Knowledge that Transforms

With the increase in new product introductions in consumer packaged goods categories, supermarkets are reluctant to accept new products. Therefore, it is very important for manufacturers to convince retailers of the high-demand potential of their products. We study how a high-demand manufacturer can use advertising, slotting allowances, and wholesale prices to signal its high demand to retailers. Specifically, we examine the relative importance of advertising and slotting allowance in signaling demand. That is, when is it optimal for the manufacturer to use high advertising support, and when is it optimal for it to offer slotting allowance as a signal of its demand? We show that when a high-demand manufacturer is trying to signal its demand to retailers, advertising and slotting allowance are partial substitutes of one another in the sense that the manufacturer can increase one in order to compensate for a reduction in the other. We find that the high-demand manufacturer's signaling strategy depends on three factors: the retailer's stocking costs, the intensity of retail competition, and the advertising response rate in the given product market. We begin with a model of one manufacturer dealing with one retailer. The manufacturer has private information about the potential demand for its new product. The retailer is uncertain about the likely demand of the new product and is willing to accept the product only if it is convinced that the demand is high. We characterize the high-demand manufacturer's separating equilibrium strategies. We find that the slotting allowance plays an important role in signaling when the retailer's stocking costs are high and the advertising effectiveness is low. On the other hand, the manufacturer does not offer any slotting allowance, and advertising plays a bigger role when the stocking costs are low or the advertising effectiveness is high. We then examine the effects of retail competition on the manufacturer strategy. We find that the slotting allowance plays a more important role when the retail level competition is very intense. The manufacturer may have to offer a positive slotting allowance even in the absence of retailers' demand uncertainty when the retail competition is sufficiently intense. This result shows that the slotting allowance may have an important role to play even in the absence of signaling or screening considerations. Thus, our analysis of competitive setting provides an alternative explanation for slotting allowances. It also offers support to the views of many retailers who believe that slotting allowances can help retailers recover high stocking costs in highly competitive retail markets. In the presence of retailers' demand uncertainty, the manufacturer offers a higher slotting allowance in order to signal its high demand. We also investigate the effect of retailer's uncertainty about the effectiveness of the manufacturer's advertising. We show that if the high-demand manufacturer also has a higher advertising response rate, the manufacturer provides even higher advertising support to alleviate the retailer's advertising-related uncertainty. By increasing the advertising support, the manufacturer credibly tells the retailer; that it would not be optimal for the manufacturer to provide such high advertising support unless it had high enough advertising effectiveness.

It is a popular contention that products launched today diffuse faster than products launched in the past. However, the evidence of diffusion acceleration is rather scant, and the methodology used in previous studies has several weaknesses. Also, little is known about why such acceleration would have occurred. This study investigates changes in diffusion speed in the United States over a period of 74 years (1923–1996) using data on 31 electrical household durables. This study defines diffusion speed as the time it takes to go from one penetration level to a higher level, and it measures speed using the slope coefficient of the logistic diffusion model. This metric relates unambiguously both to speed as just defined and to the empirical growth rate, a measure of instantaneous penetration growth. The data are analyzed using a single-stage hierarchical modeling approach for all products simultaneously in which parameters capturing the adoption ceilings are estimated jointly with diffusion speed parameters. The variance in diffusion speed across and within products is represented separately but analyzed simultaneously. The focus of this study is on description and explanation rather than forecasting or normative prescription. There are three main findings. 1. On average, there has been an increase in diffusion speed that is statistically significant and rather sizable. For the set of 31 consumer durables, the average value of the slope parameter in the logistic model's hazard function was roughly 0.48, increasing with 0.09 about every 10 years. It took an innovation reaching 5% household penetration in 1946 an estimated 13.8 years to go from 10% to 90% of its estimated maximum adoption ceiling. For an innovation reaching 5% penetration in 1980, that time would have been halved to 6.9 years. This corresponds to a compound growth rate in diffusion speed of roughly 2% between 1946 and 1980. 2. Economic conditions and demographic change are related to diffusion speed. Whether the innovation is an expensive item also has a sizable effect. Finally, products that required large investments in complementary infrastructure (radio, black and white television, color television, cellular telephone) and products for which multiple competing standards were available early on (PCs and VCRs) diffused faster than other products once 5% household penetration had been achieved. 3. Almost all the variance in diffusion speed among the products in this study can be explained by (1) the systematic increase in purchasing power and variations in the business cycle (unemployment), (2) demographic changes, and (3) the changing nature of the products studied (e.g., products with competing standards appear only late in the data set). After controlling for these factors, no systematic trend in diffusion speed remains unaccounted for. These findings are of interest to researchers attempting to identify patterns of difference and similarity among the diffusion paths of many innovations, either by jointly modeling the diffusion of multiple products (as in this study) or by retrospective meta-analysis. The finding that purchasing power, demographics, and the nature of the products capture nearly all the variance is of particular interest. Specifically, one does not need to invoke unobserved changes in tastes and values, as some researchers have done, to account for long-term changes in the speed at which households adopt new products. The findings also suggest that new product diffusion modelers should attempt to control not only for marketing mix variables but also for broader environmental factors. The hierarchical model structure and the findings on the systematic variance in diffusion speed across products are also of interest to forecasting applications when very little or no data are available.

Recent work in marketing has drawn on behavioral decision theory to advance the notion that consumers evaluate attributes (and therefore choice alternatives) not only in absolute terms, but as deviations from a reference point. The theory has important substantive and practical implications for the timing and execution of price promotions and other marketing activities. Choice modelers using scanner panel data have tested for the presence of these “reference effects” in consumer response to an attribute such as price. In applications of the theory of reference-dependent choice (Tversky and Kahneman 1991), some modelers report empirical evidence of loss aversion: When a consumer encounters a price above his or her established reference point (a “loss”), the response is greater than for a price below the reference point (a “gain”). Researchers have gone so far as to suggest that evidence for the so-called reference effect make it an empirical generalization in marketing (e.g., Kalyanaram and Winer 1995, Meyer and Johnson 1995). It is our contention that the measurement of loss aversion in empirical applications of the reference-dependent choice model is confounded by the presence of unaccounted-for heterogeneity in consumer price responsiveness. Our reasoning is that the kinked price response curve implied by loss aversion is confounded with the slopes of the response curves across segments that are differentially responsive to price. A more price-responsive consumer (with a steeper response function) tends to have a lower price level as a reference point. This consumer faces a larger proportion of prices above his reference point, thus the response curve is steeperin the domain of losses. Similarly, the less price-responsive consumer sees a greater proportion of prices below his reference point, so the response curve is less steep within the domain of gains. As a result, any cross-sectional estimate of loss aversion that does not take this into account will be biased upward—researchers who do not control for heterogeneity in price responsiveness may arrive at incorrect substantive conclusions about the phenomenon. It is interesting to note that in this instance, failure to control for heterogeneity induces a bias in favor of finding an effect, rather than the more typical case of attenuation of the effect toward zero. We first test our assertion regarding the referencedependent model using scanner panel data on refrigerated orange juice and subsequently extend this analysis to 11 additional product categories. In all cases we find, as predicted, that accounting for price-response heterogeneity leads to lower and frequently nonsignificant estimates of loss aversion. We do, however, find some categories in which the effect does not disappear altogether. We also estimate loss aversion using a “sticker shock” model of brand choice in which the reference prices are brand-specific. In line with the results of the majority of prior literature, we find smaller and insignificant estimates of loss aversion in this model. We show that this is because in the sticker shock model, there is no apparent correlation between the price responsiveness of the consumer and the representation of reference effects as losses or gains. Our findings strongly suggest that loss aversion may not in fact be a universal phenomenon, at least in the context of frequently purchased grocery products.

In spite of the high financial stakes involved in marketing new motion pictures, marketing science models have not been applied to the prerelease market evaluation of motion pictures. The motion picture industry poses some unique challenges. For example, the consumer adoption process for movies is very sensitive to word-of-mouth interactions, which are difficult to measure and predict before the movie has been released. In this article, we undertake the challenge to develop and implement MOVIEMOD—a prerelease market evaluation model for the motion picture industry. MOVIEMOD is designed to generate box-office forecasts and to support marketing decisions for a new movie after the movie has been produced (or when it is available in a rough cut) but before it has been released. Unlike other forecasting models for motion pictures, the calibration of MOVIEMOD does not require any actual sales data. Also, the data collection time for a product with a limited lifetime such as a movie should not take too long. For MOVIEMOD it takes only three hours in a “consumer clinic” to collect the data needed for the prediction of box-office sales and the evaluation of alternative marketing plans. The model is based on a behavioral representation of the consumer adoption process for movies as a macroflow process. The heart of MOVIEMOD is an interactive Markov chain model describing the macro-flow process. According to this model, at any point in time with respect to the movie under study, a consumer can be found in one of the following behavioral states: undecided, considerer, rejecter, positive spreader, negative spreader, and inactive. The progression of consumers through the behavioral states depends on a set of movie-specific factors that are related to the marketing mix, as well as on a set of more general behavioral factors that characterize the movie-going behavior in the population of interest. This interactive Markov chain model allows us to account for word-of-mouth interactions among potential adopters and several types of word-of-mouth spreaders in the population. Marketing variables that influence the transitions among the states are movie theme acceptability, promotion strategy, distribution strategy, and the movie experience. The model is calibrated in a consumer clinic experiment. Respondents fill out a questionnaire with general items related to their movie-going and movie communication behavior, they are exposed to different sets of information stimuli, they are actually shown the movie, and finally, they fill outpostmovie evaluations, including word-of-mouth intentions.These measures are used to estimate the word-of-mouth parameters and other behavioral factors, as well as the movie-specific parameters of the model. MOVIEMOD produces forecasts of the awareness, adoption intention, and cumulative penetration for a new movie within the population of interest for a given base marketing plan. It also provides diagnostic information on the likely impact of alternative marketing plans on the commercial performance of a new movie. We describe two applications of MOVIEMOD: One is a pilot study conducted without studio cooperation in the United States, and the other is a full-fledged implementation conducted with cooperation of the movie's distributor and exhibitor in the Netherlands. The implementations suggest that MOVIEMOD produces reasonably accurate forecasts of box-office performance. More importantly, the model offers the opportunity to simulate the effects of alternative marketing plans. In the Dutch application, the effects of extra advertising, extra magazine articles, extra TV commercials, and higher trailer intensity (compared to the base marketing plan of the distributor) were analyzed. We demonstrate the value of these decision-support capabilities of MOVIEMOD in assisting managers to identify a final plan that resulted in an almost 50% increase in the test movie's revenue performance, compared to the marketing plan initially contemplated. Management implemented this recommended plan, which resulted in box-office sales that were within 5% of the MOVIEMOD prediction. MOVIEMOD was also tested against several benchmark models, and its prediction was better in all cases. An evaluation of MOVIEMOD jointly by the Dutch exhibitor and the distributor showed that both parties were positive about and appreciated its performance as a decision-support tool. In particular, the distributor, who has more stakes in the domestic performance of its movies, showed a great interest in using MOVIEMOD for subsequent evaluations of new movies prior to their release. Based on such evaluations and the initial validation results, MOVIEMOD can fruitfully (and inexpensively) be used to provide researchers and managers with a deeper understanding of the factors that drive audience response to new motion pictures, and it can be instrumental in developing other decision-support systems that can improve the odds of commercial success of new experiential products.

In collaborating to compete, firms forge different types of strategic alliances: same-function alliances, parallel development of new products, and cross-functional alliances. A major challenge in the management of these alliances is how to control the resource commitment of partners to the collaboration. In this research we examine both theoretically and experimentally how the type of an alliance and the prescribed profit-sharing arrangement affect the resource commitments of partners. We model the interaction within an alliance as a noncooperative variable-sum game, in which each firm invests part of its resources to increase the utility of a new product offering. Different types of alliances are modeled by varying how the resources committed by partners in an alliance determine the utility of the jointly-developed new product. We then model the interalliance competition by nesting two independent intra-alliance games in a supergame in which the groups compete for a market. The partners of the winning alliance share the profits in one of two ways: equally or proportionally to their investments. The Nash equilibrium solutions for the resulting games are examined. In the case of same-function alliances, when the market is large the predicted investment patterns under both profit-sharing rules are comparable. Partners developing new products in parallel, unlike the partners in a same-function alliance, commit fewer resources to their alliance. Further, the profit-sharing arrangement matters in such alliances—partners commit more resources when profits are shared proportionally rather than equally. We test the predictions of the model in two laboratory experiments. We find that the aggregate behavior of the subjects is accounted for remarkably well by the equilibrium solution. As predicted, profit-sharing arrangement did not affect the investment pattern of subjects in same-function alliances when they were in the high-reward condition. Subjects developing products in parallel invested less than subjects in same-function alliance, irrespective of the reward condition. We notice that theory seems to underpredict investments in low-reward conditions. Aplausible explanation for this departure from the normative benchmark is that subjects in the low-reward condition were influenced by altruistic regard for their partners. These experiments also clarify the support for the mixed strategy equilibrium: aggregate behavior conforms to the equilibrium solution, though the behavior of individual subjects varies substantially from the norm. Individual-level analysis suggests that subjects employ mixed strategies, but not as fully as the theory demands. This inertia in choice of strategies is consistent with learning trends observed in the investment pattern. A new analysis of Robertson and Gatignon's (1998) field survey data on the conduct of corporate partners in technology alliances is also consistent with our model of samefunction alliances. We extend the model to consider asymmetric distribution of endowments among partners in a same-function alliance. Then we examine the implication of extending the strategy space to include more levels of investment. Finally, we outline an extension of the model to consider cross-functional alliances.

The number of brands in the marketplace has vastly increased in the 1980s and 1990s, and the amount of money spent on advertising has run parallel. Print advertising is a major communication instrument for advertisers, but print media have become cluttered with advertisements for brands. Therefore, it has become difficult to attract and keep consumers' attention. Advertisements that fail to gain and retain consumers' attention cannot be effective, but attention is not sufficient: Advertising needs to leave durable traces of brands in memory. Eye movements are eminent indicators of visual attention. However, what is currently missing in eye movementresearch is a serious account of the processing that takes place to store information in long-term memory. We attempt to provide such an account through the development of a formal model. We model the process by which eye fixations on print advertisements lead to memory for the advertised brands, using a hierarchical Bayesian model, but, rather than postulating such a model as a mere data-analysis tool, we derive it from substantive theory on attention and memory. The model is calibrated to eye-movement data that are collected during exposure of subjects to ads in magazines, and subsequent recognition of the brand in a perceptual memory task. During exposure to the ads we record the frequencies of fixations on three ad elements; brand, pictorial and text and, during the memory task, the accuracy and latency of memory. Thus, the available data for each subject consist of the frequency of fixations on the ad elements and the accuracy and the latency of memory. The model that we develop is grounded in attention and memory theory and describes information extraction and accumulation during ad exposure and their effect on the accuracy and latency of brand memory. In formulating it, we assume that subjects have different eye-fixation rates for the different ad elements, because of which a negative binomial model of fixation frequency arises, and we specify the influence of the size of the ad elements. It is assumed that the number of fixations, not their duration, is related to the amount of information a consumer extracts from an ad. The information chunks extracted at each fixation are assumed to be random, varying across ads and consumers, and are estimated from the observed data. The accumulation of information across multiple fixations to the ad elements in long-term memory is assumed to be additive. The total amount of accumulated information that is not directly observed but estimated using our model influences both the accuracy and latency of subsequent brand memory. Accurate memory is assumed to occur when the accumulated information exceeds a threshold that varies randomly across ads and consumers in a binary probit-type of model component. The effect of two media-planning variables, the ad's serial position in a magazine and the ad's location on the double page, on the brand memory threshold are specified. We formulate hypotheses on the effects of ad element surface, serial position, and location. The model is applied in a study involving a sample of 88 consumers who were exposed to 65 print ads appearing in their natural context in two magazines. The frequency of eye fixations was recorded for each consumer and advertisement with infrared eye-tracking methodology. In a subsequent indirect memory task, consumers identified the brands from pixelated images of the ads. Across the two magazines, fixations to the pictorial and the brand systematically promote accurate brand memory, but text fixations do not. Brand surface has a particularly prominent effect. The more information is extracted from an ad during fixations, the shorter the latency of brand memory is. We find a systematic recency effect: When subjects are exposed to an ad later, they tend to identify it better. In addition, there is a small primacy effect. The effect of the ad's location on the right or left of the page depends on the advertising context. We show how the model supports advertising planning and testing and offer recommendations for further research on the effectiveness of brand communication. In future research the model may be extended to accommodate the effects of repeated exposure to ads, to further detail the representation of strength and association of memory, and to include the effects of creative tactics and media planning variables beyond the ones we included in the present study.

A fundamental dilemma confronts retailers with stand-alone sites on the World Wide Web and those attempting to build electronic malls for delivery via the Internet, online services, or interactive television (Alba et al. 1997). For consumers, the main potential advantage of electronic shopping over other channels is a reduction in search costs for products and product-related information. Retailers, however, fear that such lowering of consumers' search costs will intensify competition and lower margins by expanding the scope of competition from local to national and international. Some retailers' electronic offerings have been constructed to thwart comparison shopping and to ward off price competition, dimming the appeal of many initial electronic shopping services. Ceteris paribus, if electronic shopping lowers the cost of acquiring price information, it should increase price sensitivity, just as is the case for price advertising. In a similar vein, though, electronic shopping can lower the cost of search for quality information. Most analyses ignore the offsetting potential of the latter effect to lower price sensitivity in the current period. They also ignore the potential of maximally transparent shopping systems to produce welfare gains that give consumers a long-term reason to give repeat business to electronic merchants (cf. Alba et al. 1997, Bakos 1997). We test conditions under which lowered search costs should increase or decrease price sensitivity. We conducted an experiment in which we varied independently three different search costs via electronic shopping: search cost for price information, search cost for quality information within a given store, and search cost for comparing across two competing electronic wine stores. Consumers spent their own money purchasing wines from two competing electronic merchants selling some overlapping and some unique wines. We show four primary empirical results. First, for differentiated products like wines, lowering the cost of search for quality information reduced price sensitivity. Second, price sensitivity for wines common to both stores increased when cross-store comparison was made easy, as many analysts have assumed. However, easy cross-store comparison had no effect on price sensitivity for unique wines. Third, making information environments more transparent by lowering all three search costs produced welfare gains for consumers. They liked the shopping experience more, selected wines they liked more in subsequent tasting, and their retention probability was higher when they were contacted two months later and invited to continue using the electronic shopping service from home. Fourth, we examined the implications of these results for manufacturers and examined how market shares of wines sold by two stores or one were affected by search costs. When store comparison was difficult, results showed that the market share of common wines was proportional to share of distribution; but when store comparison was made easy, the market share returns to distribution decreased signi.cantly. All these results suggest incentives for retailers carrying differentiated goods to make information environments maximally transparent, but to avoid price competition by carrying more unique merchandise.

The issue of “power” in the marketing channels for consumer products has received considerable attention in both academic and practitioner journals as well as in the popular press. Our objective in this paper is to provide an empirical method to measure the power of channel members and to understand the reasons (demand factors, cost factors, nature of channel interactions) for this power. We confine our analysis to pricing power in channels. We use methods from the game-theory literature in marketing on channel interactions to obtain the theoretical framework for our empirical model. This literature provides us a definition of power—one that is based on the proportion (or percentage) of channel profits that accrue to each of the channel members. There can be a variety of possible channel interactions between manufacturers and retailers in channels. The theoretical literature has examined some of these games. For example, Choi (1991) examines how channel profits for manufacturers and retailer vary if channel interactions are either vertical Nash, or if they are Stackelberg leaderfollower with either the manufacturer or the retailer being the price leader. Each of these three channel interaction games has different implications for profits made by manufacturers and retailers, and consequently for the relative power of the channel members. In contrast to the previous literature that has focused largely on the above three channel interaction games, our model extends the game-theoretic literature by allowing for a continuum of possible channel interactions between manufacturers and a retailer. Furthermore, for a given product market, we empirically estimate from the data where the channel interactions lie in this continuum. More critically, we obtain measures of how channel profits are divided between manufacturers and the retailer in the product market, where a higher share of channel profit is associated with higher channel power. We then examine how channel power is related to demand conditions facing various brands and cost parameters of various manufacturers. In going from game-theory-based theoretical models of channel interactions to empirical estimation, we use the “new empirical industrial organization” framework (Bresnahan 1988). As part of this structural modeling framework, we build retail-level demand functions for the various brands (manufacturer and private label) in a given product category. Given these demand functions, we obtain optimal pricing rules for manufacturers and the retailer. In determining their optimal prices, manufacturers and the retailer account for how all the players in the channel choose their optimal prices. That is, we account for dependencies in decision making across channel members. These dependencies are characterized by a set of “conduct parameters,” which are estimated from market data. The conduct parameters enable us to identify the nature of channel interactions between manufacturers and the retailer (along the continuum mentioned previously). In addition to the demand and conduct parameters, manufacturers' marginal costs are also estimated in the model. These marginal cost estimates, along with the manufacturer prices and retail prices available in our dataset, enable us to compute the division of channel profits among the channel members. Hence, we are able to obtain insights into who has pricing power in the channel. In the empirical application of the model, we analyze a local market for two product categories: refrigerated juice and tuna. In both categories, there are three major brands. The difference between them is that the private label has an insignificant market share in the tuna category. Our main empirical results show that the usual games examined in the marketing literature do not hold for the given data. We also .nd that the retailer's market power is very significant in both these product categories, and that the estimated demand and cost parameters are consistent with the estimated pattern of conduct between the manufacturers and the retailer. Given the evidence from the trade press of intense manufacturer competition in these categories, as well as the “commodity” nature of these products, the result of retailer power appears intuitive.

The Internet has signi.cantly reduced the marginal cost of producing and distributing digital information goods. It also coincides with the emergence of new competitive strategies such as large-scale bundling. In this paper, we show that bundling can create “economies of aggregation” for information goods if their marginal costs are very low, even in the absence of network externalities or economies of scale or scope. We extend the Bakos-Brynjolfsson bundling model (1999) to settings with several different types of competition, including both upstream and downstream, as well as competition between a bundler and single good and competition between two bundlers. Our key results are based on the “predictive value of bundling,” the fact that it is easier for a seller to predict how a consumer will value a collection of goods than it is to value any good individually. Using a model with fully rational and informed consumers, we use the Law of Large Numbers to show that this will be true as long as the goods are not perfectly correlated and do not affect each other's valuations significantly. As a result, a seller typically can extract more value from each information good when it is part of a bundle than when it is sold separately. Moreover, at the optimal price, more consumers will find the bundle worth buying than would have bought the same goods sold separately. Because of the predictive value of bundling, large aggregators will often be more pro.table than small aggregators, including sellers of single goods. We find that these economies of aggregation have several important competitive implications: 1. When competing for upstream content, larger bundlers are able to outbid smaller ones, all else being equal. This is because the predictive value of bundling enables bundlers to extract more value from any given good. 2. When competing for downstream consumers, the act of bundling information goods makes an incumbent seem “tougher” to single-product competitors selling similar goods. The resulting equilibrium is less profitable for potential entrants and can discourage entry in the bundler's markets, even when the entrants have a superior cost structure or quality. 3. Conversely, by simply adding an information good to an existing bundle, a bundler may be able to profitably enter a new market and dislodge an incumbent who does not bundle, capturing most of the market share from the incumbent firm and even driving the incumbent out of business. 4. Because a bundler can potentially capture a large share of profits in new markets, single-product firms may have lower incentives to innovate and create such markets. At the same time, bundlers may have higher incentives to innovate. For most physical goods, which have nontrivial marginal costs, the potential impact of large-scale aggregation is limited. However, we find that these effects can be decisive for the success or failure of information goods. Our results have particular empirical relevance to the markets for software and Internet content and suggest that aggregation strategies may take on particular relevance in these markets.

Despite the explosive growth of electronic commerce and the rapidly increasing number of consumers who use interactive media (such as the World Wide Web) for prepurchase information search and online shopping, very little is known about how consumers make purchase decisions in such settings. A unique characteristic of online shopping environments is that they allow vendors to create retail interfaces with highly interactive features. One desirable form of interactivity from a consumer perspective is the implementation of sophisticated tools to assist shoppers in their purchase decisions by customizing the electronic shopping environment to their individual preferences. The availability of such tools, which we refer to as interactive decision aids for consumers, may lead to a transformation of the way in which shoppers search for product information and make purchase decisions. The primary objective of this paper is to investigate the nature of the effects that interactive decision aids may have on consumer decision making in online shopping environments. While making purchase decisions, consumers are often unable to evaluate all available alternatives in great depth and, thus, tend to use two-stage processes to reach their decisions. At the first stage, consumers typically screen a large set of available products and identify a subset of the most promising alternatives. Subsequently, they evaluate the latter in more depth, perform relative comparisons across products on important attributes, and make a purchase decision. Given the different tasks to be performed in such a two-stage process, interactive tools that provide support to consumers in the following respects are particularly valuable: (1) the initial screening of available products to determine which ones are worth considering further, and (2) the in-depth comparison of selected products before making the actual purchase decision. This paper examines the effects of two decision aids, each designed to assist consumers in performing one of the above tasks, on purchase decision making in an online store. The first interactive tool, a recommendation agent (RA), allows consumers to more efficiently screen the (potentially very large) set of alternatives available in an online shopping environment. Based on self-explicated information about a consumer's own utility function (attribute importance weights and minimum acceptable attribute levels), the RA generates a personalized list of recommended alternatives. The second decision aid, a comparison matrix (CM), is designed to help consumers make in-depth comparisons among selected alternatives. The CM allows consumers to organize attribute information about multiple products in an alternatives × attributes matrix and to have alternatives sorted by any attribute. Based on theoretical and empirical work in marketing, judgment and decision making, psychology, and decision support systems, we develop a set of hypotheses pertaining to the effects of these two decision aids on various aspects of consumer decision making. In particular, we focus on how use of the RA and CM affects consumers' search for product information, the size and quality of their consideration sets, and the quality of their purchase decisions in an online shopping environment. A controlled experiment using a simulated online store was conducted to test the hypotheses. The results indicate that both interactive decision aids have a substantial impact on consumer decision making. As predicted, use of the RA reduces consumers' search effort for product information, decreases the size but increases the quality of their consideration sets, and improves the quality of their purchase decisions. Use of the CM also leads to a decrease in the size but an increase in the quality of consumers' consideration sets, and has a favorable effect on some indicators of decision quality. In sum, our findings suggest that interactive tools designed to assist consumers in the initial screening of available alternatives and to facilitate in-depth comparisons among selected alternatives in an online shopping environment may have strong favorable effects on both the quality and the efficiency of purchase decisions—shoppers can make much better decisions while expending substantially less effort. This suggests that interactive decision aids have the potential to drastically transform the way in which consumers search for product information and make purchase decisions.