CONSUMERS ’ KNOWLEDGE , MAXIMIZING TENDENCIES , AND POST-DECISION INFORMATION SEARCH Conhecimento do consumidor , tendência à maximização

Nowadays consumers have more previous knowledge about products and services before making decisions. This study sheds light on the effects of consumers’ previous knowledge on post-decision information search. Previous studies argue that cognitive dissonance and feelings of regret or dissatisfaction elicit this search. However, we show through one experimental and two correlational studies that this view is incomplete. Our findings indicate that knowledgeable consumers search for more information at the post-decision stage, even when the decision cannot be modified. This main effect is stronger (weaker) for maximizers (satisficers). Also, cognitive dissonance affects the postdecision information search behavior. Therefore, we suggest a new variable, consumers’ previous knowledge, for consideration in the post-decision information search model.


INTRODUCTION
Collectively, 264 million US smartphone users view their devices 12 billion times per day (Deloitte, 2017), Brazilians alone spend 9 hours per day on the Internet (Kemp, 2018). These statistics indicate that the tools to access and acquire information are part of consumers' daily lives (Dholakia, Zwick, & Denegri-Knott, 2013). For instance, consumers often engage in showrooming because they can easily look up information (Mehra, Kumar, & Raju, 2017). Consequently, they become more knowledgeable about the available options and product features before making decisions (Dholakia et al., 2013).
Since information search still occurs after purchasing, it affects consumers' perceptions about their purchases (Teodorescu, Sang, & Todd, 2018). However, researchers and marketers have been neglecting this behavior. Researchers focus mainly on behaviors such as word of mouth (e.g., Chu & Kim, 2011;Hennig-Thurau, Gwinner, Walsh, & Gremler, 2004), consumer satisfaction, and feelings of regret (e.g., Lee & Kim, 2008;Wirtz, Matilla & Tan, 2007;Zeelenberg & Pietres, 2007) when they investigate consumers' behavior at the post-decision stage. The few studies on post-decision information search mention that this search is primarily driven by cognitive dissonance and feelings of regret/ dissatisfaction (Oliver, 2014;Shani & Zeelenberg, 2007). However, we claim that this view is incomplete as researchers do not consider consumers' previous knowledge in their post-decision information search models. The effect of consumers' previous knowledge on information search has already been explored for the initial stages of the customer journey (Hadar, Sood, & Fox, 2013;Carlson, Vincent, Hardesty, & Bearden, 2009;Brucks, 1985). However, to the best of our knowledge, prior studies have not extended this effect to a post-decision context.
This study addresses this gap in the literature by exploring the effect of consumers' previous knowledge at the pre-purchase stage on post-decision information search. Also, we present an interaction between consumers' previous knowledge and maximizing tendencies. We achieve this objective while controlling for cognitive dissonance and (dis)satisfaction as predictors of postdecision information search. Across three studies (one experiment and two correlational studies), we show that knowledgeable consumers search for more information the post-decision phase, even when they cannot change their decisions, because they can easily access information and their knowledge makes the search an easier task (Dholakia et al., 2013;Brucks, 1985). High levels of maximizing tendencies strengthen this effect. Additionally, cognitive dissonance has a significant effect on post-decision information search, and satisfaction does not.
Our study has several theoretical contributions. First, it sheds light on a new predictor of post-decision information search, the consumer's previous knowledge, updating the literature on this topic. This new predictor is in tune with an important phenomenon, characterized by the dramatic increase in consumers' knowledge through the widespread access of information (Dholakia et al., 2013). Second, to the best of our knowledge, we present a new interaction effect between previous knowledge and maximizing tendencies on post-decision information search, thus advancing the literature on maximizing tendencies as well. Few studies have investigated the effects of this trait on consumer behavior in the post-decision stage (Ma & Roese, 2014). Third, this research explores the role of cognitive dissonance, which, so far, has not been empirically tested, even though Donelly and Ivancevich (1970) and Oliver (2014) state it to be an important driver of post-decision information search.
Fourth, based on Lemon and Verhoef (2016)'s claim, we analyze consumer behavior through the link between two different stages of their journey: the pre-and post-decision phases. Thus, we present a more complete view of consumer behavior during the decision-making process.
For marketers, we elucidate the relevant effect of consumers' knowledge that increases the post-decision information search, and where consumers look up information at the post-decision stage. Companies usually invest less effort in the post-decision phase (Brega, 2018), and hence may lose the opportunity to diminish back-out behaviors and reinforce consumer choice (Donelly & Ivancevich, 1970;Oliver, 2014). If companies understand and invest in this phase, they could increase their capabilities to create a "loyalty loop" (Brega, 2018).
Helping marketers know more about consumers' information search behavior at the post-decision stage, and which variables affect this behavior, may help business companies to get the loyalty loop.

Consumers' previous knowledge
The decision-maker's previous knowledge attracts the attention of researchers in different areas (Bettman & Park, 1980;Brucks, 1985;Hadar et al. 2013). In this work, we comply with Brucks' approach (1985), which states that previous knowledge consists of information stored in individuals' memories in a specific moment.
In our case, this knowledge is the knowledge consumers hold before decision-making. In consumption settings, previous knowledge would reflect an individual's degree of knowledge of brand names, usage procedures, and product features, among other attributes (Carlson et al., 2009).
Previous knowledge can be categorized into types: objective and subjective. Objective knowledge is the knowledge that is truly stored in an individual's memory. This knowledge depends on the consumer's actual ability to evaluate and use a product (Alba & Hutchinson, 2000). When measuring objective knowledge, researchers might test subjects in a focal topic. Alternatively, researchers try to keep the objective knowledge constant to all participants to avoid confounds produced by such knowledge (Brucks, 1985). They can do this by using an unknow product and giving the same information about this product to the participants, for example (Brucks, 1985;Alba & Hutchinson, 2000). Subjective knowledge refers to a process in which an individual scans her memory for cues to evaluate her level of knowledge concerning a certain domain (Park, Monthersbaugh, & Feick, 1994). This knowledge is built on the information an individual thinks she has stored in memory (Hadar et al., 2013). It is conventionally measured using items that request individuals to report the knowledge they believe they hold about a focal topic (Hadar et al., 2013). Though many researchers have used subjective knowledge measures in their studies (Brucks, 1985;Hadar et al. 2013), there is some concern about their effectiveness as a proxy for objective knowledge. Particularly, Carlson et al. (2009) andBrucks (1985) suggest that self-confidence influences subjective knowledge measures because individuals may overestimate (or underestimate) their knowledge depending on their confidence. However, some critics argue that self-reported knowledge often approximates actual knowledge (Alba & Hutchinson, 2000). Carlson et al. (2009) demonstrated through a meta-analysis that subjective and objective knowledge are significantly and positively correlated.
An important issue for studies that analyze the effect of previous knowledge on information search is knowledge calibration, which reflects the agreement between objective and self-assessed knowledge (Alba & Hutchinson, 2000;Carlson et al., 2009). Considering this discussion, the studies presented herein investigated both objective and subjective previous knowledge.

Consumers' previous knowledge and information search
Knowledge is a key concept in information processing research (Raju et al., 1995) There exists a traditional stream of research that particularly explores the effects of consumers' previous knowledge on information search at the initial stage of the customer journey. These studies have found positive as well as negative relationships between these two variables (Hadar et al., 2013;Carlson et al., 2009;Brucks, 1985;Kiel & Layton, 1983). Positive effects occur because knowledgeable individuals perceive the processing of new information to be an easier task (Carlson et al. 2009;Punj & Staelin 1983). Consequently, they may formulate more questions related to the focal domain and search for more information to answer their own questions. Furthermore, high levels of knowledge increase the benefits of new information (Brucks, 1985).
Conversely, researchers who found negative effects justify their results by arguing that knowledgeable individuals know more about the attributes of available options. Thus, they do not need to acquire more external information (Brucks, 1985). Although Bettman and Park (1980) explained this apparent contradiction (i.e., the existence of both negative and positive effects) as an inverted U-shaped relationship, other authors have found significant linear effects in prior inquiry (e.g., Brucks, 1985;Hadar et al., 2013;Kiel & Layton, 1983). Because the literature favors the positive effect (see Carlson et al., 2009;Hadar et al., 2013), we follow the logic that knowledge increases the information search.
Studies on the relationship between consumer's previous knowledge and information search have focused on the information search that occurs prior to the final decision (Brucks 1985;Hadar et al., 2013). However, consumers still search for information after finalizing their choices (Teodorescu et al., 2018) and the knowledge they have acquire before purchasing might affect this search as well. Their knowledge might increase post-decision information search because knowledgeable consumers have better awareness of where they can search for information; they can also process new information easier than non-knowledgeable consumers (Brucks, 1985).
Studies on post-decision information search have listed two main predictors of such behavior: cognitive dissonance and feelings of regret/dissatisfaction (see Donelly & Ivancevich, 1970;Ehrlich, Guttman, Schönbach, & Mills, 1957;Oliver, 2014;Shani & Zeelenberg, 2007;). For instance, Ehrlich et al. (1957) found that consumers sought post-decision information to confirm their choices. The authors argumentatively used the cognitive dissonance theory to explain this behavior. Their perspective assumed that consumers search information after deciding to minimize their experience of dissonance. Nevertheless, these authors did not empirically explore the effect of cognitive dissonance. Likewise, Donelly and Ivancevich (1970) showed that consumers did not desire to switch their cars when they had information about their purchased cars. This information reinforced consumers' decisions, minimizing cognitive dissonance.
However, the authors did not empirically test the role of cognitive dissonance either. Finally, Shani and Zeelenberg (2007) showed that feelings of regret or dissatisfaction motivate individuals to either acquire or avoid more information after purchasing.
Cognitive dissonance and feelings of regret/dissatisfaction are elicited during the post-decision stage of the consumer journey. However, consumers might not be affected only by the variables emerging at this stage after making their decisions since variables that influence consumer behavior during the initial stage of the customer journey still influence the end of this journey (Lemon & Verhoef, 2016). In fact, the costumer journey stages are linked, even though researchers in consumer behavior treat them as isolated stages (Lemon & Verhoef, 2016). Thus, we propose that consumers' previous knowledge, acquired at the pre-decision stage, increases post-decision information search: H1: The knowledge consumers hold before making their final decision will increase post-decision information search.

Interaction between consumers' previous knowledge and maximizing tendencies
Although the limited information-processing capacities of individuals make perfect maximization impossible (Simon, 1956), studies have shown that some individuals have higher maximizing tendencies than others (e.g., Goldsmith, Roux, & Ma, 2018;Luan & Li, 2017). Maximizers are usually more engaged with their decisions (Iyengar, Wells, & Schwartz, 2006). More than satisficers, they exhaustively engage in the decision-making process, by searching information and checking as many as available options they can, because they want to choose the best option (Iyengar et al., 2006). Therefore, we propose that maximizing tendencies influence the effect of consumers' previous knowledge on postdecision information search. Assuming a positive relationship between previous knowledge and post-decision information search (H1) and predicting a positive relationship between maximization and need for more information, we infer that maximizers are more likely to use their previous knowledge to its full extent as a base for information search. Therefore, we propose that the effect of previous knowledge is intensified, or facilitated, by the "natural" maximizer's tendency to search more information.
In contrast, satisficers have pre-defined parameters to judge when an option is satisfactory, or not, before making the final decision (Iyengar et al., 2006). When satisficers find an option that fits within these parameters, they cease the search and assume that the benefit of further investing in maximizing their choice does not pay off (Iyengar et al., 2006). Therefore, the more "satisficer" a consumer is, the weaker the positive relationship between her or his previous knowledge and post-decision information search. Thus, previous knowledge may decrease post-decision information search for low levels of maximizing tendencies. In tune with this, the more previous knowledge satisficers hold before the final decision, the more assured they may be that they made a choice that fits their parameters and that gathering more information after deciding will not benefit them.
H2: The effect of consumers' previous knowledge on the amount of post-decision information search will be stronger for maximizers than for satisficers.
In Study 1, we investigated the main effect proposed in H1 through an experimental study to understand the existence of the main effect and to increase the internal validity (Field & Hole, 2003). In Study 2A and Study 2B, the correlational studies, we investigate the effects proposed by H1 and H2 in a more naturalistic setting to increase the external validity (Field & Hole, 2003). Through these studies, we explore two moments during the post-decision stage: 1) when consumers have made a purchase but have not experienced the product yet (e.g., when they buy online and are waiting for the purchase delivery), and 2) when they have made a purchase and experienced it (e.g., when the get the product and have used it). In Study 2A, participants were requested to recall an online purchase they had not received yet, which is a common experience for consumers. For instance, Brazilians spent R$ 2 bn in online purchases between March 2017 and March 2018 (Redação Forbes, 2018) and Americans spent US$ 517.36 bn in 2018 (Ali, 2019). These consumers do not immediately experience their purchases after buying. Instead, they need to wait until the purchased product is delivered. During this period, they may search for more information about their purchases. In Study 2B, the participants recalled a purchase they had already experienced. In these studies, we tested the proposed interaction as well as the effect of cognitive dissonance and satisfaction.

STUDY 1: THE MAIN EFFECT Procedures
We investigated the proposed main effect through a single factor experiment with data collected on Amazon Mechanical Turk. We manipulated the objective previous knowledge to maintain greater control under the participants' knowledge. Since subjective knowledge may be affected by individuals' self-confidence (Hadar et al., 2013), by manipulating objective knowledge we could minimize the effect of other variables on the participants' knowledge.
The participants were randomly assigned to one of two conditions (absence vs. presence of objective previous knowledge). They were informed that, during a task, they must choose a "brain game" that would improve reasoning. As an additional reward for participation, they would win a password to access the game of their choice on a brain games website for free for one month. All available options were games to improve reasoning or memory, namely, Towers of Hanoi, Rotation Game, and Logic Puzzles.
This deception served as the motivation for the gamechoosing component of the experiment. We presented participants to unusual games to avoid confounds arising from the their knowledge about and familiarity with the games before taking part in the study as we manipulated objective knowledge. The "absence of previous knowledge" group received only the names of the games and no other information about them before making their decision. On the other hand, the "presence of previous knowledge" group received an informative text about each game along with their names and a representative image. It is important to stress that by providing participants with (or with no) information about the games, contrary to asking participants to report how much they knew about them, we might have generated different levels of objective previous knowledge.
After choosing the game, we asked the participants to search for online information about the games from the list without imposing any limits on them. We measured the time each participant spent searching as a proxy of their levels of postdecision information search. To ensure the participants searched for information related to the game decision, they pasted the URLs they accessed during the activity in an appropriate space.
Furthermore, we controlled for the participants' prior experience with the games and for their involvement with the decision using two items adapted from Mittal (1989) ("How important was it to you to make a right choice of this decision?"-1: "Not at all" and 7: "Extremely important"; "In making your selection of this game, how concerned were you about the outcome of your choice?"-1: "Not at all concerned" and 7: "Very much concerned") (where, r = .722).

Sample
Eighty-three participants were recruited. They spent approximately 14 minutes (on average) to complete the task. For controlling the main effect, we asked participants to report the extent to which they rushed into the information search activity to finish the task quickly ("I did my online research as fast as I could to get over with this task as quick as possible"-1: "I strongly disagree" and 7: "I strongly agree"). Individuals who checked "7" were filtered out (seven cases). Moreover, to investigate participants' attention, we included an attention check item ("How important was this choice for you? Please ignore this question and go to the item below"). Six participants failed this attention check, and thus were removed. Finally, 13 participants were eliminated from the analysis. The final sample consisted of 70 participants (42 females; M age = 37.56).

Manipulation check
Since we manipulated the objective knowledge, to control for knowledge accumulated before the experiment, participants reported their experience with the games options we presented ("I am experienced with the games options I was presented"-1: "I strongly disagree" and 7: "Strongly agree"). The participants'

Results
An independent sample t-test was conducted. Individuals

Discussion
The results suggest that, even after a final decision was made (and could not be changed), participants exposed to previous knowledge dedicated more time to search for additional information after deciding than individuals who were not exposed to this knowledge. This result supports H1. A positive relationship between previous knowledge and post-decision information search may exist because individuals with knowledge about a focal domain can formulate more questions about a topic and can process new information using less effort (Brucks, 1985).
It is important to note that the study was conducted in a controlled setting, wherein the participants' decision-making involved purchases they had not desired in reality. This may, in turn, explain post-decision information search (Jones, 2002).
In Study 1, although the participants were responsible for their choices, accounting for cognitive dissonance effects would be inadequate as the participants' decision would not have been as important to them as a real purchase, for example. Importantly, we did not investigate their satisfaction with the decision because the participants did not played the game. Satisfaction may affect consumers' information search behavior (Shani & Zeelenberg, 2007) as well. Thus, Studies 2A and 2B cover the potential effect of the interaction between consumer's previous knowledge and maximizing tendencies and the role of cognitive dissonance.
Study 2B also addresses satisfaction as a covariate. Additionally, in the following studies, we investigated consumers' subjective previous knowledge to determine whether the effect we found in Study 1 occurs when consumers report the knowledge they think they had before purchasing a product.

STUDY 2A: STUDY WITH A REAL PURCHASE, BEFORE EXPERIENCING THE PRODUCT Procedure
One hundred and fifty-one participants were recruited from Amazon Mechanical Turk. They recalled a recent, planned purchase they had made online, but had not had the opportunity to experience it yet. We asked them to complete the survey with the recalled purchase in mind. As the participants had not experienced the product, their postdecision information search behavior might have been higher than if they had already experienced it (Oliver, 2014). This allowed us to investigate the information search behavior in a more "extreme condition." We explicitly asked the participants not to think of an everyday purchase (e.g., milk, bread, etc.), because such products tend to be purchased routinely. This way, everyday purchases might attenuate information search behavior regardless of previous knowledge and maximizing tendencies.
In fact, they might not even generate experiences of strong cognitive dissonance. In this study, we did not investigate consumers' satisfaction as they had not experienced their purchases.
A single item measured consumers' subjective previous knowledge ("I had a lot of knowledge about this purchase before making my final decision"-1: "I strongly disagree" and 7: "I strongly agree") (Hadar et al., 2013). To measure postdecision information search, the participants used a slider scale to report how intensely they used specific approaches to find information about their purchase (where 0: "I did not use this approach" and 100: "I used this approach very intensely").
We show some approaches (e.g., browsing websites or talking to friends/relatives) to them according with previous research on information search behavior. The sum of the scores that the participants assigned to the items served as our measure for information search. The maximizing tendencies were measured using 13 items from Schwartz et al. (2002) (1 = "I strongly disagree" and 7 = "I strongly agree," with α = .737). The cognitive dissonance was measured through an adapted scale from Montgomery and Barnes (1993) with 11 items (1 = "I strongly disagree" and 7 = "I strongly agree," with α = .843).

Sample
As in Study 1, an attention-check item served as the filter and led 50 cases to be removed from the sample. We also checked whether participants thought of an everyday purchase, but no further case removal was necessary for this criterion. The final sample included 101 participants (53 females, M age = 36.86).

Results
We conducted a hierarchical regression to determine how the introduction of different variables affected the results in the model (Field, 2013). In the full model, previous knowledge, maximizing tendencies, the interaction term, and cognitive dissonance were included as predictors. The data met the assumption of noncollinearity (previous knowledge: Tolerance = .939, VIF = 1.065; maximizing tendencies: Tolerance = .887 and VIF = 1.127; cognitive dissonance: Tolerance = .862 and VIF = 1.160) (Field, 2013). Postdecision information search was included as the dependent variable. The models entered at levels 1 and 2 of the hierarchical regression analysis indicated nonsignificant effects of previous knowledge. This raises the possibility that previous knowledge might be sensitive to suppressors (Mackinnon, Krull, & Lockwood, 2000). We replicated the above study in Study 2B. However, this time, the participants were asked to recall a planned purchase they had already experienced. This way, we could explore the influence of satisfaction on information search behavior, and further determine whether the findings from

STUDY 2B: STUDY WITH A REAL PURCHASE, AFTER EXPERIENCING THE PRODUCT Procedures
One hundred and sixty participants from Amazon Mechanical Turk were recruited. The participants recalled a recent, planned purchase they had made and experienced. They accordingly completed a survey about their purchase. We measured previous knowledge, maximizing tendencies, and the post-decision information search as in Study 2A. However, to measure the participants' cognitive dissonance after their final decision, we used 21 items from Sweeney et al. (2000) (1: "I strongly disagree" and 7: "I strongly agree," with α = .972) because these authors presented an alternative to Montgomery and Barnes' (1993) scale (used in Study 2A). Satisfaction with the purchase was measured through 4 items adapted from Oliver (2014) (1: "I strongly disagree" and 7: "I strongly agree," with α = .960).

Sample
We used the same filters from Study 2A and removed eight cases from the sample for not following the instructions regarding the purchase to be recalled. We also filtered out participants who failed on the attention check (47 cases) by following the same procedure used in the previous studies. Fifty-five cases were removed, and the final sample consisted of 105 participants (65 females; M age = 36.96).

Results
A correlation analysis tested the relationship between satisfaction and post-decision information search (r = .068, n.s). The same procedure used in Study 2A was conducted to investigate the effect of previous knowledge, maximizing tendency, their interaction and cognitive dissonance on post-decision information search. Table 2 reports the results from the hierarchical multiple regression analysis.

Discussion
Studies 2A  previous knowledge) that in affects consumer behavior at the post-purchase stage (i.e., information search).
Our study presents managerial implications as well.
Consumers' knowledge may be a concern for business companies because consumers seek more information about goods and services before purchasing, being more knowledgeable than the sales person as well as more empowered by the knowledge they have (Gensler, Neslin, & Verhoef, 2017). Through three studies, we show that knowledgeable consumers continue searching for more information about the products they have purchased.
This way, marketers should assist consumers in this task. For instance, companies can provide more information about the purchased product (e.g., reinforce consumers' decision and give tips about how to use the purchased product, which may improve their experience with it) a few hours or days after the consumer has made a purchase. Furthermore, the information consumers find after a choice may serve as a "calibrating element" for their judgements. This may exert further influence on variables that are important for marketing performance, such as satisfaction and repurchase intentions (Shani & Zeelenberg, 2007). If business companies provide to consumers more information at the post-purchase stage, they might have more control under the information consumers are accessing.
We also present approaches used by consumers to search for information at the post-purchase phase. If retailers know where consumers search for information, they can choose the best approach to reinforce consumers' decision, and, consequently, diminish back-out behaviors (Donelly & Ivancevich, 1970) and feelings of regret (Shani & Zeelenberg, 2007).
Finally, our study is not without limitations. First, we could not replicate the findings from Study 1 across Studies 2A and 2B. Nevertheless, we found a marginal significant effect of previous knowledge in the model at level 3 after introducing the maximizing tendencies, the interaction term, and cognitive dissonance. This may indicate the suppression effect of other variables (Mackinnon et al., 2000) when we investigate consumers' previous knowledge in a more naturalistic condition.
Across the three studies, the effect of previous knowledge has the same direction as we proposed in H1. Second, in Study 2B, we did not measure feelings of regret, which may increase or decrease information search behavior (Shani & Zeleenberg, 2007). We used a scale to measure satisfaction, which may be an indicator of individuals' feelings of regret, but it is not a measure of it. Finally, we did not explore the effects of postdecision information search on participants' satisfaction with the decision and their repurchase intentions. Further studies should investigate these variables because they are important to improve marketing performance (Brega, 2018).
This research amplifies the understanding of consumers' post-decision experience, diverging from previous marketing literature that focuses mainly on variables such as satisfaction, repurchase intention, and service failures topics (e.g., Bach & Kim, 2012;Wirtz et al., 2007). Differently from this literature, we investigated the post-decision information search behavior, which may strongly affect consumers' actions toward brands, goods, and services (Oliver, 2014).

NOTE OF APPRECIATION
We would like to thank CAPES for the financial support to research.