The Impact on Management Experience on the Performance of Start-Ups within Accelerators

We investigate the effects that the experience level of accelerator management teams has on the performance of the accelerators they manage. In particular, we examine how the collective business experience of the accelerator managers influences the survival and growth of tenant firms within the accelerator. The experience of accelerator managers is assessed from two perspectives: their own direct knowledge from operating entrepreneurial startups, and their ability to access the knowledge of others from their professional networks. The survival and growth of tenant firms is assessed as the hazard rates for successful exits (acquisitions) and unsuccessful exits (firm failures). We find evidence to suggest that increased knowledge of accelerator managers reduces the risk of firm failures and that this reduction can be attributed more to differences in the amount of direct experience the accelerator management team has as founders in startups, than to differences in connectedness to the ecosystem.


Dave valliere
is an associate professor at the Ted Rogers School of Management, Ryerson University, in Toronto, ON. valliere@ryerson.ca E ntrepreneurship is a force driving job creation, innovation, economic growth, and fueling healthy and sustainable communities. New venture founders contribute to economic growth through a diverse range of behaviors, including innovation, resource combination, and increased competitive pressures (Wong et al. [2005]). But starting a new venture is difficult. Involuntary failure rates of entrepreneurial ventures are high: Some studies indicate failure rates for new ventures might be as high as 30% over the first two years of operations (Headd [2003]), with failure hazard rates peaking at 5.2% after approximately 24 months of business operations (Van Praag [2003]).
A number of schemes have been devised to provide supports that could help mitigate the high rate of failure for nascent ventures. These include science parks, incubators, small business centers, and accelerators. According to Phan et al. [2005, p. 166], "science parks and business incubators are property-based organizations with identifiable administrative centers focused on the mission of business acceleration through knowledge agglomeration and resource sharing." Incubators are designed to be a laboratory to encourage innovation through the use of technology and to use entrepreneurial spirit to spur new venture creation and growth (Al-Mubaraki and Busler [2011]).
In contrast, a small business center is a nonresident source of expertise open to all members of a community. Typically funded by local governments, small business centers are unique in the fact that no explicit third-party selection takes place. Instead, all entrepreneurs in the governmental jurisdiction are supported. Finally, an accelerator is a program that offers mentorship, office space, and a small investment in exchange for equity. Accelerators provide young companies with mentorship to help build a powerful network in the early stages of the firm's development. Affiliation with the top accelerator programs can also provide a strong branding and legitimizing mechanism for young companies seeking resources from other sources.
Three key elements underlie the value added by accelerators: mentorship and the ability to learn from others who are more experienced, connectivity to a powerful network that can be leveraged by the new firm, and brand enhancement and signaling of legitimacy. The primary difference between an incubator and an accelerator is that the latter provides equity funds in addition to the space, network, and mentorship. These funds enable the founders to work full-time on the new venture. Unlike incubators, accelerators are mostly cohort-based and strictly limit the time that a tenant firm can remain in the accelerator.

WinTer 2014
Instances of accelerators are currently growing exponentially. One of the earliest institutions to use the model of accelerators was The Foundry, in 1998(Fishback et al. [2007). This was followed by Y-Combinator in 2005 and Techstars in 2006, which have since been credited as the leaders of the current accelerator models. There are now more than 150 private accelerators across the United States and Canada. Al-Mubaraki and Busler [2011] estimated that globally there are 7,500 incubators worldwide and that number is growing at an annual rate 33%.
With such a dramatic shift toward the use of accelerators to support start-ups, it is surprising that very little research has been done into the actual performance of this model, and the key drivers of this performance. As many accelerators are nascent themselves it can be difficult to assess their long-term viability, let alone short-term impact. It is unclear how performance of an accelerator should be defined and measured. It might be defined in terms of firm creation and job creation (e.g., number of successful exits from the accelerator, average lifespan of tenant companies, number of employees in tenant firms) or in terms of economic returns (e.g., investment ROI, revenues generated by the accelerator, amount of follow-on investment raised, increase in valuation of tenant firms). Where accelerator performance has been defined in some manner, very few accelerators make their performance data available for assessment. Moreover, even when performance is defined and measured, there is very little understanding of the factors that drive accelerator performance.
Since the value of accelerators is very much dependent on the support services they provide to tenant firms, performance probably depends significantly on the quality of the accelerator management team that provides the services. Accelerator management can add value it at least two different ways: they can have relevant personal knowledge and skills developed through their own direct experience in start-ups, and they can access the knowledge and skills of others through their professional networks. Accordingly, in this article we aim to examine the effects of these two dimensions of accelerator management on the performance of accelerators, where performance is defined through the number of successful and unsuccessful exits of tenant firms.

LITERATURE REVIEW
Business incubators first arose in the United States in the 1950s and spread widely from there during the 1980s (Adkins [2002]). Initially they aimed to provide basic infrastructure to start-up firms, such as working space, and shared resources, such as IT equipment (Allen and McCluskey [1990]). Over time they grew to add a range of support services, such as public relations, recruiting, legal, accounting, and group buying (Smilor and Gill [1986]). The latest generation of incubators augments these services with active coaching and mentoring, acceleration funding, and connection to external networks that facilitate access to suppliers, customers, and partners (Rice [2002]; Scillitoe and Chakrabarthi [2010]). These external network connections also provide greater legitimacy to the new firms (Bøllingtoft and Ulhøi [2005]). But much of the value of these services depends on the ability of tenant firms to make use of them and thereby enable the f low of knowledge and commercialization of innovation. This depends not only on the degree of support services provided by the incubator, but also on the absorptive capacity of the tenant firms (Rothaermel and Thursby [2005]).
The rapid increase in the number of incubators and in the services they provide has led to the emergence of a range of incubator types, with different stakeholders, objectives, and strategies. Types of incubators include those affiliated with universities and focused on the commercialization of academic research, independent commercial incubators seeking investment returns, government-based incubators for regional development, internal corporate incubators that empower innovation among employees, and virtual incubators that serve anyone with an internet connection (von Zedtwitz [2003]; Grimaldi and Grandi [2005]). This range of objectives is echoed in the diverse ways in which incubators may def ine their performancewhich may include financial ROI, portfolio hit ratio, number of firms created or sustained, or number of jobs created. Typically, the performance metrics of an incubator must ref lect the fact that its stakeholders include investors, who have portfolio risk, and the tenant entrepreneurs, who take on the risks associated with new venture creation. Von Zedtwitz [2003] has argued that, despite the range of stakeholder objectives and strategies, all incubators should be run as businesses-even the noncommercial ones. Such an approach is thought to increase the prevalence in incubators of the following success factors: clear performance metrics, entrepreneurial leadership, value-added services, selection entry criteria, and access to resources for tenants (Wiggins and Gibson [2003]; Aerts et al. [2007]).
Over time, there has been a co-evolution of the needs and expectations of tenant firms and the suite of services offered by incubators and accelerators (Wang et al. [2008]). As incubators develop a wider range of offerings, they attract firms more in need of these services. Bruneel et al. [2012] have observed that tenants of newer incubators, with broader service offerings, make more use of the available services than do tenants of older incubators, with fewer services. These firms are more likely to grow rapidly and achieve successful exits. While many accelerators are standalone for-profit or nonprofit entities, the most common types have some type of institutional affiliation-whether to universities, to investment funds, or to operating businesses. This affiliation has been observed to affect the ability to access follow-on capital and provide networked services. Chandra et al. [2011] found that incubators with greater degrees of affiliation have more access to partner services and more access to different funding sources. These in turn are believed to have a significant inf luence on the rate of successful exits tenant firms have from their incubators (Peters et al. [2004]).
Typically, incubators take in nascent firms or founder teams according to some set of entry criteria. These new firms are co-located with other firms to share knowledge and experiences and to leverage shared infrastructure. As the firms grow, they access more of the services provided by the incubator, and begin to draw upon the capital provided by an accelerator or by affiliated investment funds. At some point the firms are expected to "graduate" from the incubator, either by having achieved some development milestone (e.g., first revenues) or by the elapsing of a limited tenancy period (Bergek and Norman [2007]).
Although it is widely believed that incubators catalyze significant economic growth by supporting the launch of new ventures and the commercialization of innovations, there has been surprisingly little rigorous research to substantiate this claim (Hansen et al. [2000]; Phan et al. [2005]). Incubator performance is often poorly defined and measured. And the widely varying strategies of incubators can confound acceleration goals with "life support" goals that simply require the ongoing existence of the firms and the few jobs they create. The systemic operations of incubators as elements of the entrepreneurial ecosystem are not well understood, especially the dynamics underpinning the success or failure of tenant firms under different incubator models and strategies. Few entrepreneurship researchers have compared survival rates between incubators, and fewer still have proposed causal relationships that result in these rates (Phan et al. [2005]).
With this context for the objectives and operations of incubators and accelerators, the quality of individual management team is likely to be one important predictor of performance. Incubator management is responsible for setting strategy, developing entrance and exit criteria for tenant firms, creating and evolving the suite of services made available, and facilitating the use of these services by tenant firms. The extent to which these firms grow to successful exits depends very much on the knowledge that f lows to them during their incubation period. This is management's primary valuecreating activity-the f low of knowledge to tenant firms through their access of incubation services. In some cases, management possesses the relevant knowledge themselves, by virtue of their own prior experience as entrepreneurs dealing with the strategic and operational challenges of new venture creation (Wise [2013]). And in other cases the relevant knowledge is possessed by others, with management acting to broker that knowledge to their tenant companies by drawing up their own professional networks (Osborne [2004]; Papagiannidis et al. [2009]; Warren et al. [2009]).
Entrepreneurs often are forced to rely on social capital to mitigate the impact of scarce financial capital (Kim and Aldrich [2005]). The ability of accelerator management teams to drive action on behalf of tenant firms, through their own social capital, is a key benefit to the tenant firms (Tötterman and Sten [2005]). Thus accelerator management teams with more connectedness would be a source of real value for start-up founders. What is presently unclear is the degree to which these aspects of management-prior relevant experience and connectedness within the ecosystem-matter. We do not have empirical evidence for how they impact tenant outcomes, nor their relative importance.

EMPIRICAL STUDY
To investigate these research questions, we undertook an empirical examination of accelerator cases having some notable differences in the background of their management teams. We examined the duration of time that tenant firms spend in each accelerator, and whether the tenancies result in successful outcomes (acquisitions) or unsuccessful outcomes (firm failures)-as a general proxy for accelerator performance.
In 2006, serial entrepreneur David Cohen founded Techstars, a U.S.-based mentorship-driven start-up accelerator, in Boulder, Colorado. The other founding partners were successful entrepreneurs David Brown and Jared Polis, and VC investor Brad Feld. Techstars subsequently expanded to six other cities and also established an online accelerator. A "white label" version of the Techstars acceleration model is now being used as the basis for industry-specific accelerators by Disney (entertainment) and Barclays (financial services). Incubation and acceleration services at Techstars include a mentorship program of three months, with preference for firms that are technology oriented, typically web-based or other software firms. Start-up funding of $118,000 is provided to every firm that enters the incubator ($18,000 as seed equity and $100,000 as a convertible debt note). These investments are provided from a fund of over $400 million raised from a syndicate of 75 major venture capital funds and angel investors.
The Digital Media Zone (DMZ) is a Canadian incubator and accelerator owned and operated by Ryerson University in Toronto. It was founded in 2010 with a mission to assist with the commercialization of student-led innovations in digital media technologies. It was initiated by the president of Ryerson and is currently led by Valerie Fox, former executive coach and customer experience practice lead at IBM Canada. DMZ also serves as context and infrastructure for experiential educational programs such as an undergraduate digital specialization and a master's program in digital media that support prospective entrepreneurs with ideation and opportunity spotting. Incubation services at DMZ include a four-month semi-structured support program and an optional eight-month extension for which fees may be charged. Admission preference is given to firms with a business plan and prototype in the digital media space, and whose projects build on the university's reputation for experiential learning. Many of the support services available are student-led, in that they are provided and managed by university students specializing in entrepreneurship. Acceleration at the DMZ is enabled by a related investment fund, Ryerson Futures Inc. (RFI), which enables our comparison to Techstars. RFI manages approximately $3 million from two venture capital funds and a group of business angels and intends to hold investments of $50,000 in up to 50 DMZ start-ups that have the potential to scale into large firms over time.
The approach taken for this study was to compare the hazard rates between the two cases, both for successful exits and for unsuccessful exits. A "hazard rate" is the instantaneous rate of occurrence of either of these two events at a given point in time, given that the firm has survived up until that time. We also examine the covariant effects of the relevant experience and connectedness of the top management in each of the accelerators. We assume that each accelerator's sector focus and services remain constant over the tenancy of a firm and that the development stage and value of firms change over time (whether increasing or decreasing).

OPERATIONALIZATION
All firm-level data about the tenant firms of the accelerators were obtained directly from the accelerator management. Successful exits were defined as firms that left the accelerator because they were acquired and provided a return to investors. Unsuccessful exits were defined as firms that left the accelerator because they ceased operations, providing no return to investors. The duration of firm tenancy in an accelerator was calculated from entry and exit dates (year and month). In the case of Techstars, entry dates were available only by year and quarter. This duration data is necessarily right censored because many firms are currently still in their accelerators, and some have left their accelerators without providing an exit of either successful or unsuccessful type.
The operational experience of each accelerator management team was operationalized as the mean value for the members of the top management team of the accelerator, at the time each firm was admitted. The experience of each individual was assessed as the number of years of working experience they had meeting two criteria: the work must have been in a new venture context (excluding work in large corporate environments), and the work must have involved operational and strategic decision making (excluding work as a junior employee or technical expert). Such qualifying work is referred to as "founder years". Top managers of each accelerator were identified by reviewing the information on the accelerator websites. The data for the calculation of founder years for each of these individuals were obtained by assessment of their résumés posted on LinkedIn.
The degree of connectedness of each accelerator management team was operationalized by taking the average connectedness of the individual managers, as measured by the number of followers they had on Twitter. We also considered triangulating these measures with the number of connections listed in their LinkedIn profiles, but LinkedIn caps the reporting of this number at 500 connections, and almost all of the individuals managing the accelerators in our study had reached this cap, eliminating any variance in this potential measure.
Scores for founder years and connectedness were calculated for each accelerator location on a month-bymonth basis. This accounts for the changing makeup of the management teams over the period of study, from 2006 to present. In this approach we are assuming that tenant firms have access to the direct and brokered knowledge of the accelerator management team that was in place on the day of their acceptance into the accelerator.
Exhibit 1 provides a summary of the key characteristics of the two accelerators under examination. Of note are the similarities of objectives and the differences in business model, scale, and breadth of services.
Both accelerators have been active for several years and have had opportunity to establish portfolios of tenant firms and to realize exits (both successfully and unsuccessfully). Exhibit 2 provides a summary of the portfolio of firms in each accelerator. Of note are the differences in size of their total portfolios and the relatively smaller number of successful exits that have occurred from the DMZ.

ANALYSIS
Survival analysis was performed using SPSS software (version 22). Because the passage of time is itself a meaningful antecedent of accelerator exits, model parameters were calculated via maximum likelihood estimation. For firms that achieve a successful exit, time in the accelerator represents ongoing development of the business. For firms that instead have an unsuccessful exit, time in the accelerator represents the ongoing consumptions of limited resources without achieving development of the business.
Because there are two possible outcomes for firms, we perform separate hazard analysis for each. In the first analysis the hazard event is acquisition of the firm (successful exit). We therefore exclude from this analysis those firms that achieved unsuccessful exits (30 firms at Techstars, 14 firms at DMZ). In the second analysis the hazard event is failure of the firm (unsuccessful exit). We therefore exclude from this analysis those firms that achieved successful exits (31 firms at Techstars, 6 firms at DMZ).
For each of these analyses, we first conduct the Kaplan-Meier procedure to determine hazard rates and survival curve characteristics. We compare the results for Techstars and for DMZ using the Mantel-Cox log rank (Mantel [1966]), Breslow [1974], and Tarone-Ware [1977] test statistics. Log rank compares the survival functions by weighting all data points equally, Breslow weights all data points by the number of cases at risk at that time, and Tarone-Ware weights all data points by the square root of the number of cases at risk at that time. These measures can indicate whether any significant different exists between the hazard functions of the two accelerators.
We then estimate Cox regression models that include as independent covariates the accelerator management characteristics of founder years and connectedness. This analysis can indicate the sensitivity of the calculated hazard functions to changes in the independent covariates-which is to say, the degree to which a change in the founder-years or connectedness of the accelerator management team will result in a change to the hazard rates.

RESULTS
Exhibit 3 shows the results of the Kaplan-Meier analysis of successful exits, as a probability or cumulative hazard of the exit event occurring. The hazard curves for Techstars and DMZ appear to be very similar, and the three test statistics confirm that our data do not display a significant difference in hazard rates for successful exits. In achieving the acquisition of tenant firms (successful exits), Techstars and DMZ appear to be equally able.
WinTer 2014 e x h i b i t 1

Accelerator Characteristics
Exhibit 4 shows the estimated parameters for the Cox regression of successful exit hazard ratio on accelerator management team characteristics. The years of start-up founder experience available in the management team (FY) appear to have a slight beneficial effect on the hazard of acquisition exit for tenant firms. With each additional year of founder experience the accelerator management team has, the possibility of an acquisition exit for tenant firms increases by 1.9% (p = 0.784). In contrast, the average connectedness of the management team (Conn) appears to have no effect (p = 0.834) on the hazard of positive exit by acquisition.
Exhibit 5 shows the results of the Kaplan-Meier analysis of unsuccessful exits. The hazard curves appear to drift apart over time. The three test statistics indicate this difference exists, with both log rank and Tarone-Ware being significant (p = 0.10). With respect to the hazard of failure (unsuccessful exits), DMZ firms are at higher risk.
Exhibit 6 shows the estimated parameters for the Cox regression of unsuccessful exit hazard ratio on accelerator management team characteristics. The years of start-up founder experience available in the management team (FY) appear to have a beneficial effect on the hazard of failure of tenant firms. With each additional year of founder experience the accelerator management team has, the possibility of a failure exit for tenant firms decreases by 6.8% (p = 0.135). In contrast, the average connectedness of the management team (Conn) appears to have no effect (p = 0.128) on the hazard of exit by firm failure.

DISCUSSION
The reported results seem to be suggesting one clear message: the direct start-up experience of accelerator managers matters more than their connectedness to the ecosystem. Accelerator managers who have greater experience in founder roles with start-ups have a more Kaplan-Meier Analysis of Successful Exits significant impact on the success of their tenant firms than do accelerator managers with less personal experience but more network connections. And the benefit of this founder experience among accelerator managers shows up primarily as mitigation in the risk of the failure of tenant firms (unsuccessful exits).
To take a hypothetical example, Exhibit 2 suggests that, for a firm with two years tenancy in the accelerator, the risk of failure is about 20% if the accelerator is DMZ and 14% if the accelerator is Techstars. If, at that time, the accelerator management team is augmented by an individual with five years of direct founder experience, these risks of failure should fall to about 14.1% for DMZ and 9.8% for Techstars. These changes illustrate the significant reduction of risk that can be realized through the makeup of the accelerator management team.
We propose that there are five ways by which an experienced management team can reduce risk. First, accelerator managers with greater founder experience are likely to be more astute in screening and selecting candidate firms for tenancy in the accelerator. Their start-up experience may have sensitized them to notice potential strengths and weaknesses in candidate firms that less-experienced accelerator managers may not notice. Second, experienced managers may be better able to advise and assist tenant firms with implementation of strategies and business plans by virtue of their firsthand experience doing so in their own firms. Third, direct experience with the twists and turns of process of creating a new venture may have made experienced managers acutely aware of the need for f lexibility in implementation, and the likelihood of pivots or other dramatic shifts in strategy along the way. As a result, they may have a better sense of when to advise tenants to do likewise, and of how to persuade those entrepreneurs to abandon their previous plans. Fourth, the personal experience of these accelerator managers may give them greater individual credibility with tenant entrepreneurs. This may be particularly important when a firm faces some of the daunting, though perfectly normal, setbacks that will call for great perseverance on the part of the tenant entrepreneur, or when the individual entrepreneur is feeling overwhelmed by the many responsibilities they must shoulder. An experienced accelerator manager can credibly say, "I know what that feels like, and here are some things you can do that will help." Fifth, and finally, accelerator managers with personal experience in start-up founder roles may be better able to spot opportunities for collaboration and synergy among the e x h i b i t 5

Kaplan-Meier Analysis of Unsuccessful Exits e x h i b i t 6
Effects of Accelerator Management on Unsuccessful Exits portfolio of firms that are tenants in the accelerator. Their experience with the inner workings of start-ups may allow them to see potential for connections at an operational level that other, less-experienced accelerator managers would miss.
Some of the benefits of greater experience on the part of accelerator managers might also be obtained, though possibly to a lesser degree, by providing highquality mentorship or advisors to tenant firms. These are individuals with direct start-up experience that can be valuable to tenant firms, but who are not fully involved in the day-to-day operations of the accelerator. Their knowledge may be valuable to the firms, but they may be somewhat less accessible when needed. Mentors are not an everyday contact for firm entrepreneurs in the same way that the accelerator management team is, so they may not be close enough to see some of the subtleties going on in the firms. And advisors are still less connected and more transactional in their assistance to the firms. But they still can provide a complementary source of expertise to that available through the accelerator management team.
This, then, is the first of the limitations that should be recognized in this present research-the omission of mentoring from the explanatory covariates of the model. The model estimated here is highly exploratory and is suitable only for initial high-level conclusions. A more inclusive and robust model is still needed. Moreover, the dataset of just two case studies is very limited. A larger and more diverse sample of accelerators would provide a better basis on which to develop and test causal propositions about accelerator performance drivers. Next, the Kaplan-Meier analysis we employed to determine the hazard rates assumes the date of entry doesn't matter (a year of tenancy is the same, regardless which year it is). For many start-ups this is a reasonable assumption, given that their challenges are the development of internal operations and the external search for initial customers and markets. But for some firms and industry niches, there may be external business drivers that are affected by macroeconomic factors, such as the business cycle, which are not captured in the Kaplan-Meier procedure. Finally, it may be that the negative results observed for the inf luence of connectedness on hazard rates can be attributed to a somewhat weak operationalization of the construct. The number of Twitter followers for accelerator managers is indeed an indicator of the size of their professional networks. But it particularly measures the outward f low of knowledge and inf luence, from the manager to the network, rather than the inward f low of knowledge from the network. Moreover, we observed that this measure of connectedness was correlated with the measure of founder-years (r = 0.726, p < 0.001), potentially diminishing its value as a separate inf luence factor.

CONCLUSIONS
This article has investigated the impact of accelerator senior management on the hazard rates of various exits for the ventures they support. While accelerator management can contribute many things to the tenants of the accelerators they lead, this research focuses on two specific types of contribution: direct start-up founder experience and levels of connectedness to the entrepreneurial ecosystem. We find that the years of start-up founder experience available in the accelerator management team has a beneficial effect on the failure hazard of tenant firms. With each additional year of founder experience possessed by the accelerator management team, the probability of a failure exit for tenant firms decreases. In contrast, the degree of connectedness of the management team is not found to have an effect on the failure exit hazard of the tenant firms. Hence, we suggest that accelerators wishing to upgrade their leadership would do best to ignore how well-connected leadership candidates are, and instead focus on leaders who have significant direct experience as founders of start-ups. Viewed from the perspective of an entrepreneurial founder considering becoming a tenant at one or more accelerators, the number of years of founder experience of the senior accelerator management should be a major consideration in deciding which accelerator to join.
New venture founders contribute to economic growth through a diverse range of behaviors, but starting a new venture is difficult. Most new venture efforts do not achieve their desired results. The explosive growth of accelerators and incubators worldwide is meant to mitigate these odds by better positioning start-ups for long-term success. But in order to maximize the return on the investment made into accelerators and incubators, stakeholders should focus on developing management teams who have "been there, done that," instead of those led by management who are simply well-connected.

WinTer 2014
This exploratory research takes an initial look at the factors that may differentiate the performance of various accelerators. Being limited in scope, it serves primarily to suggest the value of continued investigations in this area. Future research should improve on this initial work by employing larger samples (more accelerators) and more types of accelerators (with different business models and different international contexts). Such studies should also consider the inclusion of mentors and better operationalization of the connectedness construct to obtain more robust results. Finally, we think there is great value to be found in more exploratory qualitative research to try to develop a theory of acceleration that can account for the effects we describe here, and others.

ENDNOTE
We would like to thank the attendees of the 6th Indonesia International Conference on Innovation, Entrepreneurship, and Small Business (IICIES 2014) for their valuable input on the original version of this article.