Abstract
Existing research treats sales performance as a series of discrete, independent events rather than a series of sales attempts with intertemporal spillover across these attempts. This research examines whether there are systematic short-term trends (“momentum”) in sales performance. To do so, the authors use the clumpiness approach to examine the existence of sales momentum in a high-frequency call-level data set obtained from two call centers of a large European firm. They further investigate the effect of positive (negative) momentum, or the positive (negative) deviation from the long-term expected performance on subsequent sales performance. Exploiting the differences in the social environment of the call centers, the authors find that the social working environment mitigates the harmful effect of negative momentum and sustains positive momentum. Further, they demonstrate that calls made midday, early-week, and late-week boost performance by mitigating the adverse effects of negative momentum. The findings suggest that monitoring sales performance can help managers detect momentum and use timely interventions to enhance sales productivity. Managers can also leverage momentum by creating a more social working environment to optimize overall salesperson performance.
Researchers have long been interested in understanding salesperson performance (Albers, Mantrala, and Sridhar 2010; Franke and Park 2006). Much of this work focuses on identifying the drivers of sales success (Kishore et al. 2013); yet salespeople fail far more often than they succeed. A recent study of more than 1,000 sales organizations finds that an average salesperson has 11.9 sales conversations per day, but only 4.8% of these result in meaningful sales opportunities (InsideSales 2017). However, sales research typically does not explain how to overcome the influence of failure. Furthermore, most research on sales assumes that sales events are discrete and independent (Verbeke and Bagozzi 2000), although some work questions this assumption (Marinova, Singh, and Singh 2018; Misra and Nair 2011; Patil and Syam 2018). This article extends the stream of sales literature that examines the interdependence of sales events.
In this study, we systematically explore “sales momentum,” defined as the distinct short-term trends in salespeople’s performance outcomes above or below their long-term expected levels. We find that salespeople’s performance does exhibit momentum. In other words, salespeople experience periods of temporary elevation or decline in sales call outcomes compared with their overall expected performance. Such deviations are likely not just a statistical possibility of consecutive success or failure, but rather constitute a systematic clustering of outcomes over a short period (e.g., a run of successes with a small number of failures, or vice versa). Thus, given a period of sales calls, a salesperson with more successes (failures) than their long-term expected performance is said to be experiencing positive (negative) momentum.
To understand the perception of momentum and how it is managed in practice, we conducted a preliminary study with 160 salespeople (for details, see Web Appendix A), in which we gathered information on salespeople’s beliefs about momentum, their experiences, and whether their organizations actively managed momentum. We found that most salespeople believed they had experienced both positive (86%) and negative (65%) momentum. The salespeople also presumed that their organization tried to both identify (67%) and actively manage (56%) their momentum. However, no consensus emerged when we asked salespeople to report what their organization did to manage momentum. This suggests that there is no clear understanding of how to manage momentum, with most approaches based on intuition and guesswork. Some salespeople suggested actions such as trying to boost spirits, giving encouragement, or being positive, while others reported enforced breaks and meetings.
Thus, although sales practitioners have intuitive beliefs about sales momentum, there is not a great deal of systematic investigation into this phenomenon or strong managerial insights to estimate its impact and guide sale practice. To fill this research gap, we examine sales momentum and its impact on salesperson performance. Understanding whether specific working conditions can enhance momentum and its effects would further help manage momentum and augment salesperson performance. In particular, salespeople in dedicated sales environments make continuous calls throughout the day, and factors such as (1) the design of the sales environment, (2) individual worker characteristics, and (3) timing of the calls may augment the effect of momentum. To this end, we examine where the salesperson should work (i.e., a more or less social working environment), who should make the call (i.e., male or female salespeople), and when the call should be made (i.e., time of day and day of week) to strengthen or weaken the effect of momentum. Accordingly, we address three research questions:
Do short-term trends in salespeople’s performance systematically deviate from their long-term expectations (i.e., does momentum exist)? How does this momentum influence subsequent performance? How do the social working environment, salesperson gender, and timing of the call moderate the effect of momentum on performance?
We examine our research questions in the context of inside sales, in which salespeople mostly reach customers remotely rather than face-to-face (Martin 2013) because we expect momentum to be more visible here than in a field sales context. Inside salespeople in dedicated facilities engage in transactional selling activities that can be actively managed through continuous monitoring. That is, they make multiple sales pitches in a short period, which creates a fitting environment to detect momentum.
We use a novel disaggregate call–level data set from a large European firm that operates two call centers. Salespeople in these call centers make outbound sales calls to sell an identical product. The high-frequency transactional data capture key details of each sales call and its outcome, enabling us to identify positive and negative salesperson momentum. With these data, we can also estimate the likelihood of success in the subsequent sales call and suggest ways that managers can augment (mitigate) the effect of positive (negative) momentum in their sales force.
We explore momentum using two related analyses. In Analysis 1, we formally test for the existence of momentum by examining short-term trends in salespeople’s performance that deviate from their long-term expectations. To effectively capture sales momentum, a method should be able to capture moment-to-moment momentum, identify exactly when salespeople experience momentum, and account for their inherent low absolute success rate. We follow Zhang, Bradlow, and Small’s (2013) clumpiness approach, which tests for the existence of systematic clustering of a type of outcome over time, consistent with our definition of momentum. Intuitively, the method captures momentum behind sequences of outcomes that may appear completely random. By examining variation or disorderliness in a sequence of outcomes compared with the expected level of outcomes, the clumpiness approach captures salespeople’s success rates below 50% and shows whether they were experiencing momentum as they made additional calls. The results provide robust evidence for the existence of momentum in sales performance.
In Analysis 2, we examine the relationship between momentum and sales performance using a fixed-effects logistic regression to test how momentum influences subsequent performance. We find a strong influence of momentum on the probability of success in a salesperson’s next call. Specifically, we find that positive (negative) momentum enhances (decreases) the likelihood of selling in the next call. This finding can help managers estimate the likelihood of success in the next call, given the momentum state of the salesperson.
In addition, we explore the consequences of momentum in the sales context to provide implications regarding the type of work environment that can maximize salesperson performance. We focus on examining the moderating role of the social environment by exploiting a natural variation in the physical setup at two of our focal firm’s locations. The firm independently carried out a quasi-experiment to investigate the effect of social interaction on performance. In one location, referred to as the social call center (SCC), all salespeople were located in the same large space, and there were facilities such as a relaxation room and a communal kitchen. The SCC was specifically designed to enable greater social interaction than the firm’s other location. The other location, referred to as the nonsocial call center (NSCC), was a repurposed office building in which salespeople were spread across different small rooms, and there were no communal relaxation facilities as a result of limited space. We find that being in the SCC weakens the harmful effect of negative momentum and enhances the favorable effect of positive momentum on sales performance. Considering the effect of SCC, we propose a logic for a decision support system (DSS). We use simulation to test the effectiveness of the proposed DSS and find that salespeople in our data would have performed 18.4% better if the system had been in place.
We explore the moderating effect of salesperson gender, time of day, and day of week to provide further guidance on managing the effect of momentum and to discover the conditions in which performance in the next call is more (less) likely to be influenced by momentum. Unlike previous research that finds gender differences in behaviors (Cohen-Zada, Krumer, and Shtudiner 2017; He, Inman, and Mittal 2008), we find that gender does not significantly influence the relationship between momentum and sales performance. However, making calls midday, early, and late in the week mitigates the harmful effect of negative momentum.
The structure of this article is as follows: First, we describe the background literature on momentum and present details of our empirical context and data. Second, we conduct Analysis 1, which examines the existence of momentum using the test of clumpiness, and then Analysis 2, which assesses the influence of momentum on sales performance and its boundary conditions using the fixed-effects logit model. Third, we provide additional information about sales momentum and the financial implications of controlling for momentum, explain the logic for how a DSS can be implemented to detect and control momentum, and discuss the downstream consequences and long-term impact of momentum. Finally, we conclude with insights for theory and practice and directions for further research.
Literature on Momentum
Introducing the concept of momentum to the social sciences, Adler (1981) defines momentum as a phenomenon in which the probability of success in the present event is influenced by the outcome of the preceding event. Similarly, Gilovich, Vallone, and Tversky (1985) explore the existence of the “hot hand” in sports. Momentum is related to a hot hand or “streakiness” but differs in important ways. Hot hand literature examines the phenomenon as a statistical probability for the occurrence of consecutive events with the same outcome; by contrast, momentum (as used herein) examines the (short-term) trends in events (here, sales performance) that deviate significantly from long-term expectations. Momentum and similar phenomena have been explored in sports and gaming-related research for some time but only more recently in marketing and organizational contexts.
In the marketing context, research on shopping momentum shows that an initial purchase enhances the probability of a subsequent purchase (Dhar, Huber, and Khan 2007). Studies on binge-watching find that viewers commonly consume several episodes of the same television series in a condensed period (Schweidel and Moe 2016; Zhang, Bradlow, and Small 2013). Rather than momentum, Zhang, Bradlow, and Small (2014, p. 196) call this phenomenon “clumpiness” or the “degree of nonconformity to equal spacing.” Clumpiness examines the systematic clustering of an outcome over time, consistent with our definition of momentum. These studies on momentum and similar phenomena illustrate how an event in one period can lead to another event with a similar outcome (e.g., purchase, consumption, successful performance) in the following period; however, they have limited applicability to the sales context in which salespeople make multiple sales calls in a condensed period. For example, studies investigating momentum in consumption tend to examine situations in which customers make multiple purchases, restricted by time and budget. By contrast, salespeople are involved in an incentivized work task, not a consumption context. Patil and Syam (2018) examine a sales context that is similar to ours and find that high-performing salespeople can experience consecutive months of well-performing periods. However, the authors study momentum at an aggregate level by comparing quota achievement information for each month. This finding raises the question of whether sales momentum exists only at an aggregate level or whether salespeople can also experience momentum moment-by-moment in their daily sales tasks. In addition, Patil and Syam assess heterogeneity in momentum effects by salespeople’s performance levels, which calls for further research to examine other factors that moderate the effect of sales momentum.
Studies exploring the existence of momentum in sports and organizational theory also have limited implications for sales momentum because they assume that two or more players compete for victory in a single or sequential set of events (Cohen-Zada, Krumer, and Shtudiner 2017; Lehman and Hahn 2013). However, salespeople in our context do not compete for the same customer, as tennis players compete to win a point, and salespeople do not compete with customers other than in a metaphorical sense. Furthermore, in many studies momentum is a self-reported perception that is observed with surveys or qualitative approaches (Kerick, Iso-Ahola, and Hatfield 2000; Markman and Guenther 2007), so there is a lack of information on how to detect sales momentum objectively and manage its effects.
To this end, we use a formal statistical test on sales call outcomes to show that momentum indeed exists in sales call tasks. Building on prior studies suggesting the existence of sales momentum, we examine the conditions of outgoing sales calls that may influence the effect of positive and negative momentum on performance. By leveraging our extensive data, we test a moderator (salesperson gender) of the effect of momentum previously explored in sports. Cohen-Zada, Krumer, and Shtudiner’s (2017) study on professional sports players indicates that men’s performance is significantly affected by momentum, whereas women’s performance is not. The authors speculate that this effect is caused by differences in testosterone, which may increase following victory and enhance performance in sports and business contests (Verbeke and Masih 2000). We test to determine whether a similar effect would be observed in the sales context.
To assist managers’ understanding of the effect of sales momentum, we examine two additional factors (sales call environment and timing of the call) that are not explored in previous momentum studies. First, we explore whether the social working environment impacts the effect of momentum on performance. Salespeople in a more social working environment are, by definition, more likely to observe and/or interact with their coworkers. We suggest that the effects generated by these inherent features of a more social working environment moderate the momentum effect on sales performance through spillover of momentum perceptions (Moesch and Apitzsch 2012) and interruption of the workflow (Jett and George 2003). Second, we test how the timing of the calls impacts the effect of momentum on performance. Chronobiology research discusses biological rhythms that may explain differences in salespeople's behavior. For example, research often invokes circadian rhythm–based rationales for the influence of time of day on human cognitive and behavioral variables, such as working memory, alertness, and sustained attention (Kanuri, Chen, and Sridhar 2018). To account for the possible influence of biological rhythms on the effect of momentum, we explore the time of day and day of week the call is made.
Examining the moderating effects of the salesperson gender, sales call environment, and the timing of the calls enables us to provide a more complete picture of the effect of momentum on sales performance. In addition, it helps provide recommendations on how firms can better manage momentum to maximize the potential for positive effects and minimize the potential for negative effects on sales performance. Table 1 presents an overview of selected literature on momentum from various fields.
Selected Literature on Momentum.
Notes: N.A. = not applicable.
Research Design
Choice of Setting: Inside Sales Force
Momentum is most likely to occur in a context in which salespeople make multiple sales attempts in a short period. Therefore, we test our hypotheses on an inside sales force, which is an increasingly dominant channel to connect with customers (Martin 2013). Approximately 47.2% of salespeople in the United States were considered inside sales representatives in 2017 (InsideSales 2017). Inside salespeople reach customers remotely rather than through face-to-face meetings (Martin 2013) and, in general, work in dedicated facilities, augmented by technology and actively managed by continuous monitoring of performance. Inside sales forces often have transactional roles focused on opening opportunities and/or selling less complex products. In such cases, sales forces in dedicated call centers are more cost efficient for firms because they can execute many calls without geographic restrictions.
Inside sales forces have characteristics that distinguish them from traditional field sales forces and make them a good choice to observe momentum. First, inside salespeople have frequent face-to-face contact with their peers and managers, so the social effects within the inside sales force have a high potential to influence performance. Frequent face-to-face contact also enables inside sales managers to measure real-time performance. Second, inside salespeople make sales calls at a far higher rate than field salespeople. For example, whereas a typical field salesperson in a specialty market in the pharmaceutical industry makes approximately 13 calls per day (Pharmaceutical Executive 2010), an inside salesperson makes approximately 121 calls per day (Richard 2015). These factors pose distinct challenges for inside sales managers. Therefore, principles applied in traditional field sales contexts tend to have limited applicability in the inside sales context. Even so, researchers pay significantly less attention to inside sales force management than to traditional field sales force management. Our choice of setting (inside sales force) helps fill this research gap, and provides the first empirical evidence for momentum in sales performance.
Data
Inside sales forces commonly work in dedicated call center facilities, which are not necessarily connected with the rest of the organization. Indeed, many firms now specialize in running inside sales campaigns on behalf of third parties. We use novel, high-frequency transactional data from one of these firms, a large European sales specialist, to examine momentum. The empirical context of our analysis is based on an outbound sales campaign run over a three-month period, employing all salespeople in the firm. All salespeople were responsible for selling an identical product (a hard copy reference work) with the same price (€4.86, equivalent to US$6.44 at the time of study), and customers could not purchase this product through any other channel. This product category is challenging for salespeople because the product is not usually sought after and lacks a large market share and demand, especially in the digital era. Therefore, salespeople must use significant persuasion to make sales. Our context is also significant because no other marketing interventions (e.g., sales promotion, advertising) were undertaken during the studied period, so no other factors influenced demand. This makes the setting highly appropriate for our research questions, as demand for the product is minimized and the sales outcome is largely determined by the salesperson’s effort.
The individual call–level transactional data consist of 74,060 sales calls (after data cleaning) undertaken by 113 salespeople during the campaign, and the logs associated with the calls. The log of each call contains a time stamp and the outcome of the call (“successful” or “failed” sale). As is typical in many modern call centers, each salesperson is equipped with a computer terminal. The system automatically dials the customer by drawing a random number from a large database of potential customers (i.e., cold calls), and the call is then directed to a salesperson. Customer assignment is randomly generated, and salespeople do not have control over whom they can call. After completing the call, the salesperson enters the details of the call outcome in a customer relationship management system. Then, the system automatically dials another random number. The system also enabled us to capture, as a control, the psychological state (confidence) of the salesperson before each call, by asking, “What is your level of confidence in having a successful sale in the next call?” Salespeople could choose 0%, 20%, 40%, 60%, 80%, or 100%. This entire process was automated through the computerized system protocol, and salespeople could not avoid answering the question. 1
Differences Between the SCC and NSCC
The focal firm operates two call centers that are located in different cities in the same country. Salespeople work at the center closest to their homes. Both cities are highly entrepreneurial and are important business and industrial centers, containing multiple universities. Web Appendix B summarizes the similarities and differences of the call centers. All salespeople were recruited using the same process and were employed under the same terms, with identical incentive structures (unchanging over the course of the study) comprising a base hourly wage and a percentage commission. For each sale, salespeople earned a commission equivalent to US$.25. All salespeople were paid monthly, and there was no quota or target to gain a bonus. All salespeople received the same initial training course, which consisted of three days of theory and two days of practice, before being placed in one of the call centers. Both centers operated under the same hierarchical structure, with a manager and team leaders, who, in turn, supervised salespeople. All salespeople worked from the same database of potential customers.
The natural variation we exploit to assess the influence of social factors is the difference in the physical arrangement and accommodation of the salespeople at the two call centers that give differing potential for social interactions. 2 In the SCC (Figure 1, Panel A), salespeople work in a single large room. Each salesperson has a desk in an open space that is shared by all salespeople in the SCC. The selling environment is modern but crowded, and salespeople are provided with comfortable facilities for relaxation and socialization. These facilities are designed to facilitate social interactions between salespeople and consist of couches, armchairs, and a fully equipped kitchen. In the NSCC (Figure 1, Panel B), social interaction between salespeople is limited by the facilities’ space. Specifically, the salespeople work in many individual small rooms on different floors. The call center does not have dedicated socialization or relaxation facilities other than two small kitchens with a few chairs, because of the site’s size and physical structure.
Floor layout of SCC and NSCC.
Table 2 provides descriptive statistics and compares the characteristics of the two call centers. The number of salespeople in our data is sufficiently large (n = 113), and each salesperson made approximately 655 calls during the campaign period (21 calls per day). The NSCC had a total of 35,420 outgoing calls, and the SCC had a total of 34,640 outgoing calls. Observation dates for both call centers were the same except for two days when the NSCC was closed while the SCC was open. We removed these days from our data set, which left 31 observation days in both call centers. Excluding no-calls (e.g., missed calls, wrong numbers), the SCC had a higher success rate (28.12%) than the NSCC (15.06%), with an average of 20.49% across both. Although we find that salespeople in the NSCC and SCC do not differ in the number of intercall time less than 1 minute, which suggests this time frame likely captures salespeople making back-to-back calls, they differ in most other factors related to work efficiency. Our observation of the data reveals that salespeople in the SCC took a longer time in between calls (average of 47 seconds in NSCC vs. average of 54 seconds in SCC; p < .01), more frequent and longer (1–4 minutes) intercall times (average of 9.575 calls in NSCC vs. average of 12.662 calls in SCC; p < .01), and more frequent breaks each day (average of 3.8 breaks in NSCC vs. average of 4.1 breaks in SCC; p < .01). More time between calls also led salespeople to make fewer calls per minute (average of .634 calls per minute in NSCC vs. average of .539 calls in SCC; p < .01). We also find that it was more likely in the SCC that if one salesperson is taking a break, at least one other salesperson is also likely taking a break (average of 70% in NSCC vs. average of 77% in SCC; p < .05). These observations suggest that salespeople in the SCC are more likely to be interrupted in between calls and to interact with one another, thereby yielding a social effect. We exploit this natural variation in physical arrangement of the call center to examine the moderating effect of momentum on sales performance in the next period in Analysis 2.
Descriptive Statistics and Comparison of NSCC and SCC.
Notes: N.A. = not applicable.
Analyses and Results
As noted, we examine sales momentum in two related analyses. In Analysis 1, we assess the existence of sales momentum using individual sales call outcomes. In Analysis 2, we use the momentum calculation from Analysis 1 to assess the impact of momentum on the likelihood of a sale in the next period and explore boundary conditions of the momentum effect.
Analysis 1: Assessing the Existence of Momentum
We begin by providing visual evidence of momentum in the data and then presenting simple analyses to document the transition between calls. The heat map in Web Appendix C depicts the outcomes of a sample salesperson, in which we observe the clustering of sales (and no sales) at multiple times (e.g., day 16 and day 31), which could represent positive and negative momentum, respectively. We observe similar patterns across our data set.
Table 3 represents sales momentum using a first-order Markov transition probability between the outcomes of consecutive calls. The Markov transition matrix provides the probability of a salesperson’s transition from one outcome to the other in successive calls. We find that the probability of having a successful sale after a successful sale (27%) is greater than the probability of having a successful sale after a failed sale (19%). Similarly, the probability of having a failed sale after a failed sale (81%) is greater than the probability of having a failed sale after a successful sale (73%). These findings suggest an inherent “stickiness” in the performance of a salesperson, but they do not provide systematic evidence for the existence of momentum. This is because Markov transition probability infers possible dependence of outcome in two periods only and does not account for salespeople’s low success rate. Thus, we discuss the requirements for a suitable measure to systematically assess sales momentum in the next subsection.
First-Order Markov Transition Probability in Sales Calls.
Methodological requirements for sales momentum
In our context, each salesperson makes outgoing calls, which are observed multiple times a day across various days. Most salespeople make hundreds of calls and achieve at least one successful sale per day. We consider the regular and irregular clusters of failed and successful sales to examine momentum. Given this context, we highlight the preconditions for the method to assess sales momentum. First, the method should identify when salespeople are experiencing momentum. Such an understanding can provide implications to managers on how to mitigate or enhance the momentum effect. However, most prior studies on momentum (Patil and Syam 2018; Schweidel and Moe 2016) infer that momentum exists but cannot distinguish when it is happening. For example, Schweidel and Moe (2016) examine Hulu viewers to infer momentum (“binge” behavior) from customer viewing behavior but are unable to tell precisely when the viewer is experiencing momentum. Not knowing when individuals are experiencing momentum makes it difficult to control for momentum effects and lacks managerial value in the sales context.
Second, in our context, the unit of analysis for sales momentum must be at the most disaggregated, individual call level to capture the exact nature of momentum in real time. It is necessary to capture momentum at the individual call level because salespeople make multiple calls within a short time frame; thus, experienced momentum may fluctuate from call to call. Confirming this point, prior studies examining random events find that outcomes are viewed in “local” subsequences in a longer but finite “global” stream of events (Hahn and Warren 2009; Warren et al. 2018). Lehman and Hahn (2013) also examine moment-to-moment momentum but not in a way that is directly applicable to sales. They use American football data to operationalize positive (negative) momentum by the number of wins (losses) when the team is in a winning (losing) streak. However, salespeople have a significantly lower chance of success (or winning) than professional football players, who have a relatively similar chance of winning and losing on average. 3
Finally, the method must accommodate a lower than 50% chance of having a successful sale because salespeople fail far more often than they succeed. Therefore, capturing sales momentum as consecutive runs of outcomes is not ideal. 4 A suitable method to examine the existence of sales momentum thus needs to account for clusters of outcomes (e.g., success or failure of the sales call) in which one outcome is more likely to occur than the average level. To an observer, the salesperson outcomes from each sales call may seem independent of one another, and outcomes from a group of calls may also seem like a random sequence. Therefore, our objective is to find a system behind what may appear as disorderliness, and formal statistical assessment can help us fully understand this system. Given the requirements for sales momentum analysis and assumptions, we adopt the test of clumpiness (Zhang, Bradlow, and Small 2013, 2014), which captures momentum consistent with our definition, has elegant statistical properties (see Web Appendix D), and provides a framework for statistical inference. In the next subsection, we provide a summary of the test of clumpiness and discuss our assumptions to accurately capture sales momentum in our context.
Test of clumpiness
Clumpiness is an entropy-based measure, where entropy refers to the variation (disorderliness) of outcomes in a sequence of events (i.e., sales calls). We use the outcome of the individual sales calls to represent whether salesperson j on day d was successful (
We make two assumptions to accurately capture sales momentum using the test of clumpiness in our context. First, in line with prior research on random events, we assume that salespeople have a short-term memory of experiences (Farmer, Warren, and Hahn 2017) and that the limited capacity of the window of experiences slides one event at a time through a finite number of experiences (Hahn and Warren 2009; Warren et al. 2018).
6
Therefore, we test sales momentum in a rolling window by using a fixed-window size
7
that moves sequentially from the beginning to the end of the sample by adding the next observation from the sample and dropping one observation from the end of the window. We illustrate clumpiness testing with a rolling window through an example. Consider Salesperson B who makes 11 calls on Day 1 with successful sales occurring in the first, eighth, ninth, and tenth calls (i.e.,
Second, we assume that a salesperson has a fresh start every day. When salespeople finish their daily task of making back-to-back outgoing calls to customers, they have time off from work to rejuvenate. Their time off from work may be as short as 15 hours (e.g., leave work after the last outgoing call at 6
Next, we compare
Results
We run the test of clumpiness in a rolling window of nine calls to determine whether the next call was made while the salesperson experienced momentum. We find that 1.35% of calls were made while salespeople experienced momentum. 10 These momentum observations are spread out across multiple salespeople. We classify salespeople as having experienced momentum if they made at least one call while experiencing momentum. Of the 113 salespeople in our data set, we find that 81 (71.68%) experienced momentum at some point. When salespeople experienced momentum, it typically lasted for one or two more calls, then the patterns of outcomes become less clumpy.
As Table 4 shows, 58% of salespeople made .1% to 1.9% of calls while experiencing momentum during the campaign period. We also examine the distribution of momentum calls by day and hour (Table 5) to explore how momentum experience is distributed across days and hours worked. Salespeople who experienced momentum did so at least once on an average of 31.0% of the days worked. Most salespeople made at least one call under momentum on 10.0% to 19.9% or 30.0% to 39.0% of the days worked (Table 5, Column 3). Given that salespeople experienced momentum in a specific hour, salespeople made approximately 7.89% of calls in that hour while experiencing momentum.
Salespeople's Momentum Experience.
Salespeople's Momentum Experience Across Days and Hours.
Robustness assessments
We run a nonhomogeneous hidden Markov model (HMM) to show that finding the existence of sales momentum is not bound to the method used. We also assess the robustness of the choice of the rolling window size by running our tests of clumpiness and the logit model with different sized windows.
Alternative method to capture sales momentum: The clumpiness metric has been thought of as an alternative method to HMM to capture bursts of activity separated by less active periods (Zhang, Bradlow, and Small 2013). The HMM specifies outcomes to be related to the states of the process and models the transition among the latent states (Netzer, Lattin, and Srinivasan 2008). Specifically, we use a first-order Markov model, in which we assume that future states, at t + 1, depend on the present state, t, but not on any other states preceding it. With HMM, we can glean insights into how the underlying states evolve as salespeople progress through successful or failed calls. We find evidence of momentum through the transition matrix (Web Appendix G, Table W2) by examining a strong latent-state persistence from the previous call to the next call.
Alternative rolling window for momentum measure: We use a rolling window of nine calls throughout the analysis. To enhance robustness of our findings, we also explore alternative rolling windows. We run the test of clumpiness using rolling windows of 11 and 13 to find that .89% and 2.51% of calls, respectively, were made while the salesperson was experiencing momentum. The distribution of percentages of calls experienced by the salesperson was also largely similar across multiple rolling window sizes (Web Appendix G, Tables W4 and W5). As an additional robustness check, we assess across-day momentum by using outcome information from all calls made by salespeople in a day to determine whether they experienced momentum during a specific day (Table W6). We find that 37 of 113 (33%) salespeople experienced across-day momentum at least once.
Analysis 2: Assessing the Impact of Momentum on Sales
In Analysis 2, we assess the differential impact of positive and negative momentum on the outcome of the subsequent sales call using the previously calculated clumpiness scores from Analysis 1.
Measures
The dependent variable, outcome of the call, is a binary variable that captures whether the salesperson made a successful sale for a call. To explain our dependent variable, we include positive momentum and negative momentum. We operationalize positive and negative momentum following Lehman and Hahn (2013). Positive momentum takes a value equal to the clumpiness score when the success rate within nine consecutive calls is greater than or equal to the salesperson’s overall success rate and takes a value of 0 otherwise. Similarly, negative momentum takes a value equal to the clumpiness score when the success rate within nine consecutive calls is less than the salesperson’s overall success rate and takes a value of 0 otherwise. With this procedure, positive and negative momentum are mutually exclusive. We illustrate this with an example in Web Appendix J.
To see how the effect of positive and negative momentum differentially influence outcomes, we explore four types of factors: social, gender, time of day, and day of week. Social is an indicator variable to differentiate salespeople in the SCC from those in the NSCC. Female is a binary variable indicating salesperson’s gender. MidDay, EarlyAfternoon, and LateAfternoon are time-of-day indicator variables. EarlyWeek and LateWeek are day-of-week indicator variables. We interact social working environment, gender, time, and day variables with positive and negative momentum variables to understand factors that influence momentum.
We also add a set of control variables that may influence the outcome of each sales call. These include (1) salesperson self-reported precall confidence, because salespeople’s judgments about their likelihood of making a sale may be correlated with the effort they put into selling and therefore directly related to the outcome; (2) number of calls received by the customer, as customers who have already requested a callback may be more likely to make a purchase; (3) experience, to capture the impact of learning-by-doing over the tenure of the salesperson in the sales campaign; (4) number of breaks, to control for the salesperson's willingness to work for the entire day, as their motivation to work and expend effort can be related to outcome; (5) time since the last break, to control for salesperson’s (reinvigorated) mindset from the prior break and its potential influence on outcome; and (6) time spent on prior break, to control for the impact of length of breaks on outcome. We define our notation of variables and measures in Table 6.
Variable Descriptions.
Descriptive statistics of variables
The correlation table in Web Appendix K shows that none of the variables are highly correlated, which suggests no significant collinearity problems. Confidence has a low correlation with positive (r = .007) and negative (r = −.043) momentum, which suggests that salespeople are not always conscious of their momentum states, further emphasizing the importance of objective measurement of momentum rather than relying on perceived momentum, as in most prior work. The mean of positive momentum is .06, and the mean of negative momentum is .26. The comparatively low mean of positive momentum is consistent with the low sales rate (20.49%), which is typical of outbound inside sales. Salespeople fail more often than they succeed; therefore, in many cases, positive momentum is 0, while negative momentum carries a value. Web Appendix K provides correlations and descriptive statistics of the variables used in the study.
Identification challenges and strategy
The objective of Analysis 2 is to examine the differential effect of negative and positive momentum on the likelihood of a sale in a subsequent sales call. We assume the customer’s decision is based solely on salesperson factors as customers in our context make purchase decisions based on their interaction with the salesperson, not by demand-side factors such as advertising and promotion. Therefore, we use logistic regression to examine the effect of positive and negative momentum on the outcome of each call as follows:
Salespeople’s strategic break-taking behavior is the first empirical challenge. Our focal firm has a system for break allocation that gives salespeople some flexibility. Specifically, salespeople are allowed to take a five-minute break every hour. Breaks cannot be accumulated, but salespeople can choose when to take their hourly break. During this five-minute break, salespeople typically go to the bathroom or get some air. Salespeople may strategically take a break depending on their performance because they have some control 11 over their breaks. Although it is unlikely that salespeople will strategically take a break when they think they are performing better than usual (i.e., when in positive momentum), 12 a frustrated salesperson may strategically take a break when not performing well (i.e., when in negative momentum). Therefore, it is possible that negative momentum is biased while positive momentum is not.
In addition to salespeople’s strategic break-taking behavior that may bias negative momentum, other unobservable factors may impact momentum and influence performance. Even with an extensive set of control variables employed to account for the heterogeneity in salespeople’s performance, capturing all factors that may influence momentum is not possible. For example, prior research indicates that salespeople strategically manage their performance depending on prior outcomes and sales quotas and ceilings (Misra and Nair 2011), which are unobserved. Although salespeople in the studied firm were not assigned quotas or ceilings, they were verbally reminded by managers to try to make at least one sale per hour. Such pressure may influence negative momentum to be biased and have a weaker effect.
Given the previous discussion, regressing positive and negative momentum on the salesperson’s outcome in the next call with logistic regression may yield biased estimators. To control for endogeneity between momentum variables and the outcome, it would be ideal to conduct a field experiment in which observed and unobserved strategic behaviors of salespeople are controlled. For example, managers could control any incentive or pressure that would spark a temporary rise in performance. Managers could also randomly assign breaks to salespeople to help reduce their inclination to walk away from the task when they are experiencing negative momentum. However, such a randomized experiment was not feasible for two reasons. First, it is not realistic to take away the control from the salespeople to take a bathroom break when needed. Second, the focal firm was involved in its own natural experiment to examine the role of social effects in performance and did not want to interfere with salespeople’s natural behavior. Therefore, we alleviate endogeneity concerns empirically.
We correct for our negative momentum variable by including a copula term in our model to directly capture the correlation between the endogenous variable and the error of the regression (Park and Gupta 2012). Although classical endogeneity correction methods rely on instrumental variables to partition the endogenous variable into exogenous and endogenous components, the copula method does not require instrumental variables. Instead, the copula method assumes a nonnormal endogenous variable. We run the Shapiro–Wilk normality test (Datta, Ailawadi, and Van Heerde 2017) to confirm the nonnormal distribution of negative momentum. We find that the distribution of negative momentum is significantly different from the normal distribution (W = .1844, p < .01).
Our identification strategy captures the joint distribution of negative momentum and the error term using a copula, which is generated using the nonparametric density estimation method, and then finds the marginal distribution function of negative momentum. We include the copula term in the regression to obtain consistent estimates that do not suffer from endogeneity problems. In line with previous studies in marketing literature using the copula approach, we outline specific steps.
To generate the copula term, 
We include
Results
Table 7 shows the results of the estimation of the salesperson’s outcome in the next call. Model 1 (Table 7, Column 2) is the baseline model without copula correction term. Model 2 (Table 7, Column 3) shows the role of positive and negative momentum in explaining for salesperson’s outcome in the next call. As can be inferred from the table, positive momentum
Regression Analysis for Likelihood of a Sale.
*p < .10. **p < .05. ***p < .01.
Notes: Numbers reported represent coefficients; numbers in parentheses represent bootstrapped standard errors.
Model 4 (Table 7, Column 5) is our main model, which explores additional factors that may influence the effect of momentum on performance. Specifically, we examine the impact of three potential factors: (1) gender, (2) time of day (morning, midday, and early and late afternoon), and (3) day of week (early, midweek, and late week). We find that positive momentum
The interactions between both forms of momentum and the type of call center are significant (positive momentum:
We find that day and time of the call influence the effect of positive and negative momentum on performance. Compared with midweek (Wednesday), the effect of negative momentum is significantly different in the early week (Monday, Tuesday;
Finally, we find that gender does not significantly impact the effect of momentum on performance in the sales context. The insignificant effect of gender is different from prior studies on professional sports players that find a varying effect of momentum by gender, theorized to be a result of differences in testosterone (Cohen-Zada, Krumer, and Shtudiner 2017). The effect of gender in driving sales momentum may not be significant because sales is a more sedentary repetitive activity in which testosterone or risk-seeking differences in gender may not be activated.
Robustness assessments
We assess the robustness of our results using another method as an identification strategy. In addition, we evaluate the robustness of our key moderating factors: social working environment variable and time-of-day variables.
Alternative identification strategy: To assess the robustness of our findings with the copula method, we introduce an instrumental variable and use a control function approach (Petrin and Train 2010). We use the number of calls made since the last involuntary break, which is a period off from putting effort into selling due to no-calls (e.g., missed calls, wrong numbers), as an instrumental variable for negative momentum. The control function approach includes predicted first-stage residuals as additional regressors in second-stage estimation (Rutz and Watson 2019). Specifically, we derive a proxy variable (i.e., predicted residuals) that conditions on the part of negative momentum that depends on
Alternative assessment of social working environment: In our previous analysis, we found that the social effect (i.e., being in the SCC) augments the effect of positive momentum and ameliorates the effect of negative momentum. In other words, salespeople in the SCC were able to break negative momentum faster than those in the NSCC. This is because of the different magnitudes of negative (positive) momentum generated at the two call centers. Our challenge is to examine the social effect associated with (1) weaker negative momentum, which makes it easier for salespeople to “snap out” of negative momentum, and (2) stronger positive momentum, which makes salespeople “stickier” in positive momentum. Therefore, we apply matching methods that balance treatment and control groups according to observables (Avery et al. 2012). Matching methods enables us to isolate individual salesperson characteristics that influence the magnitude of momentum. Therefore, we can test the moderation of the social effect of negative (positive) momentum more robustly. We find that the social effect is associated with weaker (stronger) negative (positive) momentum in the SCC than the NSCC, independent of any other confounder (Web Appendix M).
Alternative definition of time-of-day variables: To ensure that our results are robust across a different definition of time-of-day variables, we change the operationalization of these variables (Kanuri, Chen, and Sridhar 2018). We redefine morning as 8:00
Additional Analyses
We run additional related analyses to provide supplementary information about sales momentum as observed in our context. Specifically, we (1) examine financial implications of controlling the sales momentum, (2) propose a logic for a DSS and test its performance using simulation, (3) explore possible downstream consequences of sales momentum, and (4) assess the potential long-term effect of momentum.
Financial Implications
As we find that sales momentum effects can be managed through the social effect, time of day, and day of week of making sales calls, we assess the economic value that controlling for the impact of momentum through these factors would have had to our focal firm. We use Equation 8 to estimate the financial consequences of controlling momentum.
Financial consequences of accounting for negative momentum with social effect
If the salesperson had been in the SCC, the expected decrease in odds is only 35.0%. Therefore, by weakening the effect of negative momentum, salespeople would have been able to sell 11.5% more (equivalent to 600 units) if the NSCC operated as the SCC. We evaluate the financial consequences of effectively managing negative momentum by designing a social working environment with Equation 8, and we find that the benefit of operating an SCC for our focal firm is approximately US$3,863.71 over the course of three months. In other words, if the focal firm operated the NSCC as an SCC, it would have generated US$1,287.90 more per month than what it generated in revenue. 14
Financial consequences of accounting for negative momentum with time of day and day of week
Negative momentum midweek leads to a 9.0% (13.9%) decrease in revenue compared with early (late) in the week. Consequently, compared with midweek, the financial benefit of salespeople in the NSCC making a call earlier in the week is US$3,013.54 (US$1,004.51 per month), and the financial benefit of making a call later in the week is US$4,677.86 (US$1,559.29 per month). Negative momentum in the morning leads to a 10.7% decrease in the chance of selling compared with midday. Therefore, the benefit of salespeople in the NSCC making calls midday as opposed to morning in the three months of the campaign period is approximately US$3,591.55 (US$1,197.18 per month). 15
Decision Support System (DSS) Simulation
To provide practitioners with specific directions on how to spot and manage momentum effects, we propose DSS decision-making steps (Figure 2) and simulate the extent to which salespeople’s performance would have improved if our focal firm used such a DSS to manage momentum. The layout of the DSS is as follows:
Step 1. Salesperson j on day d makes call t and enters the outcome Step 2. Determine whether salesperson j on day d is in momentum from Equations 1 and 2 using Step 2.1. If the salesperson is in negative momentum, go to Step 3. Step 2.2. If the salesperson is in positive momentum, go to Step 4. Step 2.3. If the salesperson is not in momentum, go back to Step 1 (t becomes t + 1). Step 3. The salesperson is in negative momentum. The salesperson is given a break (t becomes 1). Step 4. The salesperson is in positive momentum.
Step 4.1. If the salesperson is in positive momentum for the first time (i.e., z = 1), go back to Step 1 and build on previous outcomes to calculate the momentum calculation (i.e., t becomes t + 1). Step 4.2. If the salesperson’s clumpiness measure for z + 1th time is less than that of z (i.e., Step 4.3. If the salesperson’s clumpiness measure for z + 1th time is greater than or equal to that of t (i.e.,
DSS decision-making steps.
We test the performance of this proposed logic by comparing the actual outcome from our data with the simulated outcome using the proposed DSS logic. Doing so gives an idea of how much better (or worse) the salesperson would have performed if there was a DSS. We find that salespeople in our data would have performed 18.4% better after each momentum experience had the system been in place. Web Appendix N lays out the specific steps in simulating the outcomes.
Downstream Consequences of Momentum
Although the objective of this study is to examine the impact of sales momentum and its moderating factors, there may be unknown downstream consequences of salespeople’s momentum experience. We explore potential downstream consequences of momentum on salesperson efficiency and confidence in making sales calls. First, we assess whether salespeople are becoming more efficient in call durations by examining the correlation of positive and negative momentum from the previous call with call duration in the next call. We take the average time spent on successful calls and unsuccessful calls and assess the difference in time for each outcome type. 16 We find that stronger positive (r = −.0385) and negative (r = −.0957) momentum in the previous call leads to shorter subsequent calls. However, these correlations are weak (Anderson and Sclove 1986), and we do not have sufficient evidence to conclude that momentum leads salespeople to make calls more efficiently. Second, we assess the downstream consequences of positive (negative) momentum leading to overconfidence (underconfidence) by examining the correlation of positive (negative) momentum in the previous call and confidence in the next call. We find that confidence has a weak correlation with positive (r = .007) and negative (r = −.043) momentum. Positive (negative) momentum may lead to increased (decreased) confidence; however, the correlation is weak and we do not have sufficient evidence to conclude that positive (negative) momentum may lead to overconfidence (underconfidence).
Weak correlations in downstream consequences analysis may occur because salesperson consciousness of momentum is mostly ex post. We expect that salespeople typically realize that they were in momentum in hindsight. Given these weak correlations, we conclude that salespeople in momentum are unaware of their state of momentum and do not behave accordingly.
Long-Term Effect of Sales Momentum
Although sales momentum examined in the current context is, by nature, a short-lived phenomenon, we check the potential long-term effect of momentum. The momentum effect should be treated as having a long-term impact if a short-term impact is carried forward and sets a new trend in performance (Dekimpe and Hanssens 1995). For example, we examine whether momentum affects performance in calls made two or three calls later. To do so, we create n lags in momentum variables and determine whether they significantly influence performance made n + 1 calls later. We create lags up to n = 3. Similar to our original model, we run a logit model with lagged momentum variables and a set of control variables, as follows:
As the results in Web Appendix O show, virtually none of the lagged momentum variables used to capture the long-term impact are significant. By contrast, the short-term effect variables,
Discussion
This study provides evidence that momentum is relevant to and can be directly observed (vs. inferred) in individual sales performance. In doing so, it is, to the best of our knowledge, the first study to show direct (vs. perceived) effects of both positive and negative momentum using sales call outcomes. Specifically, we (1) detect the existence of momentum in individual disaggregated sales performance outcomes, (2) demonstrate the role of momentum in driving improvements and decrements in future sales performance, (3) show that the social environment in the workplace is an important factor that impacts the effect of momentum on performance, and (4) show that timing of the call (i.e., time of day and day of week) also influences the relationship between momentum and performance.
Theoretical Contributions
First, we contribute to the sales effectiveness literature by providing evidence of temporal spillovers of sales call outcomes. Most studies examining drivers of salesperson performance use either cross-sectional or aggregate data, thus ignoring momentum in actual outcomes across consecutive (individual) sales calls. The limited research on the influence of historical call performance might explain why existing models of individual sales performance rarely exhibit high explanatory power (for meta-analyses, see Albers, Mantrala, and Sridhar [2010] and Franke and Park [2006]). Also, aggregate data cannot capture salespeople's moment-to-moment momentum experience, which can be an important factor that explains sales call outcomes in a context where salespeople make back-to-back sales calls throughout the day. To examine momentum at the individual call level, we treat performance as a series of related sales attempts and offer a systematic way to determine the temporal spillover for each call.
Second, we contribute to sales literature by demonstrating the importance of positive and negative momentum as drivers of sales performance. A few studies explore issues similar to the momentum we study but do not explicitly distinguish the effect of positive and negative momentum. Discerning the positive and negative effects enables us to examine potential asymmetric effect sizes and moderating effects on the two types of momentum. By leveraging our rich set of disaggregate call-level data, we show how clusters of outcomes from preceding calls (i.e., momentum) influence the outcomes of subsequent calls and categorize the clusters of outcomes as positive and negative momentum. This categorization enables us to explore the distinct influence of both salesperson failure and success. We indeed find a varying effect of momentum, such that positive momentum increases while negative momentum decreases the likelihood of sales in the next call. The timing of the calls has an asymmetric impact on the effect of positive and negative momentum, such that the timing significantly moderates the effect of negative momentum but does not impact the effect of positive momentum. This asymmetric impact of moderators shows that distinguishing the bidirectional characteristics of momentum gives a holistic view of factors influencing salespeople’s performance.
Third, we contribute to the momentum literature by exploring additional factors that influence the effect of momentum. Patil and Syam (2018) explore the moderating role of salespeople’s performance, but most momentum research examines (1) interruption (Adler 1981; Markman and Guenther 2007), (2) gender (Cohen-Zada, Krumer, and Shtudiner 2017), (3) source of payment (Dhar, Huber, and Khan 2007), and (4) comparison of own performance and competitor’s performance (Lehman and Hahn 2013). Although the latter two are not applicable in a sales context, we include gender and interruption as part of our social effect in our model. Adding to this relatively limited list, we find that midday, early and late-week activities mitigate the harmful effect of negative momentum. These findings augment the nascent body of research exploring factors that moderate the effect of momentum. In addition, we overturn a moderator from prior studies (i.e., gender), which we find is not significant in the sales context.
Finally, we extend workspace management literature related to salespeople by identifying the social environment as an important factor influencing the relationship between momentum and performance. We add to previous sales research that has identified the possible influence of social effects (Chan, Li, and Pierce 2014) by showing that social effects help support positive momentum and hinder negative momentum. We also add to general workspace design and management work by showing that the interruptions inherent in social workplaces have similar substantive effects to those of the broader social environment.
Managerial Implications
For many years, scientists were confident that belief in momentum was a pervasive human bias. Nevertheless, this belief remained quite popular in many contexts. Indeed, our salesperson survey shows that most organizations try to identify salespeople who are in momentum (Web Appendix A); however, managers do not have a systematic method to detect momentum, and they do not know how to effectively manage momentum. Our study not only shows that momentum exists and has a substantive effect on performance but is the first to provide a usable method by which sales organizations can objectively detect their salespeople’s momentum state in real time and react to it. Our results, therefore, have substantial managerial implications and are likely to be generalizable to sales settings in which salespeople make multiple sales calls within a short period.
We find that the social environment has a moderating effect on both positive and negative momentum, enhancing the odds of making a successful sale. To harness these effects, managers might consider designing workspaces to increase social interaction. In addition to avoiding the physical separation of salespeople when designing new workspaces, managers could implement simple changes to their existing workplace practices. Interaction could be encouraged through gatherings such as team lunches or by creating a game room to encourage salespeople to interact during breaks. Such initiatives could help increase interaction among coworkers, enhance the effect of positive momentum, and decrease the effect of negative momentum.
Alternatively, firms could replicate the interruption mechanism of the SCC by building a predefined logic into the system that distinguishes when salespeople are in momentum and when to assign a break. This can be done by incorporating a computerized DSS, such as the one we design in this study, that can instantly control sales momentum with the latest data on outcomes, eliminating lags in the decision process. As the system randomly selects customers’ phone numbers, it could simultaneously calculate salespeople’s clumpiness on the basis of their outcomes in previous calls. If the system finds that salespeople are experiencing negative momentum, it could interrupt this momentum inconspicuously by not assigning the next phone number but perhaps temporarily assigning another task. Then, the system could start calculating the momentum measure from a new call sequence. Conversely, if the system finds that salespeople are experiencing positive momentum, it could compare the change in momentum across calls. When the strength of salespeople’s positive momentum is weakening, the system could interrupt, which our results show helps sustain positive momentum. In this case, the system calculates momentum for the next call and so on in continuation of the previous momentum calculation. We simulate this logic for the DSS and find that salespeople in our data would have performed 18.4% better than how they actually performed, if the DSS had been implemented. In practice, managers can use real-time performance data to assess salespeople’s momentum experience before each sales call.
Finally, we show that although momentum management can enhance salespeople’s short-term performance, it does not affect their long-term performance or significantly influence their selling behavior. That is, momentum does not have any significant downstream consequences for salespeople's behavior, as salespeople’s consciousness of momentum is mostly ex post, rather than during the experience. In addition, momentum is short-lived and does not have long-lasting effects. Taken together, our results show that managers need not worry about any possible change in salespeople’s selling behavior after the momentum experience, but instead managers should focus on timely detection of momentum, most likely through a DSS of the type we design here. Such a system will help enhance the likelihood of selling in the short term, therefore augmenting overall performance.
Limitations and Further Research
Our results should be considered with several caveats and limitations in mind. First, although we establish the existence of momentum, our model is not intended to address how and why salespeople enter momentum, which is limited by our ability to observe this process directly (Miller and Sanjurjo 2018). Future studies should investigate the factors that cause momentum because we find that momentum exerts an influence on performance. For example, some environmental conditions may make salespeople prone to entering momentum. Such conditions could be individual-based or situation-specific, within or outside the workplace.
Second, with the observational data collected through a natural experiment, we infer the interruption and interaction aspects of the social effect from different break-taking behavior and by showing that intercall interruptions yield similar results to the general social effect. Future studies could test the causal mechanisms behind the social effect and its influence on the relationship between momentum and performance. To directly capture interruption and interruption aspects of the social working environment, studies could conduct a field experiment to collect information on which salespeople are interacting when and preferably with whom (e.g., location maps), in addition to information on salespeople’s performance by time. We strongly recommend that future studies run experiments to examine the role of emotional or other state spillovers in social working environments to further advance our line of study. In addition, studies could exploit an exogeneous event (e.g., fire alarm drills) to determine whether the unexpected event can disrupt a salesperson’s momentum similar to social settings.
Third, throughout the analysis, we used a constant rolling window to capture momentum. Specifically, in assessing the impact of momentum on sales, we used a rolling window of nine calls and captured individual differences using salesperson fixed effects. This enabled us to compare momentum in a panel setting consistent with the scope of our research. Future studies could explore other effective ways to capture momentum using individual-specific rolling windows to account for any personal differences in momentum experience.
Fourth, we assume that salespeople have short-term memory of experiences during the nine call window. This assumption is based on prior literature that found short-term memory of experiences in a rolling window of four coin tosses (Hahn and Warren 2009; Warren et al. 2018). Future studies could examine the maximum number of call experiences that is salient, across multiple contexts.
Last, we use simulation to evaluate the performance of the proposed DSS. Specifically, we identify the first momentum experience by each salesperson each day and compare the expected outcome with the observed to find that salespeople in our data would have performed 18.4% better after experiencing momentum with DSS in place. However, it may be suboptimal in practice to assign a break after every momentum experience, especially if a salesperson experiences momentum frequently. Future studies could run a field experiment to find the ideal number of assigned breaks that maximizes a salesperson’s performance while managing momentum effects.
Supplemental Material
sj-pdf-1-mrj-10.1177_00222437221095097 - Supplemental material for Managing Positive and Negative Trends in Sales Call Outcomes: The Role of Momentum
Supplemental material, sj-pdf-1-mrj-10.1177_00222437221095097 for Managing Positive and Negative Trends in Sales Call Outcomes: The Role of Momentum by Irene Y. Nahm, Michael J. Ahearne, Nick Lee and Seshadri Tirunillai in Journal of Marketing Research
Footnotes
Acknowledgments
The authors thank marketing seminar participants at Case Western University, George Mason University, Texas Christian University, University of Missouri, University of Tennessee, University of Minnesota, Michigan State University, Tilburg University, University of Alabama at Birmingham, Korea University, 2019 Marketing Ph.D. Doctoral Symposium at University of Houston, 2018 Marketing Science Conference at Temple University, 2018 Enhancing Salesforce Productivity Conference at University of Missouri, 2017 Sales Thought Leadership Conference at HEC, 2017 AMA Winter Marketing Educator’s Conference, and 2016 Theory + Practice in Marketing Conference at Texas A&M University. The authors are grateful to the JMR review team for their valuable feedback on this article.
Associate Editor
Shrihari Sridhar
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
