whether M l (1) or not (0) a self product (existing or new, located at ix [), is among the k closet for the i-th segment. The difference between the total revenue obtained by the firm's self and new products together, -13- after entry of its new products, and that obtained by all the firm's (existing) products before entry is the objective being maximized. The DIFFSTRAT Procedure We use a fundamental insight (augmented space method) to determine the positions of multiple new products that are simultaneously opti- mal. We illustrate this idea with an example. Without loss of generality, consider a perceptual space of one dimension, and two segments SI and S2 as shown in Exhibit 2a. Let there be two existing products EP1 and EP2 located as shown in the Exhibit. Let the sales potential of segment SI be 1 unit and that of S2 be 2 units. The ordinate in Exhibit -2p represents the sales potential attainable by any one new product position. In a typical application of PRODSRCH (May, Shocker, Sudharshan (1982), Sudharshan (1982), Sudharshan, May, Shocker (1986)) or PROPOPP (Albers (1982)), or the method of Gavish, Horsky and Srikanth (1983), we would attempt to find the best single new product position for such a market. Let our problem now be to find the two new products that are jointly optimal. Consider an augmented space of two dimensions (number of new products desired times number of dimensions of perceptual space). In Exhibit 2b, we have labelled the two dimensions AS Dim 1 and AS Dim 2 (for augmented space dimensions 1 and 2 respectively). Let AS Dim 1 be the possible locations of new product 1 (NP1), and AS Dim 2 be the The numbering of the new products as 1 and 2 is purely arbitrary, since they are simultaneously introduced. However, this numbering is useful both for the example here, as well as an index to be used in our computer software implementation of DIFFSTRAT. -14- Single New Product Sales Potential £P< 51 i *l S2 ff ePi. 51 - Segment 1 ideal point ; EP1 - Existing Product 1 52 - Segment 2 ideal point ; EP2 - Existing Product 2 Exhibit 2a: Simple Joint Space Representation of a Product-Market Exhibit 2b: Augmented Space Man of Joint Sales Potential For Two New Products Sales Potent -15- space of location of all possible new product 2 (NP2) positions. We location of all possible new product 2 (NP2) positions. We then con- struct an objective function in this space which represents the sales potential of all the joint combinations of the respective positions of the two new products (not allowed to occupy the same location a 2 degenerate case). We now try to find an optimal position in this space (given by the coordinates on AS Dim 1 and on AS Dim 2) which would represent a solution to this joint optimization problem. The coordinate AS Dim 1 gives us the position of new product 1 (NP1) in the original space and AS Dim 2 gives the position of the second new product (NP2). The objective function for the problem of Exhibit 3, for the single new product problem with a probabilistic choice con- sumer model, shows that the objective function is not generally smooth. For the single choice case it would be of a step-like nature. Not discussed earlier, QRMNEW (May, 1979) is the optimizer within DIFFSTRAT. It is used to find that combination of the analyst spe- cified n (the number of new products that we wish to consider) that N would provide the highest incremental demand of all feasible com- 3 binations of n positions. The augmented space that is set up by 2 If the two new products are located at positions (in Exhibit 2a) 0.5 and 5, respectively, sales potential for new product 1 is 1 and that for new product 2 is 2. The total sales potential for this com- bination of new products is 3. Looking at Exhibit 2b, the sales poten- tial corresponding to coordinates (0.5,5) — AS1 gives the location of the first new product, and AS2 the second one — is 3. 3 It is possible to use DIFFSTRAT to try different number of new product entries and provide the positions and the expected incremental demand for each number of new products tried. If the market structure is specified correctly, it is expected that an optimal number of products for a given market can be found in this fashion. In other words, the optimal number of new product concents can be generated endogenously if desired. The decision whether to specify the number of products exogenously or endogenously would depend on the budget available for analysis. -16- Exhibit 3 A Simplified Flow Chart of DIFFSTRAT REM) INPUT PARAMETERS J£ Call MULTDRIV for driving multiple entry problem Ad r Call MULTSPACE which sets up market for QRMNEW and OBJECTIVE Call QRMNEW ri v. _ _ J Set current solution = initial solution* (AS) Set best solution = current solution (AS) ± Generate solution, rent soli new solut V Call DECODER AS + OS Call ORJECTIVE OBJECTIVE performs function pv.i 1 uf\ t i on > '.t best solution current solution (AS) JK- Terminate QRMNEW | I Call DEC( DECODER AS ♦ OS \f ±L- Report best solution * c ^_ STOP J _J KEY: AS - augmented space OS - original space AS ■*■ OS - conversion from AS to OS OS > AS - conversion from OS to AS *weighted centeroid starting solution **these involve further calls to the objective func- tion evaluator — OBJECTIVE -17- MULTSPACE is the space in which every point represents a unique com- bination of n^ product positions. For each such point, OBJECTIVE evaluates the expected incremental demand. Consumer preference models are available (as specified) in original space (OS) only. OBJECTIVE, given the positions of a set of new products (that constitute one position in augmented space (AS)) can evaluate the expected incremen- tal demand associated with this set. The conversion of a point in AS to the corresponding positions in OS thus needs to be done; and is carried out by DECODER. The rules for this conversion were explained earlier in our discussion of the example, and shown graphically in Exhibits 2a and 2b. DIFFSTRAT reports (i) the best set of new product positions, (ii) the segments that each of these products is best suited to, (iii) the expected incremental demand from this set, (iv) the expected demand for each new product, and self-products from each segment, and (v) the closest two competitors for each new and existing self product for each segment. If repositioning/deletion of existing self products has been allowed, reports (i)-(v) are provided, indicating which self-products have been deleted or repositioned and providing their expected demand from each segment. Computational Examples with DIFFSTRAT To illustrate the differences in the market shares obtained be- tween simultaneously located new products and sequentially located ones, we performed a few simulation comparisons. Table 1 shows the positions and market shares for the simultaneous and sequential new -18- products. Markets are characterized by the number of ideal points (50 or 100), the number of existing products (10 or 15), and the size of the consideration set of the consumers (k = 2,3, or 5). New products are introduced into such markets. The joint spaces of perceptions and preferences are two dimensional, which is realistic (Aaker and Myers (1972)), and which permits plotting of the market structure, enabling a visual comparison of the positions to be made. The numbers of new products introduced was also varied (2,3, or 4). Column 4 of Table 1 shows the positions of the new products positioned sequentially. Column 5 shows the same for the DIFFSTRAT new products. From this table, it is clear that there is considerable difference between the positions of the seauential and simultaneous new products. The difference in market shares is apparently not as great as one might expect. But, considering frequently purchased goods, where each share point has considerable value (e.g., from Williams (1985), for soft drinks, each share point is worth $300 million, from Giges and Freeman (1985), for cookies, it is worth $31 million, and for coffee, $45 million), such differences are definitely meaningful, and necessary to consider in decision making. Exhibit 4 provides a graphical illustration of the differences in the positions of new products obtained using PR0DSRCH sequentially, and using DIFFSTRAT (simultaneous positioning). For the same market (the locations of the ideal points are shown in Exhibit 5), first, three new products were positioned sequentially using PR0DSRCH. Their positions are designated . Second, DIFFSTRAT was used to position -iy- TABLE 1 Comparison of Sequentially Positioned vs. Simultaneously Positioned Optimal New Products of 1 ts No. of existing products Size of considera- tion set K Number of new products Sequentially located new product positions Simultaneously located new product positions Sequential ." new products market share Simultaneous new products market share Dimensions 1 2 Dimensions 1 2 10 2 3 6.6, 4.1 3.6, 4.8 8.0, 5.4 6.4, 3.8 3.6, 4.9 8.0, 5.5 > 34.98 ) 37.30 10 2 4 6.6, 4.1 3.6, 4.8 8.0, 5.4 2.7, 3.3 6.6, 3.7 3.9, 7.6 9.2, 6.4 5.9, 8.9 ) > 41.62 42.76 10 3 2 6.9, 7.1 6.9, 7.0 7.1, 6.9 4.4, 3.6 > 20.7 22.24 3 4 " 6.9, 7.1 6.9, 7.0 3.6, 3.9 4.9, 7.0 5.4, 6.9 6.6, 5.1 3.6, 3.9 4.8, 6.3 > 38.88 42.92 15 5 I t 3 7.6, 6.8 7.6, 5.7 7.8, 5.4 4.3, 7.4 4.0, 6.2 7.8, 4.5 1 > 18.62 20.14" 15 5 : l 4 7.6, 6.8 7.6, 5.7' 7.8, 5.4 4.3, 7.4 7.6, 6.8 4.6, 4.0 8.1, 6.6 4.1, 6.5 1 > 25.82 28.64 i 1 j 15 1 j i 3 i 4 6.4, 3.2 3.8, 3.8 4.2, 7.0 2.3, 5.5 6.2, 2.7 3.4, 6.0 7.4, 7.4 8.2, 4.5 1 > 28.42 J 32.12 ; io ! j i i i 3 3 7.6, 5.6 3.9, 3.4 3.6, 5.2 7.5, 4.3 3.3, 6.1 5.8, 8.1 > 37.99 \ 39.23 1 i 1 10 i 5 3 7.4, 5.2 3.8, 5.8 4.0, 5.4 8.1, 5.9 3.2, 6.2 5.9, 3.3 > 1 > 28.20 J ! 29.56 ! i ! Table 1, continued -20- No. of ideal points No. of existing products Size of considera- tion set K Number of new products Sequentially located new product positions Simultaneously located new product positions Sequential * new products market share Siraultaneoi new product market sha* Dimensions 1 2 Dimensions 1 2 15 3 3 4.6, 7.1 6.5, 7.0 5.2, 6.8 3.1, 4.6 6.4, 6.8 4.9, 6.7 > 21.04 22.69 4|| 15 5 2 7.2, 5.6 6.9, 5.5 7.6, 5.0 7.0, 1.9 > 16.17 16.45 * Correlated t = 5.718, p> 0.001, for the difference simultaneous market share - sequential market share. -21- three new products, their positions are shown using the symbol^ in Exhibit 4. Third, the fourth sequential new product was positioned using PRODSRCH, it is represented byg. Lastly, DIFFSTRAT was used to position four new products. These are shown asQ. The differences in the positions of the PRODSRCH new products and the DIFFSTRAT new products appear relatively minor for three new products. But, for four new products these positions are remarkably different! A carefully constructed simulation comparison is called for to better understand market structures. We now have relatively good procedures (PRODSRCH and DIFFSTRAT) to generate optimal new products in markets with unequal sales potentials associated with each segment, and where the segments can have a probabilistic preference share model (k > 1). We can address the issues of the equilibrium market struc- tures that would prevail under different market assumptions. This becomes especially easier if we consider that if the effects of ad- vertising and distribution are introduced using response function models (Stern and El-Ansary (1982), Urban and Hauser (1980), Little (1979), etc.), then, marketing strategies, in eauilibrium, can be separably analyzed for product positioning strategies and for adver- tising and distribution strategies, respectively (Hauser and Shugan (1983), Kumar and Sudharshan (1986)). Discussion and Conclusions Unlike existing procedures which allow for evaluation of prespe- cified new product concepts (see Shocker and Srinivasan (1979) for a review of such methods), or, those which given a set of new product -22- EXHIBIT 4: POSITIONS OF SIMULTANEOUS AND SEQUENTIAL NEW PRODUCTS ♦ 3 SEQUENTIAL o 3 simult B A Th SEQUENT D 4 SIMULT DIMENSION 2 9 T 8 7 6 5 4 34 2 1 9 5 6 7 DIMENSION 1 -23- EXHIBIT 5: JOINT SPACE OF IDEAL POINTS AND NEW PRODUCTS DIMENSION 2 10 T 9 .. 8 .. 7 .. 6 5 .. 4 3 2 1 ♦ V \ ♦ ♦ o ♦ ♦ a. ♦ ♦ o * ♦ ♦