The Cost of High-Powered Incentives: Employee Gaming in Enterprise Software Sales *

Size: px
Start display at page:

Download "The Cost of High-Powered Incentives: Employee Gaming in Enterprise Software Sales *"


1 The Cost of High-Powered Incentives: Employee Gaming in Enterprise Software Sales * Ian Larkin Harvard Business School May 17, 2008 DRAFT * I am grateful to George Baker, Max Bazerman, Constanca Esteves, Bob Gibbons, Brian Hall, Bronwyn Hall, David Mowery, David Reiley, Al Roth, Rob Seamans, Jason Snyder, Steve Tadelis, Catherine Wolfram, Oliver Williamson and seminar participants at UC Berkeley, UCLA, USC, Stanford, Arizona, Indiana, Kellogg, Harvard, MIT, Georgetown, Maryland, University of North Carolina, London Business School and INSEAD. All mistakes are my own. I thank the Ford Foundation and Harry S. Truman Foundation for graciously providing research funding for this project, and the enterprise software vendor for providing the dataset and for many useful discussions. Address for correspondence: Baker Library 463, Soldiers Field Road, Boston MA

2 The Cost of High-Powered Incentives: Employee Gaming in Enterprise Software Sales It is well known that employees game incentive systems designed to motivate effort and retain top performers, sometimes to the detriment of their own employer. However, there are very few detailed empirical studies which document the extent of this gaming, or estimate the cost of this gaming to firms. In this paper, I use a proprietary database of deals sold by a leading enterprise software vendor, together with the incentive system used to compensate salespeople, to demonstrate the scope of incentive system gaming by employees, and the revenue the vendor forgoes due to gaming. Accelerating commissions and quarterly deadlines give salespeople the incentive to concentrate deals into a single quarter, and to avoid making any sales in other quarters. Since the sales cycle is months long, salespeople have strong control over the quarter in which sales close. Salespeople also partially control discounts, allowing them to offer lower prices to customers who go along with their preferred deal timing. After documenting some clear patterns of gaming, I use instrumental variable and matching techniques to compare deals that were gamed to very similar deals that appear not to have been, and estimate that gaming leads to excess discounts that cost the vendor 6-8% of revenue. This figure is remarkably similar to the 7-8% of revenue the vendor spends directly on salesperson commissions, suggesting the vendor may spend twice what it thinks it does on attracting, motivating and retaining top salespeople. 2

3 1 Introduction Attracting and retaining highly skilled employees is a key strategic task facing firms. To appeal to and motivate top employees, firms have increasingly turned to incentive systems that pay based on performance. The use of stock options to compensate senior executives is the most visible example of a performance-based pay system, but in many industries, such as high technology and professional services, performance determine a significant portion of compensation for nearly every employee on the payroll (Culpepper, 2006). It is well known that performance-based incentives can have perverse consequences, as employees take unforeseen actions to increase their compensation. Such behavior, commonly referred to as gaming, can sometimes harm the employer. Some gaming behavior, such as falsifying invoices or backdating stock option grants, is clearly illegal or unethical. However, employees often manipulate incentive systems in straightforward, ethical ways, especially when it is impossible to write a contract specifying all relevant actions or outcomes. One prominent example is the gaming of nonlinear, period-based compensation systems, which give employees the incentive to alter the timing of tasks to take advantage of the non-linearity in the compensation schedule. This timing gaming (Oyer, 1998) can affect business outcomes by changing revenue flow, pricing, or other key variables in ways beneficial to the employee but detrimental to the firm. While the prevalence of timing gaming in the face of non-linear, period-based incentives is well-known, little is known about the scope or cost of this gaming to employers. Existing empirical studies on timing gaming focus on macro-level company data (Oyer, 1998), illegal employee actions (Healy, 1985), or non-business settings (Asch, 1990). In this paper, I use a proprietary database of deals for a leading enterprise software vendor to investigate the extent of timing gaming by the vendor s salespeople, and to measure the cost the vendor incurs due to this gaming in terms of foregone revenue. Like many business-to-business (B2B) technology companies, the vendor compensates salespeople largely on an accelerating commission scale, meaning a salesperson s commission for the same deal can vary dramatically depending on the quarter in which the deal closes. The sales cycle in enterprise software is usually well over a year, giving 3

4 salespeople considerable control over the exact quarter in which a deal closes. Since salespeople partially control the level of discount granted to a customer, they have the ability to give the customer a strong incentive a lower price to purchase in the quarter preferred by the salesperson. The paper s empirical results suggest that gaming is widespread and costly to the vendor. In the paper, I first demonstrate that salesperson compensation concerns appear to influence the timing of the large majority of deals in the database. I next use several statistical techniques, including an instrumental variables approach to control for omitted variable bias and matching techniques that are based on gaming propensity, to demonstrate that this timing gaming results in excess customer discounts representing 6-8% of vendor revenue, all else held equal. The results are stable across estimation techniques and therefore appear to be statistically robust. Salespeople receive on average 8% of booked revenue in the form of sales commissions, so this result suggests that the incentive system costs the vendor approximately twice what it may think it is spending on salesperson compensation. That the firm bears an additional indirect cost to motivate salespeople, above and beyond the commissions it pays them, does not prove that the incentive system is suboptimal. However, the magnitude of this indirect (and hitherto hidden) cost begs the question of why most business-to-business high technology firms use highly non-linear incentives with relatively short deadlines. At a minimum, the level of indirect cost suggested in this paper demonstrates the premium technology firms place on attracting and motivating top employees. More fundamentally, the results may suggest that it is more costly than firms realize to rely on incentive systems that have always been used or are common throughout the industry, a rationale commonly given by executives in industry interviews 1. This paper makes several contributions. It is, to my knowledge, the first paper to use detailed internal sales, employee and incentive system data to estimate the extent and cost of gaming in terms of foregone revenue due to mis-pricing. In this way, it goes beyond the question of do incentives matter? to show the extent to which they matter, in 1 Indeed, some enterprise software firms have recently begun to move away from using quarterly deadlines in their salesperson incentives, citing the very gaming costs noted in this paper. These experiences are detailed in the last section of this paper. 4

5 a business setting. The estimate of the effect on pricing is also novel, in that many theoretical and empirical studies assume that higher employee wages, not excessive discounts to customers, are the primary cost of high-powered incentive systems. Many empirical studies on the perverse impact of incentives focus on illegal or clearly unethical behavior, but this study examines the negative business impact of employees doing exactly what the incentive system asks them to do: make sales. Finally, the results contain some interesting directions for further research, including the interaction between employee and executive incentive systems, the effect of employee tenure on the likelihood that salespeople manipulate the compensation system, and differences in customer participation in the gaming process. The paper is laid out as follows. In the next section, I briefly introduce incentive and sales dynamics in enterprise software and other B2B technology industries. Section three reviews the relevant theoretical and empirical literatures on the use (and misuse) of incentives in organizations. In section four, I build hypotheses on the effect of non-linear, period-based incentive systems on deal timing and pricing. In section five, I review the data, estimation strategy and empirical results. The paper concludes with a discussion of the results in light of the strategic rationale for the incentive system, and briefly examines alternative incentive systems. In the final section I also review the limitations of this study and discuss potential avenues for further research. 2 Institutional background 2.1 Salesperson incentives in enterprise software The enterprise software industry, which produces the server- and mainframe-based applications and software infrastructure that manage and report the vast information flows corporations need for strategic decision making, is an ideal setting in which to study the impact of non-linear, period-based incentive systems. Nearly every enterprise software vendor uses non-linear commissions, bonuses and other high-powered incentive schemes to compensate salespeople, product development teams, service personnel, and other employees (Gartner, 2004). Executives and industry analysts claim that the use of high-powered incentives is 5

6 critical in enterprise software due to two facets of industry dynamics: extremely fast innovation cycles, and cost structures that are typified by very large product development costs and very low production costs. Major product upgrades, which can cost up to $3 billion to develop, occur every four to seven years. Marginal production costs are close to zero, and vendors feel that building significant market share is critical to the sustained customer lock-in that is necessary for long-term profitability 2. These dynamics lead to an intense battle for market share, largely focused on the sales function. Furthermore, most software vendors know little about the employee attributes that are correlated with performance, making it difficult for them to build hiring practices which attract top talent (Cusumano and Selby, 1998). Firms respond to these industry forces by offering performance-based compensation with large differences in compensation between employees who perform strongly and those who do not (Gartner, 2004). Salesperson compensation in enterprise software is based on an aggressive compensation schedule with rapidly accelerating commissions over the course of the financial quarter depending on total revenue already booked. Unlike incentives for most job functions, the compensation system is completely explicit 3. Table 1 lays out a typical compensation scheme in enterprise software, loosely based on the compensation scheme used at the vendor which provided data for this research, and the author s experience working in the industry 4. The same data are presented graphically in Figure 1. As noted, a salesperson will only make a quarterly base salary of $12,000 if she makes no sales, which is considered a starvation wage for the industry. More importantly, the commission she receives on a sale rapidly accelerates as her total sales in a quarter rise. For example, on her first $250,000 deal in a quarter, she will make a commission of 2%, or $5,000. However, if she has already closed deals totaling $6 million in a quarter, the same $250,000 deal will result in a commission of 25%, or $62,500. Depending on how much other revenue she has booked in a quarter, her sales commission 2 The industry is therefore a classic example of an information good industry, the competitive dynamics of which are discussed at length in Shapiro and Varian (1998). 3 Employees in other functions, such as product development and professional services, tend to be partially compensated by explicit performance measures, but usually are also partly compensated based on subjective measures such as managers view of performance. Reasons for this difference are discussed in the final section of this paper. 4 It should be noted that all compensation calculations used in the proceeding econometrics use the vendor s actual compensation schedule; the vendor requested that its actual compensation schedule be disguised. 6

7 on an identical deal potentially increases by an order of magnitude. As made clear by the compensation schedule, the accelerating commissions are reset on a quarterly basis, meaning that all salespeople restart at the lowest commission rate for new deals at the beginning of every financial quarter 5. Industry executives point to the strategic importance of the most highly skilled salespeople as the primary rationale for this incentive system. Software vendors feel that there are very few good salespeople, many average ones, and even more poor ones, and that market wages for top salespeople are extremely high 6. This leads to an interesting phenomenon in enterprise software and other high technology industries, where the most highly compensated employee is often not the CEO, but the top salesperson (Gartner, 2004). This rationale confirms a basic holding of the agency literature: high-powered incentives can be used to sort among employee type (Lazear, 1986; Lazear 2000a). However, according to industry executives, the importance of top salespeople goes beyond the amount of revenue they directly book. Top salespeople are held to be better at selling new products and major upgrades to existing products. Successful market penetration by these products provides the basis for long-term profitability, since minor upgrade purchases, which are much more easily sold by average salespeople, are a large source of revenue. Industry executives therefore believe that there is a spillover effect from sales of critical products to sales of minor upgrades and other non-critical products 7. A basic breakdown of sales by product type for the vendor that provided data for this research, shown in Table 2, supports the notion that top salespeople are more productive in selling new products and critical upgrades. As shown in the table, the top 10% of salespeople (measured by total revenue booked) are responsible for 32% of revenue for the company. However, they are responsible for 46% of revenue for new 5 Most vendors use a combination of dedicated quotas and accelerating commissions; that is, sales commissions are equal to zero until a salesperson surpasses their quota, after which commissions accelerate. The vendor providing data for this research did not use a quota-based system, but its accelerating commission schedule was quite typical for the industry. 6 This phenomenon is not unique to software or even to high tech, although the quick innovation cycles and experience good nature of the product exacerbate the differences between good and bad salespeople in software and tech environments. 7 Industry executives also feel that employing top salespeople helps with recruitment, since their identities are widely known in the industry, and young or inexperienced salespeople want to learn how to sell from the best. Therefore, there may be a recruiting spillover in the hiring of top salespeople as well. 7

8 products, and 42% of revenue for upgrades defined as mission critical by the vendor; furthermore, they are responsible for only 24% of sales for non-critical, existing products 8. If revenues from non-critical upgrades of existing products, which account for over 60% of the vendor s total revenue, are partly dependent on previous sales success for new and critical products, the advantages for corporate performance provided by top salespeople go well beyond their direct revenue contributions Other evidence supports the view that the vendor examined in this paper uses the incentive system to effectively sort among employee type. In the past decade, the firm has changed the commission percentages three times in 1997, 2003 and Table 3 shows a disguised example of how commissions changed during these periods. As seen in the table, the vendor substantially raised commissions in 1997; it reports that this was necessary because the emerging Internet bubble made it hard to recruit top salespeople. After the Internet bubble collapsed in the early 2000s, the vendor twice lowered commissions, reporting that it no longer had any difficulty recruiting or keeping salespeople. Interviews with other industry executives suggest that commissions have come down throughout the industry in the last few years due to the collapse of the Internet bubble, although the basic non-linear structure remains intact. The tying of the non-linear commission schedule to the financial quarter has several rationales. First, given the increasingly quick innovation cycles in the industry, executives feel that basing commissions on annual sales risks missing critical upgrade cycles. Oyer (1998) shows that salespeople in annual quota systems appear to shirk for a number of months at the start of the financial year, and vendors feel they cannot risk this type of behavior. Just as importantly, the senior executives at enterprise software vendors are compensated (and retained) largely on the basis of the company s stock market performance, which is closely tied to the ability of their companies to hit quarterly Wall Street financial targets. Finally, executives report that it is industry standard practice to use quarterly periods to determine commissions, and are afraid that good salespeople will leave if the company deviates from this practice. 8 These results do not stem from any differences in the assignment of sales territories or products. The vendor uses an annual process of assigning sales territories whereby every salesperson has an approximately equal chance to reach a predefined quota representing the vendor s average expected sales per salesperson. Furthermore, salespeople are specific to territories, and do not have any product specialization. 8

9 There is one other relevant aspect of the incentive system: the degree of salesperson control over price. As is common in large B2B procurement environments with intense price competition (Bhardwaj, 2001), enterprise software vendors give salespeople a great deal of flexibility to control discounts. However, as the discount negotiated with the end customer increases, the level of authority needed to authorize the discount increases as well. Table 4 shows a disguised example of the deal approval process for the vendor which provided data for this research. The key idea is that salespeople do have a great degree of control over pricing, but the likelihood of getting a high discount approved goes down as the level of discount, and therefore the level of needed approval, goes up. 2.2 Prevalence of high-powered, period-based incentives Although enterprise software may be one of the most extreme cases, the use of accelerating commissions, non-linear bonuses and other output-based pay for non-executives is the rule, not the exception, in many industry environments. Compensation specialists note that highpowered incentives are particularly prevalent in high technology and professional service industries, since effort is so hard to monitor, and outcomes so uncertain (Culpepper, 2006). In industries marked by high rates of innovation or technical change, high-powered incentives are held to be critically important to attract and retain the best employees, among the most critical components of successful innovation (Mansfield et al, 1971; Teece, 1986; Zenger, 1994; Brynjolfsson et al, 1993). In business functions as diverse as product development, research and analysis and even administrative support, performancebased pay often accounts for the majority of total compensation (Zenger, 1994; Culpepper, 2006). Firms typically spend about 7% of revenue on the sales function (Godes, 2003), demonstrating its importance to most companies. Non-linear incentives are also prevalent in salesperson compensation contracts. In a recent survey, 95% of salespeople reported their salaries were partly based on commissions and bonuses (Joseph and Kalwani, 1998). Oyer (2000) notes that well over three-quarters of salespeople were compensated by a non- 9

10 linear scheme, giving them incentives to time sales to maximize their own compensation 9. Industry executives and analysts in many technology and other industries, including supercomputing, pharmaceuticals, defense equipment, telecom equipment, semiconductors, and large real estate sales, have reported to the author that salespeople in their industry are compensated using some form of accelerating commission, although none reported that the rate of acceleration was as high as that of enterprise software. 3 Existing literature on the use (and misuse) of highpowered incentives This section discusses the existing theory and empirical evidence on high-powered incentives in firms. The use of high-powered incentives and the resulting impact on firm performance are key strategic issues facing firms, and many authors have posited that firms using more incentive-heavy compensation structures for everyday employees perform better, and have found some support for this theory in certain empirical settings (Teece, 1986; Zenger, 1994; Zenger and Lazzarrini, 2004). 3.1 The rationale for high-powered incentives: agency and other theories Although output-based incentives 10 such as sales commissions have been prevalent for certain job functions for well over a century, they have attracted the attention of scholars only recently. Williamson (1975, 1985) was among the first to discuss the positive motivational effects of linking compensation to outcomes, arguing that they induced effort. The groundbreaking paper by Holmstrom (1979) on tying incentives to measurable outputs established criteria for improving performance-based contracts when first-best contracts are not available, and became the cornerstone for the subsequent agency literature. Lazear (1986) added an important element to the rationale for output-based incentives: they could 9 It is also common in sales environments to use decelerating commissions, such as a set quota, rather than accelerating commissions, such as those used by the vendor in this study. Both of these non-linear commission schedules will affect incentives to time the closing of deals. 10 This paper uses the terms high-powered incentives and output-based incentives interchangeably. It should be noted, however, that not all output-based incentives are directly tied to measurable output. Bonuses, for example, are often based on subjective performance measures. 10

11 effectively sort among worker types, inducing workers of sufficient skill to choose to stay at the company, and those with insufficient skills to leave. These two agency theory-based rationales inducing effort and sorting have become the leading explanations in the economics literature for the increasing use of output-based incentives (Lazear, 2000a) 11. There is a vast empirical literature supporting the notion that the use of outputbased incentives for employees can have a positive effect on performance. A number of studies examine the change in a performance measure after a switch to high-powered incentives, and nearly every study finds a significant, positive effect. Examples include productivity in installing windshields (Lazear, 2000b), sales productivity for retail stores (Banker, Lee and Potter, 1996), productivity in collective agriculture concerns in China (McMillian, Whalley and Zhu, 1989), and productivity in tree planting (Paarsch and Shearer, 1999). One downside to these and similar studies is that they rarely distinguish between the effort and sorting effects of output-based incentives (Prendergast, 1999). In addition, a number of studies in the marketing literature examine the use of commissions and other output-based incentives for salesforce motivation, finding a strong agency-based rationale for these practices (e.g. Basu et al, 1985; Lal and Srinivasan, 1993; Shaw et. al, 2000). Salespeople tend to be less risk averse than the average employee (Coughlan and Narasimhan, 1992), and the high prevalence of commission-based compensation for salespeople may induce self selection of risk loving types to the sales function The unintended consequence of incentives The use of high-powered incentives has become so prevalent exactly because they so strongly influence actions, and therefore outcomes. However, it is well-known that not all actions or outcomes induced by high-powered incentives are intended or beneficial. One clear problem for the use of high-powered incentives arises when no measurable output neatly corresponds to the principal s goals, leading to a potential mismatch between 11 Institutional theory, which explores the role that industry norms, management fads and other behavioralbased factors play in corporate control mechanisms, is also held to explain the rise of high-powered incentives (Zucker 1987). Empirical work comparing agency and institutional theory find the two are largely complementary (Eisenhardt, 1988). 12 Camerer et al (1997) make a similar argument around employee self-selection based on compensation plans in the market for taxi drivers in New York City. 11

12 desired outcome and the behavior motivated by the compensation system. The basic logic and several examples of this phenomenon were given in the classic piece by Steven Kerr titled On the Folly of Rewarding A While Hoping for B (1975). The most commonly cited business example of detrimental gaming is Sears experience with its automotive mechanic arm. Sears moved from hourly compensation for mechanics to a revenuesharing arrangement, and mechanics predictably responded to this incentive scheme by ordering unneeded repairs. The negative publicity stemming from employee gaming is held to have caused long-term underperformance for this unit of Sears, due to a widespread belief among consumers that Sears mechanics were not trustworthy (Patterson, 1992; quoted in Baker, 2000). The literature on multitasking, inspired by Holmstrom and Milgrom (1991) and Baker (1992), recognizes that job functions are complex, and employees will opportunistically shift their effort towards those tasks that make up parts of their compensation scheme. Since these tend to be tasks with measurable outputs, too little effort may be put into tacit, but important tasks. One interesting empirical application of the multi-tasking model is Johnson, Reiley and Muñoz (2006), which shows that private bus operators in Chile, rewarded solely on the basis of their total number of passengers carried, drive at excessive speeds in the war for fare and cause a disproportionate number of accidents, as compared with drivers of state-run busses, who are paid an hourly wage. A related stream of research examines non-linear compensation structures, particularly related to deadlines 13. Deadlines are usually not correlated with underlying demand for a product or service, yet employees can game the timing of a task so they maximize their compensation. In one of the earliest empirical studies on the topic of deadlines, Asch (1990) demonstrated that Navy recruiters were very susceptible to a period-based award system the Navy used to recognize and compensate outstanding recruiters. They would strategically stockpile potential recruits until eligible for and likely to achieve an award, resulting in an unsmooth recruitment rate which was not explainable by the underlying demand to enlist. 13 Several studies look at the perverse impact of non-linear incentives without examining deadlines; these include Chevalier and Ellison (1997) on the mutual fund industry and Leventis (1997) on the market for cardiac surgery in New York. 12

13 Oyer (1998) extended this logic to the business setting, looking at revenue streams and resulting margins for companies with similar products and customers but different financial periods. In Oyer s example industries, salespeople are compensated on a nonlinear incentive scheme based on an annual sales quota. He found that the revenue flows increased and margins decreased for such companies as the end of their financial reporting year approached. This empirical regularity, Oyer demonstrated, was consistent with timing gaming by salespeople, who had substantial (but incomplete) control over both deal timing and pricing. He took advantage of exogenous changes in the financial reporting period of some companies to show that demand characteristics did not explain these differences in revenue flow and pricing. Healy (1985) demonstrated that senior executives made similar decisions on revenue recognition when their compensation was non-linear and based on fiscal year firm performance. In sum, there is considerable theoretical and empirical evidence that incentives matter tremendously in organizations and may have unanticipated consequences. Researchers have made considerable progress in the development of methods to overcome incentive problems when outputs cannot be easily measured 14. However, economists and business strategists have made remarkably little progress in understanding non-linear, period-based incentive schemes (Prendergast, 1999). 4 Theory development and hypotheses In building hypotheses around timing gaming and the effect of this gaming on business outcomes, it is most useful to consider the factors influencing the actions of two actors: salespeople and customers. While other participants are relevant, including executives, salesforce management personnel and shareholders, I will treat the actions of these actors as exogenous when building hypotheses, and empirically control for the effects of their actions when possible. 14 For example, there is a significant literature on the use of subjective performance measures, which partly looks at overcoming the problems of measurability and the link between desired outcome and induced employee action. See, for example, Baker, Gibbons and Murphy (1994). 13

14 Agency theory predicts that salespeople will take advantage of any aspect of their incentive system that increases their compensation. If the salesperson faces a non-linear, period-based commission system, she has two mechanisms by which to increase her compensation: influencing the timing of a deal, and influencing the price paid. Influencing timing can increase compensation, because of the non-linearity in compensation: making more deals happen in a single period will leave her better off than having a smooth flow of the same deals across periods. Salespeople can influence timing covertly, by deliberately slowing negotiations in a bad quarter, or openly, by promising better deals to customers if deals are closed on the salesperson s preferred timeline. In terms of pricing, a salesperson would prefer to sell at a higher price (all else equal) because it results in a higher commission. However, she would be willing to sell at a lower price if doing so would result in higher overall compensation, due to the non-linearity in compensation. On the customer side, I assume that utility is based on two factors: price paid, and timing of purchase. I assume that there are two customer types: one cares greatly about timing, and one has weak timing preferences. Both customer types care greatly about price. Even a customer with weak timing preferences faces disutility as a deal s timing moves away from its preferred timing 15. The assumptions about the salesperson s mechanisms to maximize compensation, and the elements of customer utility, suggest strongly that salespeople will strategically time the closing of deals where they are able to do so, in order to maximize their compensation. This is essentially because salespeople have a much stronger motivation to affect timing than customers. A salesperson who has already generated a large amount of revenue in a quarter will try to pull forward some deals which would otherwise close later. Conversely, a salesperson who has not or expects not to close many deals in a quarter will try to push out deals which would otherwise close in that quarter. Formally, I hypothesize: 15 I examine the assumption that some enterprise software customers have very strict timing preferences, while some do not, in the following section. 14

15 H1: Deals for which customers do not have strong timing preferences will close in the quarter which maximizes salesperson compensation, compared to quarters immediately around the salesperson s preferred quarter. Again, this hypothesis recognizes that customers have some underlying time preference, and the salesperson s ability to game therefore cannot stretch infinitely across quarters. Hypothesis 1 focuses on the timing of deals across quarters, but a natural extension posits that the incentive system also affects the timing of deals within quarters. Deals that are pulled from later quarters will naturally close late in a quarter, as the salesperson attempts to convince the customer to purchase earlier than the customer prefers but tries to avoid inflicting significant disutility on these customers. The converse statement also would appear to have merit: deals pushed from earlier quarters will naturally close early in a quarter. While it may be natural to think that a salesperson would prefer never to close a deal early to have an option around pushing the deal into even later quarters, it is important to remember that customers face increasing disutility by moving their purchase date away from their preferred timing. Salespeople who expect to have a big quarter will, therefore, will push a deal so that it enters their planned big quarter, but not so far as to risk losing the deal altogether. This effect will lead to a prevalence of early deals. Formally, I hypothesize: H1a: The timing gaming of deals will lead to a natural bunching of deals at the beginning and the end of the financial period that determines the salesperson s commission. Again, some deals will occur away from the beginning and end of the period, for customers with strong timing preferences. Customers with weak timing preferences are only willing to change their preferred purchase date if it positively affects their utility via a lower price. Of course, all else held equal, salespeople would prefer not to lower price, since it lowers their commission. 15

16 However, if giving a customer a few more points off of list price helps convince the customer to purchase in a period where the salesperson has already closed many other deals, the discount effect on compensation is easily swamped by the commission effect inherent in the non-linear compensation schedule. To the extent that customers derive more utility from paying less than from having exact control over deal timing, they will hold out for a good deal in a period where the salesperson has strong incentives to close a deal before the quarter ends, believing they can use this fact to drive a bigger discount. Of course, as noted, this outome requires that the customer does not have strong preferences around exact deal timing, which is not true in all instances. From this discussion comes the paper s second hypothesis: H2: Deals whose timing was strategically manipulated result in significantly higher discounts for customers than deals whose timing was not. The identification of a set of deals whose timing appears not to have been manipulated is critical to the empirical strategy for identifying this effect. 5 Data, estimation strategy and results 5.1 Data The data for this study were provided by a leading enterprise software vendor, representing all deals closed by 225 salespeople based in North America, selected randomly from all salespeople employed by the company for at least two quarters between 1997 and In total the dataset contains 4,020 deals closed over the course of 22 financial quarters. The database excludes two types of deals booked by salespeople: deals under $50,000, which are usually small add-on purchases sold by a telephone representative and are not the result of negotiations; and site license deals, which give the customer the right to use as many licenses as it wishes for a particular product. Site license contracts were not available, and 16

17 much of the data used later in identification is not relevant to them, since, for example, there is no notion of the level of discount granted 16. Still, deals in the database account for nearly 90% of total direct sales revenue for the salespeople in question. The dataset was also augmented with publicly available information on customers. The final dataset contains five classes of information: 1. Deal outcomes, which includes products bought (licenses, maintenance and services), list price, and price paid. 2. Deal timing, which is the date of record for the sale (for both compensation and revenue recognition purposes). 3. Salesperson information, which includes a unique salesperson identifier, tenure, age, gender, full sales and compensation history, territory history, and mobility history across sales districts. 4. Customer information, including name, number of employees, revenues, market capitalization, some information on IT use, and previous customer purchases of the vendor s products Deal s contribution to total quarterly compensation for the salesperson, which is the marginal commission the salesperson earned on the sale in the quarter in question 18. I also calculate what the marginal commission on each sale would have been had it closed one quarter earlier and one quarter later. One unfortunate aspect of these data is that the commission schedule stayed constant, meaning there is not a straightforward experiment utilizing changes in incentives. List prices, however, were largely increasing during the period in question, and average discounts stayed about the same, meaning that overall salaries rose. Also, list prices on some products changed dramatically, producing an exogenous change in incentive intensity that I exploit in the econometric analysis. 16 This study therefore does not investigate the timing of site license deals. However, I do use their contribution to commission in the quarterly measure of salesperson compensation, so the incentive effect of site license deals on deals in the database is taken into account. 17 I only observe the products bought and total number of licenses for repeat customers in periods outside of the database. I do not observe pricing or discounts. 18 For this calculation, I used the actual compensation structure used by the vendor, similar to but not the same as the schedule given in Table 1. 17

18 I focus only on license revenue in the empirical analysis. Salespeople do get a commission on service revenue generated, but their commission is a set percentage of service revenue sold, is therefore not based on any kind of financial period, and does not count as revenue generated towards the non-linear commission schedule used for licenses. Therefore, commissions earned on service revenue are irrelevant to the incentive system under investigation here. I do control for spending by customers on service when assessing pricing on licenses, since salespeople may give better deals on licenses to customers who buy more services. Table 5 shows summary statistics for the dataset, using the deal as the unit of observation, and reveals some interesting deal characteristics. First, the deals are large, with an average size of over $950,000. Second, they are heavily discounted, with an average discount over 35%, while some discounts reached 95%. Most tellingly, nearly 75% of deals closed on the last day of the financial quarter, suggesting that the presence of the quarterly deadline in the incentive system carries a dramatic effect. The average commission of $73,000 represents a gross average commission rate of about 8% 19. The average realized commission of $73,000 is statistically significantly greater than the commission on the deal had it closed a quarter earlier ($62,900) or later ($59,800), which demonstrates that deal timing is correlated with marginal salary benefit to salespeople. Table 6 shows summary statistics on the dataset organized by salesperson-quarter, and again reveals preliminary evidence that salespeople manipulate the timing of major sales. Salespeople make no sales at all for a full third of the salesperson-quarters in the dataset which, in combination with the prevalence of deadline deals, is difficult to rationalize on the basis of demand-side characteristics alone. Over 95% of salespeople employed for at least 4 quarters in the dataset have at least one quarter in which they booked no sales. Even beyond the fact that 73% of deals close on the last day of the quarter, deal timing over the course of the quarter is not smooth. Figure 2 breaks the financial quarter into 13 weeks, and shows the total number of deals in the database which close in each 19 This figure is close to that reported by Godes (2003), who stated that sales expenses were, on average, 7% of revenues for large corporations. Of course, the vendor incurs other selling expenses beyond commissions, although commissions make up the bulk of selling cost. 18

19 week 20. Except for the beginning and end of the financial quarter, there are around 40 or 50 sales in the average week. However, there is a large spike not only on the very last day in the quarter, but also in the weeks leading up to the end of the quarter. Even more strikingly, there is a spike of deals in the beginning weeks of a quarter. The line in the figure shows the average discount for deals closing in each week. For the middle weeks of the quarter, average discounts hover around 30%; both at the start and the end of the quarter, however, discounts rise to 35-37%. These data appear very consistent with the timing gaming hypothesis. As the end of the quarter approaches, salespeople attempt to close as many deals as possible in order to take advantage of the convexity of their commission schedule. Since, as is seen in Table 6, so many salesperson-quarters have no sales at all, it is logical to think that salespeople are likely to pull forward deals that would naturally close in the preceding quarter. To do this, they are willing to give a higher discount, because the discount effect on their salary is swamped by the commission effect. It is comforting that the corresponding case at the start of the quarter is evident in the data, since the incentive story is the same. Salespeople who are in the midst of a poor quarter, or who expect many deals to close the next quarter, may push out deals that would otherwise close a quarter earlier. Again, they are able to motivate customers to go along by giving them higher discounts. The critical insight here, which is key to the identification strategy discussed in the next section, is that the expected marginal commission across quarters influences when a deal is closed within a quarter. Middle-of-quarter deals are least likely to have been influenced by the periodicity of the incentive system, and therefore are most likely to be deals around which the customer had strong timing preferences. This reasoning depends heavily on the assumption, discussed in Section 3, that some customers have weak preferences over deal timing, while others have very strong preferences. Interviews with numerous software executives and salespeople, as well as end customer interviews, suggest that for the majority of projects, customers are willing and able to be flexible on the timing of major software purchases. In general, sales cycles range from 12 to 24 months, and it can take another two to five years to implement a 20 Week 13 represents the last week in the financial quarter, except the last day. Week 14 represents the last day of the financial quarter. All the other weeks correspond to five-business day weeks. 19

20 software package once purchased. Most customers, therefore, do not have preferences over deal timing down to the exact week or month, since the ongoing project schedule is not known in detail in advance. This qualitative finding from interviews is corroborated by the data, which show that the vendor s financial deadline, which is irrelevant to most customers (except that they can use it to get a better deal), seems to drive the timing of the majority of deals 21. However, interviews suggest that for a subset of projects, customers do have strict timing requirements. For example, exogenous budgeting or administrative deadlines may require that allocated funds be spent by a certain date. Interviews suggest customers sometimes have accounting requirements which require an invoice by a certain date, or they may have internal budgeting deadlines for certain projects which mean that unspent funds will go back to the corporate pool and be lost to the business unit in question. Also, sometimes senior executives artificially give a hard deadline to the IT department to begin implementing a package, usually due to perceived competitive threats. As noted in Table 7, which breaks down deal timing by customer type, the government and education segments, which are arguably the most likely to have hard budgeting deadlines, buy 18.6% of their projects in the middle of a quarter, versus 12.2% for private companies. This statistic lends credence to the hypothesis that mid-quarter deals represent those for which customer-side budgetary deadlines were tight. The data suggest that time-sensitive projects, as measured by the total percentage of deals happening in the middle of a financial quarter, are just as likely for current as new customers. As noted in Table 7, middle-of-quarter deals make up 13.4% of new projects for current customers who previously only bought around the vendor s deadline, and 12.6% of new projects for customers that are completely new to the vendor. This is further evidence that, outside of government and education customers, hard deadlines are an artifact of specific projects, not of customers. 5.2 Estimation strategy and results 21 Customers financial reporting deadlines were introduced in several places in the empirical analysis, but never had a statistically significant effect on deal timing, discounts or other variables of interest. 20

Get the Right People:

Get the Right People: WHITEPAPER Get the Right People: 9 Critical Design Questions for Securing and Keeping the Best Hires Steven Hunt & Susan Van Klink Get the Right People: 9 Critical Design Questions for Securing and Keeping

More information

Procrastination, Present-Biased Preferences, and Financial Behaviors

Procrastination, Present-Biased Preferences, and Financial Behaviors Procrastination, Present-Biased Preferences, and Financial Behaviors Jeffrey Brown University of Illinois at Urbana-Champaign and Alessandro Previtero University of Western Ontario Prepared for the 16

More information

Module 9: IS operational and security issues

Module 9: IS operational and security issues file:///f /Courses/2010-11/CGA/MS2/06course/m09intro.htm Module 9: IS operational and security issues Overview Firms that make extensive use of information systems must contend with a number of ongoing

More information

Factors Affecting Housing Build-out Rates

Factors Affecting Housing Build-out Rates CLG Housing Markets and Planning Analysis Expert Panel Factors Affecting Housing Build-out Rates A report by Professor David Adams and Dr Chris Leishman Department of Urban Studies University of Glasgow

More information



More information



More information

Creating Value with Homogeneous Technology: OCTOBER 2007

Creating Value with Homogeneous Technology: OCTOBER 2007 Creating Value with Homogeneous Technology: Electronic Business Service Providers* Shane Greenstein** and Sarit Markovich*** OCTOBER 2007 Abstract We study the determinants of market pricing for vendors

More information

How Do Venture Capitalists Choose Investments?

How Do Venture Capitalists Choose Investments? Preliminary and incomplete Do not quote without authors permission How Do Venture Capitalists Choose Investments? by Steven N. Kaplan and Per Strömberg* First Draft: August 2000 This Draft: August 2000

More information

A Simpler Plan for Start-ups

A Simpler Plan for Start-ups A Simpler Plan for Start-ups Business advisors, experienced entrepreneurs, bankers, and investors generally agree that you should develop a business plan before you start a business. A plan can help you

More information

Strategic Sourcing for Services: Assessing the Balance between Outsourcing and Insourcing

Strategic Sourcing for Services: Assessing the Balance between Outsourcing and Insourcing Strategic Sourcing for Services: Assessing the Balance between Outsourcing and Insourcing Eric K. Clemons Lorin M. Hitt Discussion Draft 4.2 First Draft 19 May 1997 Current Draft 16 June 1997 1. Introduction

More information

Occupation-Specific Human Capital and Local Labor Markets. Jeffrey A. Groen, U.S. Bureau of Labor Statistics

Occupation-Specific Human Capital and Local Labor Markets. Jeffrey A. Groen, U.S. Bureau of Labor Statistics BLS WORKING PAPERS U.S. DEPARTMENT OF LABOR Bureau of Labor Statistics OFFICE OF EMPLOYMENT AND UNEMPLOYMENT STATISTICS Occupation-Specific Human Capital and Local Labor Markets Jeffrey A. Groen, U.S.

More information



More information

Making Smart IT Choices

Making Smart IT Choices Making Smart IT Choices Understanding Value and Risk in Government IT Investments Sharon S. Dawes Theresa A. Pardo Stephanie Simon Anthony M. Cresswell Mark F. LaVigne David F. Andersen Peter A. Bloniarz

More information


THE ECONOMICS OF EMPLOYEE BENEFITS THE ECONOMICS OF EMPLOYEE BENEFITS Darren Lubotsky Department of Economics and Institute of Labor and Industrial Relations University of Illinois at Urbana-Champaign August 2004 Chapter prepared for Employee

More information

MSc Facility & Environment Management The Bartlett School of Graduate Studies MSc Report

MSc Facility & Environment Management The Bartlett School of Graduate Studies MSc Report MSc Facility & Environment Management The Bartlett School of Graduate Studies MSc Report Matt Woodhouse 15 September 2010 1 If outsourcing sells, is anyone buying? Table of Contents Abstract... page 4

More information

The Employment Cost Index (ECI) is a quarterly

The Employment Cost Index (ECI) is a quarterly The Employment Cost Index: what is it? Understanding the characteristics of the Employment Cost Index aids in identifying the cost pressures measured by this important Government statistic pressures that

More information

As Science Evolves, How Can Science Policy? *

As Science Evolves, How Can Science Policy? * As Science Evolves, How Can Science Policy? * Benjamin F. Jones Northwestern University and NBER May 2010 Abstract Getting science policy right is a core objective of government that bears on scientific

More information

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. The Tenuous Trade-Off between Risk and Incentives Author(s): Canice Prendergast Source: The Journal of Political Economy, Vol. 110, No. 5 (Oct., 2002), pp. 1071-1102 Published by: The University of Chicago

More information

Behind the Help Desk: Evolution of a Knowledge Management System in a Large Organization

Behind the Help Desk: Evolution of a Knowledge Management System in a Large Organization Behind the Help Desk: Evolution of a Knowledge Management System in a Large Organization Christine A. Halverson IBM Research 650 Harry Rd San Jose, CA. 95120, USA Thomas Erickson IBM Research

More information

Valuing Young, Start-up and Growth Companies: Estimation Issues and Valuation Challenges

Valuing Young, Start-up and Growth Companies: Estimation Issues and Valuation Challenges 1 Valuing Young, Start-up and Growth Companies: Estimation Issues and Valuation Challenges Aswath Damodaran Stern School of Business, New York University May 2009 2 Valuing Young,

More information


ICT SYSTEMS MARKETING PLAN ICT SYSTEMS MARKETING PLAN by Ladan Mehrabi Graduate Diploma in Business Administration, Simon Fraser University, 2006 Bachelors Degree in Electronics Engineering, Tehran Azad University, 2001 PROJECT

More information

You re In then You re Out: The Incidence and Effects of Being Outsourced

You re In then You re Out: The Incidence and Effects of Being Outsourced You re In then You re Out: The Incidence and Effects of Being Outsourced Deborah Goldschmidt Johannes F. Schmieder Boston University Boston University NBER, and IZA November 2014 Abstract The nature of

More information

Designing Multi-Person Tournaments with Asymmetric Contestants: An Experimental Study. Hua Chen, Sung H. Ham and Noah Lim +

Designing Multi-Person Tournaments with Asymmetric Contestants: An Experimental Study. Hua Chen, Sung H. Ham and Noah Lim + Designing Multi-Person Tournaments with Asymmetric Contestants: An Experimental Study Hua Chen, Sung H. Ham and Noah Lim + + All authors contributed equally. Chen: Bauer College of Business, University

More information

Evaluating the Effectiveness of Internet Marketing Initiatives

Evaluating the Effectiveness of Internet Marketing Initiatives MANAGEMENT STRATEGY MEASUREMENT MANAGEMENT ACCOUNTING GUIDELINE Evaluating the Effectiveness of Internet Marketing Initiatives By Marc J. Epstein and Kristi Yuthas Published by The Society of Management

More information

Unlocking the business value of social, mobile, cloud and big data

Unlocking the business value of social, mobile, cloud and big data Unlocking the business value of social, mobile, cloud and big data How businesses are turning to Managed Service Providers to optimize transformative technologies June 2013 Contents Scope of research/methodology

More information


HOW SAAS CHANGES AN ISV S BUSINESS HOW SAAS CHANGES AN ISV S BUSINESS A GUIDE FOR ISV LEADERS Sponsored by Microsoft Corporation Copyright 2012 Chappell & Associates Contents Understanding the Move to SaaS... 3 Assessing SaaS...3 Benefits

More information

Do Retiree Health Benefits Cause Early Retirement?

Do Retiree Health Benefits Cause Early Retirement? Do Retiree Health Benefits Cause Early Retirement? David M. Linsenmeier Princeton University November 2002 I would like to thank Orley Ashenfelter, Linda Bilheimer, David Blau, Melissa Clark, Henry Farber,

More information



More information

Focus Groups as Qualitative Research

Focus Groups as Qualitative Research Focus Groups as Qualitative Research PLANNING AND RESEARCH DESIGN FOR FOCUS GROUPS Contributors: David L. Morgan Print Pub. Date: 1997 Online Pub. Date: Print ISBN: 9780761903437 Online ISBN: 9781412984287

More information

Table of Contents. A Matter of Design: Job design theory and application to the voluntary sector 1

Table of Contents. A Matter of Design: Job design theory and application to the voluntary sector 1 Table of Contents Introduction................................................. 3 Setting the Context........................................... 5 Why Job Design?.............................................

More information