Venture Capitalists Investment Process, Criteria, and Performance
Frdric Martel*
Draft: November 2006
Abstract
According to previous studies on venture capital, deal selection skills are critical to financial success. The objective of this study is to examine and validate empirically the venture capitalists decision criteria (1) relevant for their investment decision that are (2) predictive of subsequent deal success, and/or (3) survival outcomes.
Using the information collected by a successful European VC firm while evaluating its entire deal flow of 2219 investment opportunities received between September 1996 and August 2001 (e.g. ex-ante data) plus information gathered from a questionnaire answered in September 2001 by the same VC (e.g. ex-post data), we observe that our analyses results differ significantly depending on whether we use the ex-ante or the ex-post questionnaire answers. This observation is consistent with the findings of recent cognitive psychology studies on ex-post research methodologies, as for instance, interviews and surveys.
Focusing on our ex-ante data, we identify the criteria that are significant for investment decision: Management and Financials. We also determine the criteria that are predictive of deals future financial success: Management and Financials and/or predictive of company survival, such as Resistance to Risk, Barriers to Entry, and Innovation. We note that a structured investment approach may assist VCs to adapt their investment criteria to market conditions; thereby improving their prospects of financial success.
*Frederic Martel, University of Lausannes Hautes Etudes Commerciales (HEC), e-mail: [email protected]
2
Content I. Introduction ........................................................................................................................... 3 II. Literature review .................................................................................................................. 5
2.1. Investment process: background................................................................................................................... 5 2.2. Data Sources: background ............................................................................................................................ 5 2.3. Criteria linked to investment decision .......................................................................................................... 6 2.4. Criteria linked to success .............................................................................................................................. 8 2.5. Criteria linked to survival ............................................................................................................................. 8 2.6. Limitations encountered in past research ..................................................................................................... 9
(1) Data accessibility issues.............................................................................................................. 9 (2) Reporting bias issues................................................................................................................. 10 (3) Investment criteria identification issues ................................................................................... 10 (4) Sample selection representativeness issues .............................................................................. 10 (5) Data pairing issues .................................................................................................................... 11 (6) Decision model construction issues.......................................................................................... 11
III. Research Method .............................................................................................................. 11 3.1. Empirical context ........................................................................................................................................ 11 3.2. Data Collection............................................................................................................................................ 12
(1) Interview Data ........................................................................................................................... 12 (2) Deal Flow Data.......................................................................................................................... 12 (3) Ex-ante data............................................................................................................................... 12 (4) Outcome data............................................................................................................................. 13 (5) Market Data............................................................................................................................... 13 (6) Ex-post data............................................................................................................................... 14
3.3. Validity, analysis and reliability ................................................................................................................. 14 (1) Data sample specificity ............................................................................................................. 14 (2) Data sample validation.............................................................................................................. 15 (3) Univariate analysis .................................................................................................................... 18
IV. Results and Discussion ..................................................................................................... 19 4.1. Representativeness Analysis....................................................................................................................... 19 4.2. Deal Flow Analysis ..................................................................................................................................... 19 4.3. Investment Process Analysis....................................................................................................................... 20 4.4. Investment Criteria Analysis....................................................................................................................... 20
(1) Interview Analysis..................................................................................................................... 21 Criteria analysis - Differences between 3 criteria types ................................................................ 21 (2) Enacted Criteria Analysis.......................................................................................................... 23 (3) What other VCs seem to do differently .................................................................................... 26
4.5. Decision models comparison ...................................................................................................................... 27 V. Conclusion and Future Research........................................................................................ 28 Acknowledgements................................................................................................................. 30 References............................................................................................................................... 31
3
I. Introduction
Venture capitalists (VCs) typically set up specially designed legal structures such as limited
partnerships, commonly called funds, to raise money and then to invest in a diversified portfolio
of companies via multiple rounds of financings. Only the most promising ventures receive follow-
on investments. According to Venture Economics (1988), over a 16-year period, more than one-
third of 383 investments made by a group of VCs resulted in an absolute loss, and about two-thirds
resulted in capital returns of less than double the original amount invested, with only a select
number of deals generating outstanding returns. Looking at the industrys reported returns, wide
gaps exist among the impressive returns generated by a few top-performing VCs, the industry
average returns, and the industrys worst performers.
Consistently delivering above-average returns while mitigating risks is crucial for VCs. It is
also a key challenge because, according to recent studies such as Kaplan and Schoar (2003) and
Cochrane (2005), on average, VC investing is not more profitable than investing in the stock
market, but is much riskier1 due to the high stochasticity of returns (returns come from few
financial successes), low liquidity, lack of transparency, and diversity (jurisdictions, industries,
market cycles, and financing stages differ). Yet the most successful VCs manage to deliver year-on-
year and fund-after-fund consistently high above-average performance, typically thanks to
proprietary skills and/or experience (Kaplan and Schoar, 2003). While published research on the
possible nature of these proprietary skills is extensive, it is far from conclusive.
Research on performance generating skills2 is far from comprehensive because, in part, of
the complexity of the VCs tasks, the opacity of their operating procedures, and the lack of reliable
research data. To date, there has been practically no study published on VCs investment screening
capabilities that is based on unbiased data recorded live at the time of the deals evaluation (e.g.
ex-ante) on real VC deals over an extended period of time that are paired with their subsequent
financial performance. So far, most studies have used either experimental designs or ex-post
evaluations, where possible biases cannot be controlled or quantified.
The purpose of this study is to identify and validate empirically criteria (1) relevant for
VCs investment decision, (2) predictive of subsequent deal success, and/or (3) survival3 outcomes,
using principally ex-ante data, where identified biases are controlled for. In addition, our
investigations aim to identify and segregate some of the assumptions flaws underlying published
studies that rely solely on ex-post design and expose some of the potential biases in deal selection
(by showing that different criteria predict VC investment and success).
1 Shepherd (1999) defines a ventures risk as the potential for the venture failure or bankruptcy 2 e.g. individual and organizational skills, know-how, good practices, or processes. 3 e.g. Shepherd (1999) defines survival as the probability that this venture will continue to participate in the market.
4
To our knowledge, this study is unique because it is the first that combines (1) interview
data, (2) ex-ante data gathered on a single representative VCs complete deal flow of 2219 deals,
(e.g. this VC), (3) ex-post data on a subset of no less than 198 deals, and (4) performance data on
all 2219 deals present in the database. All this information was collected to analyze one VCs
investment behavior over a 4.5 years period of turbulent financial markets in the
Telecommunication, Information Technology, Media, Entertainment Industry, (e.g. TIME) industry
sectors.
Our data on deal flow and investment criteria is analyzed on 5 main axes. First, we use
interviews to review the VC investment process and identify the investment criteria this VC says it
used, or its espoused criteria. These are: (1) Management, (2) Scalability, (3) Barriers to Entry,
(4) Exit Opportunities. Second, we regress its investment decisions to the ex-ante information it
collected on deals in order to derive its enacted or actual investment criteria, which are more
related to the Management and the Financials of the venture. Third, also via a regression analysis
using information on other VCs decisions and our ex-ante data, we find predictors that can explain
other VCs investment decisions, which seem to be based also on Management and Financials plus
Product Development, Potential Partnerships, and Existing Customer base.
Fourth, we pair venture criteria to three possible deal outcomes: (1) financial success, (2)
low growth maturity (examined within this paper under survival), or (3) bankruptcy or venture
failure to identify the predictors linked to company survival (e.g. Barriers to Entry, Resistance to
Risk, and Management) and the predictors linked to financial success (e.g. Management and
Financials). Should these predictors be different from the investment criteria, it would imply that
VCs are not always using the right investment decision criteria needed to optimize their financial
success or the investments survival.
We note that this VCs significant investment criteria are highly correlated to the predictors
of venture success. This would lead us to conclude that this VC was following a strategy focused on
investing in deals with a higher probability of financial success where one or two deals in a
portfolio compensate for a high level of total write-offs with few just surviving firms in the
portfolio. This conclusion is further corroborated by the actual performance this VC published in
December 2005.
Also, expanding on Mainprize et al. (2002), we derive a decision model based on our
ex-ante data and evaluate its effectiveness as a decision aid in comparison to the performances of
other published investment models. We find that our model may be a better decision aid than some
of the models presented in other studies.
We also note that a structured approach to venture investing may assist VCs to stay flexible
and in tune with their constantly evolving market conditions when armed with the relevant
investment criteria.
5
II. Literature review
2.1. Investment process: background
The importance of the selection criteria can be best understood by first reviewing the
published research on the wider framework of which they are part of: the Investment Process.
In Table 1, the findings of 6 studies on the multiple-stage investment process are presented.
These investigations all point out that process efficiency combined with deal selection skills and the
relevant investment criteria are crucial for financial success. Especially since, according to Fried
and Hisrich (1994), the VC investment decision-making process is also designed to reduce the risk
of adverse selection.
However, two factors make previous research inconclusive. First, according to Hisrich and
Peters (2002), the evaluation process combines objective information gathering and analysis with
the VCs intuition, gut feeling and creative thinking that is difficult to capture via traditional
research methods. Second, recent research suggests that VCs lack a strong understanding of how
they make investment and divestment decisions (Zacharakis and Shepherd, 2001).
Nonetheless, past studies have made a number of important assertions on the links between
investment process and investment criteria that could be also observed in our data. Firstly, Hall and
Hofer (1993) presents different criteria used at each of the screening and due diligence stages,
thereby supporting our observation that VCs tend to adapt their criteria at each stage of the
investment process. Secondly, since Shepherd et al. (2003) states that most VCs also adapt their
decision process to include experience gathered, in support of Mainprize et al. (2002), VCs may
benefit from structuring their investment processes so as to limit certain inherent potential biases
discussed in Section 2.6, which hamper retrospection analysis and experience building.
2.2. Data Sources: background
Previous studies have relied on data gathered via multiple investigations methods. Table 2a
presents a taxonomy compiled from findings of past studies on the investment criteria and their
investigation methods. Almost all studies rely on data collected via ex-post information gathering
methods reporting only espoused4 selection criteria, while only Kaplan et al. (2005a) is based on ex-
ante (e.g. Enacted) data.
In line with Sandberg et al. (1988), Bar et al. (1992), Muzyka et al. (1996), and Shepherd
(1997), which all question the validity of ex-post questionnaire based research, we develop the
following proposal:
6
Hypothesis 1: Espoused4 criteria data are different from enacted5 criteria data and
a-posteriori reported6 criteria data, which leads to significant differences in
analyses results depending on the data source used.
Support for hypothesis 1 would also credit research presented in Zacharakis (1995) and
Zacharakis and Shepherd (2001), which asserts that VCs, like many experts, appear to be poor at
ex-post introspection. Indeed, a-posteriori, VCs seem to overestimate their ability to predict the
future financial performance of potential deals.
2.3. Criteria linked to investment decision
According to Fried and Hisrich (1994), the VCs investment decision is crucial and bears
serious adverse selection risk due to the investments illiquidity nature. While Table 2a provides an
overview of a number of studies that investigated the VC investment criteria and their method of
investigations, Table 2b summarises the investment decision criteria identified in some of these
studies. Past research defines mainly four types of selection criteria in order of importance: the
competency of the management team, the attractiveness of the market, the attractiveness of the
opportunity (service or product), and the deals terms.
From the criteria presented in Table 2b, we select the 14 criteria that are present in our
database to develop the following hypothesis:
Hypothesis 2a: The 14 selection criteria Management, Innovation Potential, Stage of
Product Development, Resistance to Risks, Scalability, Barriers to
Entry, Existing Customer Base, Customer Potential, Potential for
Partnerships, Market Potential, Competition, Financials, Exit
Potential, and Valuation Attractiveness are statistically significant
predictors of investment.
Supporting evidence for hypothesis 2a would further credit the research methods used by
past researchers presented in Table 2a. It would also support the findings of the studies presented in
Table 2b. However, only some of these 14 criteria might be significant predictors of VC
investment. This may support Clarysse et al. (2005) that states that financial investment criteria are
more important than human resource criteria.
In addition, Burch (1986) comes to the conclusion that VCs would prefer an average
product/service idea with a top management team rather than a good product/service idea with a
4 As per Shepherd (1999), espoused criteria are the criteria VCs report or state they use when evaluating new venture proposals 5 As per Shepherd (1999), enacted or in-use can be defined as actual or truly used 6 As per Shepherd (1999), a-posteriori reported can be defined as reported via questionnaire or interview after the event
7
bad management team. Kaplan et al. (2006) contradicts this finding by implying that this is a bad
strategy.
Rah et al. (1994) finds that Market Attractiveness is by far the most predictive investment
criteria (30% more predictive then management capabilities).
Wright and Robbie (1998) claims that VCs place considerable emphasis and scrutinize in
detail of all aspects of a business. This normally includes sensitivity analysis of financial
information, discussions with personnel, and assessment of a great deal of intangible and subjective
information.
Interestingly, two studies report that the criteria used in different stages of the due diligence
process vary. Wells (1974) found that different criteria were applied at the screening and the
evaluation phases, moving from broad questions, such as portfolio fit, etc., to more deal specific
ones. Whereas, Hall and Hofer (1993) also identifies the two key criteria most used for the initial
screening stage: a) fit of the venture seeking financing with the VC firm's investment guidelines,
and b) long-term growth and profitability of the industry in which the business will operate.
According to them, however, in the proposal assessment stages, the key criteria change to: 1)
source of the business proposal that played a role in the VCs interest in the plan, and 2) proposal
previously reviewed by persons known and trusted by the VC.
To verify these two studies findings, we formulate the following hypothesis:
Hypothesis 2b: Selection criteria importance change at each stage of the due diligence
process.
Supporting evidence for Hypothesis 2b would credit research that link the use of different
criteria to different stages within the due diligence process. The supporting evidence would also
cast some questions on the more encompassing studies that do not take this observation under
considerations when making their analyses.
More generally, although all these studies determine certain investment criteria, contrary to
this study, none actually validate whether the VCs are right to use these criteria to optimize their
success rate.
Furthermore, just thinking that only these selection criteria are the most significant for VC
investment may be too limitative. Other factors, such as changes in the environment or market
conditions, may influence the venture (Abell 1978, Aaker and Day 1986). Also, requirements for
market success are likely to change with market evolution (Abell 1978, Shepherd et al. 2000).
Besides, Guidici and Paleari (2000), Inderst and Mller (2002), and Rider (2005) all report that
Market Conditions influence all aspects of the relationship between the VCs and the venture, and
specifically Valuation, Investment Terms, and Risk/Return expectations. Further empirical research
is certainly necessary in this area.
8
2.4. Criteria linked to success
Studies based on a-posteriori collected information demonstrate that VCs are successful at
predicting new ventures future successes (Dorsey, 1979; Sandberg et al., 1988; Kahn, 1987).
Riquelme and Watson (2000) presents a taxonomy of criteria associated with small and medium
sized company success. These analyses tend to indicate that emphasizing the qualifications of the
management, intensifying cooperation between the VC and portfolio companies, and ensuring a
strong (minority) shareholder position of VCs coincide with above average success (Schefczyk,
2001).
In Table 2c, we summarize the success criteria identified in a number of studies on venture
investing. From the criteria presented, we select the 14 criteria that are present in our database and
formulate the following hypothesis:
Hypothesis 3: The 14 selection criteria Management, Innovation Potential, Stage of
Product Development, Resistance to Risks, Scalability, Barriers to
Entry, Existing Customer Base, Customer Potential, Potential for
Partnerships, Market Potential, Competition, Financials, Exit
Potential, and Valuation Attractiveness are statistically significant
predictors of venture success.
Support for Hypothesis 3 would credit research presented in Table 2c. It would also support
the only correlational study, from Schefczyk (2001), that points out a possible correlation of one
selection criterion, emphasizing portfolio companies' managers' qualifications, to success.
While most VCs try to invest in ventures with both strong businesses and strong
management, some VCs claim to weigh one or the other more heavily (Kaplan et al., 2006).
Although it is possible for founders to adapt their style and become successful at running a
larger business, these founders often have neither the interest nor the skills necessary to do so
(Jayaraman et al., 2000). Stevenson and Jarillo (1990) argue that different skills are needed to
effectively manage the entrepreneurial challenges of a start-up versus the later administrative
challenges of an established firm. Hambrick and Crozier (1985) finds that successful start-ups are
more proactive in adding managers with more experience.
2.5. Criteria linked to survival
Table 2d offers a summary of the research findings on the investment criteria predictors of
company survival or non failure. Taking from this table the criteria for which we have data, we
formulate the following proposition:
9
Hypothesis 4: The 14 selection criteria Management, Innovation Potential, Stage of
Product Development, Resistance to Risks, Scalability, Barriers to
Entry, Existing Customer Base, Customer Potential, Potential for
Partnerships, Market Potential, Competition, Financials, Exit
Potential, and Valuation Attractiveness are statistically significant
predictors of venture survival.
Support of Hypothesis 4 would further credit the research made by Shepherd (1999) which,
among others state in Table 2d, identifies survival predictors for ventures and ranks them in order
of importance: (1) industry related competence, (2) high educational capability, (3) competitive
rivalry, (4) key success factor stability (i.e. Barriers to entry), and (5) lead time as innovators.
Furthermore, in line with some of the criteria mentioned above, according to Brderl et al. (1992)
and Lee and Lee (2004), failure of technology based ventures could be categorized into three
groups: (1) personal traits of the entrepreneur; (2) strategies and resource capabilities of the
venture; and (3) environmental conditions of a new venture.
2.6. Limitations encountered in past research
A number of publications raise suspicions on the VCs conspicuously accurate predictive
skills captured via a-posteriori research (Hofer and Sandberg, 1987; Hall and Hofer, 1993). Recent
studies even exist on the research limitations and possible biases in the field of VC investment
criteria ex-post analyses (i.e. Muzyka et al., 1996; Shepherd, 1997; Rider and Tetlock, 2005). These
studies highlight 6 key potential research limitations:
(1) Data accessibility issues
Conducting large-scale studies by gathering data about the investment process, criteria, and
investment financial performance of many VC firms is challenging due to the scarcity of reliable
unbiased data.
This lack of information is not surprising since VCs have several incentives to keep their
data opaque, including, amongst other things, to protect their competitive edge, to mitigate greater
potential fiscal scrutiny from their local tax authorities, and to avoid easy benchmarking to their
competitors, not to forget the well documented survivorship bias resulting from unsuccessful funds
going into liquidation without ever having published performance data. Also to be mentioned is the
issue of the time required to fill-out questionnaires when much research agrees that the scarcest
commodity a VC has is time, not capital (see e.g. Gladstone, 1988 and Quindlen, 2000).
Furthermore, as pointed out by Muzyka et al. (1996) and Birley et al. (1994), in sharp
contrast to the United States where a tradition of answering candidly to detailed questionnaires is
well established, European VCs can be much more restrictive with respect to cooperating with
researchers. In addition, European VCs may be somewhat annoyed by the abundance in recent
10
years of questionnaires sent indiscriminately by researchers attempting to bring clarity and
transparency to what some VCs often view as trade secrets (Muzyka et al., 1996).
(2) Reporting bias issues
Most past studies are based on VCs retrospective self-reporting (e.g. Tyebjee and Bruno,
1984), interviews and verbal protocols, and statistical analyses (see Mainprize et al., 2002, for a
review), or questionnaire responses rather than actual evaluations (for example, MacMillan et al.,
1987 and Robinson, 1987), and observations in laboratory setting, where typically VCs are asked
by researchers to evaluate deals and decide if they would invest, (Beim and Lavesque, 2004;
Mainprize et al., 2002). However, Hofer and Sandberg (1987), and later Shepherd (1997) and Rider
and Tetlock (2005), have shown that results obtained using these methodologies may diverge from
reality as VCs may not report accurately what they do and how they think. Even fewer studies
address whether VCs are right to use the investment criteria they say they use.
Indeed, according to Shepherd et al. (2003), decision-makers, especially experienced ones,
tend to overlook established objectives and instead rely on intuition and various heuristics when
deciding. Cognitive biases that may affect the way VCs address decisions include overconfidence
and anchoring, one form of which is to follow past practice and shun innovative alternatives
(Keeney, 1992).
Self-reporting has been shown to overstate the number of criteria actually used and to
understate the weighting of the most important criteria when compared to more sophisticated
decision-making techniques (see Stahl and Zimmerer, 1984; Riquelme and Rickards, 1992).
Ex-post methodologies (e.g. interviews, questionnaires, or surveys) assume that VCs can
accurately recall and explain without biases their own decision process. Zacharakis and Meyer
(1998) propose that methods that use surveys to ask VCs to revisit previous choices and use those
choices as a base to assess their decision process are biased.
(3) Investment criteria identification issues
Depending on the scope of the studies7 the most salient investment criteria vary (see
Riquelme and Watson, 2000). In addition, investment criteria may vary with market conditions (see
hypothesis 4 presented in section 2.4), VC investment agenda, VC style, etc. Therefore, to identify
the relevant investment criteria is challenging.
(4) Sample selection representativeness issues
Another important issue is the potential for sample selectivity (or non-random) bias in the
data. Since we are using sub-samples of the full database, it is important to show that these
7 Including or not deal sourcing, initial screening, ratings of business plans, due diligence, etc.
11
subs-samples are representative of the average deals applying for finance in our data set. The
Heckman procedure tests if such potential bias is present.
Unfortunately, there is a large number of other potential selection biases (e.g. is the deal
flow we analyze representative of all VCs deal flows, are the deal rankings representative of the
way other VCs rank deals, etc.) that can not be controlled for within our data sample. These biases
have to do with heterogeneity, i.e. VCs are not all alike and do not all follow the same objectives
with the same risk/reward investment profiles. Such biases would also influence more general
studies. We only have data from one of them so we can not generalize our findings to the industry
as a whole.
(5) Data pairing issues
Researchers lack paired ex-ante data on VC deal selection skills and the resulting
investment performance (Schefczyk, 2001). Therefore, so far, no meaningful comparison between
relevant investment criteria and subsequent performance can be found in literature.
(6) Decision model construction issues
Past research on the modeling of investment decision suggests that VCs do not necessarily
have a strong understanding of their decision criteria (e.g. Shepherd, 1997; Zacharakis and Meyer,
1998) and, instead, rely heavily on intuition (e.g. Hall and Hofer, 1993; Zacharakis and Meyer,
1998); face environments that are not conducive to learning and/or improving judgment (e.g.
Shepherd and Zacharakis, 2002); and may need to use decision aids to significantly improve their
judgment and decision making (e.g. Zacharakis and Meyer, 2000; Shepherd and Zacharakis, 2002).
However, decision models found in previous studies, such as Mainprize et al. (2002), are
limitative in their capacity to encompass all elements affecting VC decisions. Also, to present the
diversity of VCs individual and organizational deal-selection practices with a common model is an
issue.
III. Research Method
3.1. Empirical context
Our source
Most of our unique data comes from one VC company that was established in September
1996 in Switzerland. In 1998, it employed about 10 professionals, including staff. By August 2001,
its team was composed of 5 partners, 8 investment managers, plus assistants, accountants and
secretaries; a total of 25 professionals. Section 4.1 should demonstrate the representativeness of this
VC.
12
3.2. Data Collection
This study relies on the combination of analyses done on the following six data sources:
(1) Interview Data
Our first source of data consists of interview data on how the VC structured its investment
process and on its modus operandi. This data was collected via multiple interviews with its
investment managers and partners between 1999 and 2001. These interviews were intended as very
open discussions in order to capture a maximum of information. Subsequent interviews helped
obtain any necessary clarifications.
These interviews serve as the basis for Section 4.2 that describes a VCs investment process
and which is presented in Table 3. The VCs feedback to our studys preliminary findings also
helped us to further develop our understanding of investment process and investment criteria by
providing some reality checks and an opportunity to collect additional information and
perspectives.
(2) Deal Flow Data
The second source consists of the empirical data collected at the time of the due diligence
on the 2219 deals it screened between September 1996 and August 2001; all in all about 20
columns of potential information for each deal. Much information on the deal flow specifications
can be found in Table 4.
The quantity of deal flow information available for each 2219 observations varies
somewhat, as the investment opportunities that were more carefully evaluated have more complete
information (see Table 5). To improve the completeness of our database, in some cases, additional
information was hand-collected from the ventures websites, public and private databases, etc.
(3) Ex-ante data
The third data source consists of the ratings on 14 evaluation criteria selected and defined
by the VC in 1999, gathered on the deals at the time of their individual due diligence. The criteria
can be best illustrated by the 14 main questions the VC asked itself for each investment
opportunity:
Investment Criteria and Description Questions
Criteria Description question Management 1 Management How capable is the management to execute its business plan?
Product / Service 2 Innovation Potential How innovative is the product or service? 3 Stage of Product Development What is the current stage of the product development? 4 Resistance to Risks How risky is the project and what is the worst case scenario? 5 Scalability How scalable is the business? 6 Barriers to Entry How much is the product or company protected from copiers?
13
7 Existing Customer Base What is the quality of the current customer base? 8 Customer Potential How attractive/large is the potential customer base? 9 Potential for Partnerships What is the potential for this VC to create value added for the company
via partnerships with vendors, other portfolio companies, etc.? Market 10 Market Potential What is the market potential for the product or service? 11 Competition What is the status of the competitive landscape? Financials 12 Financials Are the financial plans reasonable and attractive? 13 Exit Potential What is the likelihood of having a successful exit via trade sale or IPO? 14 Valuation Attractiveness How attractive is the investment proposition for the VC?
Source: our reference VC
In our database, these ratings are available for a sub-sample of 198 deals.
The VC ascertained that the grades recorded for each project are the grades attributed to it
at the last stage reached within the due diligence process. Hence, if a project was defined as having
a management team with a grade of 2 (not very good) at the first evaluation stage, but then, during
the due diligence process, this VC revised its opinion on the management (for example because it
was strengthened by the arrival of an experienced CFO), the deal may have ended with a grade of 5
for Management (which would be captured in the database).
This VC also allocated to each criterion an investment decision weight. All criteria
multiplied by their respective weight could be added to compute a deals overall attractiveness
score. Unfortunately, as this VC was still experimenting with its evaluation model, it did not
actively modify the weights from deal to deal so that the variation of the weights in the ex-ante
database can not be properly analyzed. During our interviews, this VC informed us that it intended
to analyze these weights at some later date to refine and improve its decision process, provided
enough quality data was then available.
(4) Outcome data
The fourth data source consists of information collected on the performance of the deals,
the number of successive financing rounds, and financial outcomes (i.e. trade sale, IPO, still
private, or bankruptcy) on the 2219 deals between September 2001 and December 2004. The
outcome data was collected by consulting a number of VC specific databases (e.g. Venture One,
Venture Economics, Capital IQ, etc), company websites or web page archives (such as
web.archive.org) in order to establish whether the firm became a successful investment, is alive but
not a successful VC investment, or was bankrupt by December 2004. The financial outcomes of
the ventures are used in Section IV to evaluate whether certain investment criteria could be
predictive of performance.
(5) Market Data
Because stock market indices and deal flow are highly correlated, our data set does not
allow us to discriminate among (1) quarterly variations in market conditions measured by European
14
IT stock index, deal flow (as a proxy for demand for VC money), other VCs investment rate (as a
proxy for offer of VC money, survival, and success rates), (2) variations in this VCs activity
measured by the number of deals encountered every quarter, and (3) variations in the VCs
experience (as a function of time, the cumulated number of investments realized or exited).
As shown below, four time periods have been defined. Between September 1996 and
August 2001, the TIME industry, this VCs industry focus, underwent first a major expansion until
end of 1st Quarter 1999. A boom period followed from the 2nd Quarter 1999 to the 2nd Quarter 2000.
The boom period lead to a correction period between the 2nd Quarter 2000 and the 4th Quarter 2000.
Finally, the bust period started in the 1st Quarter of 2001, which is the last quarter for which we
have deal flow and/or ex-ante data.
EXPANSIONBOOM
FALLBUST
1996 1997 1998 1999 2000 2001
our VC's TIME industry focus
Figure 1: TIME Industry Chronology
(6) Ex-post data
The sixth data source consists of 98 answers to an ex-post multiple-choice survey 8 with 32
questions completed in August 2001 by the VCs investment managers. These 98 deals, 79 rejected
projects plus 19 investments, are all part of the 198 deals in the deal flow for which ex-ante
investment criteria ratings are available. The questions include the 14 investment criteria answered
ex-ante at the time of due diligence. The 98 deals were initially screened between the first quarter
of 1999 and the second quarter of 2001; e.g. 1 to 5 years before the questionnaire was answered.
The 19 ventures in which the VC invested involved in some cases multiple rounds of financings.
3.3. Validity, analysis and reliability
(1) Data sample specificity
As data from only one VC is available to us, our findings can only be specific to this VC
and no generalization is possible. The two main questions that can not be answered are: (1) Is the
VC representative of successful VCs in general and therefore can this studys findings be
8 For more information on our ex-post questionnaires design and an initial analysis, please review Birrer (2002). Alexander Birrer is the
author of the questionnaire also used in this study.
15
generalized to successful VCs?, and (2) Are the subset of VCs deal flow for which we have data
representative of the entire deal flow of this VC?
(2) Data sample validation
To support our analyses and validate our findings, of the following tests are done.
First, a correlation analysis between the ex-ante and ex-post data is made to determine
whether any salient difference exists. Ideally the analysis of the ex-ante and ex-post data should
lead to the same conclusions. This can however only be the case if the correlations between the two
data samples are high. If the correlations are low, we need to concentrate our analyses on the ex-
ante data, which we deem less prone to some of the biases presented in Section 2.6.
Second, to address the possibility of selectivity bias, we use the Heckman two-step
estimation procedure, or Heckit model9, on our ex-ante data. This analysis enables us to eliminate
biases in regression weight calculation due to censorship.
As describe by ., the following conditions need to be fulfilled to built the Heckman
model: 1) create a probit model for the propensity u, 2) compute the Mills ratio (lambda ) from u, 3) use the Mills ratio and the model variables x to build a regression model using only the
uncensored records, record with non-zero v, and iv) apply the model to the whole dataset
suppressing the dependence on x.
The Heckman model posits i) a value w for participating in an activity (working, stealing,
etc.), ii) a propensity u for engaging in this activity, and iii) a latent correlation between
participation and value. If the correlation between participation and value is model by explicit
variables, the correction is not needed and conveniently disappears. Nonetheless, censoring does
not always hurt a model. It is sometime possible to fit an accurate model using only the uncensored
records. Censoring hurts when it is biased in the sense that it is dependant on the independent
variable w. As previously said, the Heckman model eliminates bias caused by censoring on a
variable correlated to w and the scheme is defined as follows:
V = s(u) w
where:
V: censored value. It is a directly observable value censored by a biased selection process.
w: Value. Latent Gaussian r.v.
s( ): Step function. Equal to 1 for positive u; 0 otherwise. s(u) is observable
u: selection propensity. Latent Gaussian r.v.
Having described the scheme, the Heckman model is the following:
w = xt +
9 Heckman developed the Heckit model to investigate the value of womens wage (J. Heckman, 1974, 1979). It uses a binary probit model, a kind of single-neuron neural Network.
16
u = xt + where
X: vector of observable model variables possibly governing w and u.
(beta) and (gamma): vectors model of coefficients governing "w" and "u". (epsilon) and (eta): latent variables governing w and u. Having described the Heckit model, we can apply this rational to our data. Here, this
procedure allows the identification and isolation of any sample specific bias due to the potential
non-representativeness of our datas sub-samples on the total data sample. For this test, it is
necessary to segregate our data by dependent variables (i.e. invested/not invested, successful/not
successful, alive/bankrupt, etc.) and then compare each cluster groups so as to determine if
significant differences exist. The influence of the unbiased criteria (e.g. Predictors) on the possible
outcomes (e.g. Dependent Variables) is then highlighted as the inverse Mill's ratio.
So, in the first step, a probit regression selection model is estimated to calculate the inverse
Mill's ratio (because the error term of this model is normally distributed, which is one of the main
assumptions underlying the Heckman model) with a dummy variable (i.e. 1 if this VC invested; 0,
otherwise) as a dependent variable (Equation 1).
Our selection equation is as follows:
Selection equation 1: Dependent variables (DVi)i = a0 + 1 n (exa1) + 2 n (exa2) + 3 n (exa3) + ei ... Description of the variables Control variables a1 Source a2 Age a3 Country of Origin a4 Industry a5 Develop. Stage a6 Raising first round a7 Known Founder a8 Serial Founder a9 Successful Founder a10 Financed a11 OtherVcLeads_AK a12 OtherInvestorCommitted_AD
Predictors (exai) exa1 Financials exa2 ExitStrategy exa3 ValueAttractiveness exa4 RiskResistance exa5 MarketPotential exa6 Competition exa7 Innovation exa8 Management exa9 Scalability exa10 PotentialCustomers exa11 BarriersToEntry exa12 Partnership exa13 ProductDevelopment exa14 ExistingCustomers
Dependent variables (DVi) slctd Invested by this VC vcsbsq Financed by other VCs sccss Successful investment srvl Live company in Dec.04
n = Lambda n
Note that in the Heckman procedure, the residuals of the selection equation are used to
construct a selection bias control factor, which is called Lambda (), which is equivalent to the Inverse Mill's Ratio. is a summarizing measure that reflects the effects of all unmeasured
18
characteristics, which are related to observed selectivity biases to be isolated from our investment
criteria into and additional variable, Lambda..
In a second step, the estimated probabilities (Lambda n (n)) (including the inverse Mill's ratio to account for selection bias) are used as regressor in an Ordinary Least Square (OLS)
regression model to estimate the probability that a venture will receive VC backing. This model
employs the White (1980) robust standard error procedure that accounts the sample selectivity
separately via a heteroskedasticity10-consistent covariance matrix estimator (Equations 2)11.
Equation 2 - regression or observation equation: y* (unobserved)='w+u u~N(0,1 )
where y is observed (equals 1) if and only if y* is greater than 0; and y is not observed if
smaller or equal to 0. The variance of u is normalized to 1 only because, not z*, is observed. The
error terms, u and e, are assumed to be bivariate, normally distributed with correlation coefficient,
, and and are the parameter vectors.
The Heckman 2-step procedure should indicate if any selectivity bias due to sub-sampling
is present.
(3) Univariate analysis
Using our ex-ante data, we derive the significance of each selection criteria used via linear
regression for 4 situations. This gives us a preliminary description or summarization of individual
criteria. Thanks to our univariate analysis of multivariate outcomes, we can explore each variable
separately. We look at the range of values, as well as the central tendency of the values. We also
describe the pattern of response to the variable. We are conscious that linear models are very coarse
approximations of the Boolean logic underlying the mental processes. However, they are very
parsimonious, which is a key advantage for small data sets.
Yet, because deal characteristics may be inter-correlated, these univariate analyses do not
say a lot about the independent contribution of each criterion to the decision. For instance, this
VCs preference for network deals may actually be a preference for deals with good management or
conversely.
When evaluating possible limitations of our analysis, we realize that univariate analysis
alone may not be sufficient, especially for our complex yet in some case limited data sets.
Additional, and sometimes even contradictory, results may be found using multivariate analysis.
During the course of data analysis, a common practice is to include in multivariate analysis only
those variables that are statistically significant in univariate analysis. Such a habit is risky as some
10 A sequence or a vector of random variables is heteroskedastic if the random variables in the sequence or vector may have different
variances. The complement is called homoskedasticity. Source: Wikipedia. 11 The results are reported with and without allowing for sample selectivity.
19
variables not significant in univariate analysis may become significant in multivariate analysis. In
this study, we identify, with examples, four possible scenarios in which the above situation could
occur: (1) the effect of unbalanced sample size; (2) the influence of missing data; (3) an extremely
large within group variation, relative to between group variation; and (4) the presence of
interaction.
Further analysis could include principal component analysis with multivariate regression
analysis. It is highly probable that these analyses would yield entirely different findings. However
such analysis will not be covered in this version of the paper but could be part of future research.
IV. Results and Discussion
Our data can be analyzed from four different perspectives: (1) Deal Flow analysis, (2)
Investment Process Analysis, (3) Investment Criteria Analysis, and (4) Decision Model
Comparison.
4.1. Representativeness Analysis
This VC invested about Euro 140 million via three funds and created more than Euro 342
million in capital gains for its investors, generating an audited internal rate of return (IRR) of more
than 30% with an average exit multiple of 3.4. It managed to raise a new fund in 2000-2001 when
fundraising conditions were not favorable, and is still a very active investor that has made a number
of successful exits between 2001 and end of 2004.
The deal flow followed the variations of the stock market with a time lag of about one
quarter. It ranged from about 25-50 deals per quarter to a peak at 250-350 mostly early stage deals.
By the end of August 2001, this VC had invested in 62 individual ventures and made 17
follow-on investments selected from within its deal flow of 2219 deals.
The 79 investments resulted in 16 IPOs and 15 trade sales. From the remaining
investments, by the end of August 2001, only 17 firms had ceased activity, while 14 were still
active portfolio companies that in some cases had received additional financing by December 2004.
4.2. Deal Flow Analysis
Table 5 divides the information collected by the VC on each potential investment according
to a) information about deal focus in terms of Country of Origin and Industry, b) information about
the companys development stage, such as its Financing Stage, if it was Raising 1st Round
financing or if it was a Follow-up Deal, its Age, and if it already Has Revenues; c) information
about the origin of the deal, its Source, and d) information about what other VCs did, if Other VCs
lead the deal and whether Other Investors Committed; and e) information about the entrepreneur,
20
whether he was a Known Entrepreneur to the VC, a Serial Entrepreneur, and/or a Successful
Entrepreneur with a positive track record.
Table 6a and Table 6b compare the survival and the financial outcomes of the deals
received by this VC, and whether they have been financed by this VC, other VCs, or not financed at
all. Overall, it shows that this VC, on average, made 52% of Good Investment Decisions (GID),
significantly better than other VCs (29%) active during the same period (Table 6c), which confirm
this VCs above average selection skills.
In addition, after formally implementing its structured investment process in August 1999,
this VC showed a relative resistance to worsening market conditions: its success rate decreased to
36% (14/35) from 69% (18/26) while, in parallel, that of the average VC decreased steeper to 3%
from 54% (35/65) (see Table 6a).
The potential success rate of other VCs, had they followed this VC investment schedule,
would have been between 31% and 36% depending on the method, leaving an over-performance for
this VC of 16% to 21% (52% minus 31% to 36%) (see Table 6b), whether we take into account all
deals or only deals that we know became successes. Besides, this VCs over-performance also
exists within the Boom (fraction of deals invested: 9/16 = 56% vs. 13/61 = 21%) and the Fall
(fraction of deals invested 4/11 = 36% vs. 1/32 = 3%) of the financial markets (Table 6a). This
would tend to show that this VC was more successful than other VCs in selecting future profitable
deals.
Overall, as demonstrated in Table 6b, this VCs performance should qualify it as a
successful VC worthy of being considered for an insightful academic study.
4.3. Investment Process Analysis
Within its investment process, this VC defined itself 7 due diligence stages. The most
selective stages were the Manager (only 27% of the deals passed through), then the Team (32%),
then the Screening (61%) and finally the Investment Committee (IC) (80%). Table 5d shows that all
due diligence stages except the IC were good at identifying future successes from future failures,
thereby making good investment decisions. It seems that the Investment Committee (IC) actually
made, proportionally, the worst rejection decisions.
4.4. Investment Criteria Analysis
Background
Researchers generally agree that the VCs espoused criteria are not the sole basis for the
VCs investment decisions (Pries and Guild, 2002). A priori, VCs decision to invest in a specific
deal, or to promote it to the next level of due diligence, is an obscure, implicit mental and social
process integrating information about deals, market conditions, and VCs fund requirements. Many
factors may enter into consideration when a VC makes an investment. The ultimate objective of
21
decision analysis is to translate a decision into a structured set of explicit rules. If we suppose that
financial success is a major objective of VCs, we can expect these rules to be predictive for success.
For VC investors, a great differentiation can be made between survival goals and success
goals, as companies pursuing survival goals may require different strategies and resources than
those aiming for a rapid financial success. Accordingly, VCs focused on the maximization of
financial profits for their investors and themselves may pursue different investment strategies as
they get relatively little rewards for creating strong stable businesses that can not be sold quickly
with a large profit
As discussed above, the investment criteria can be examined from at least 5 angles: (1)
What the VC, during our interviews, says it does its Espoused Criteria; (2) what it does its
Enacted Criteria, and (3) what other VCs do differently, (4) what it reports ex-post to have done via
an ex-post questionnaire, and (5) what this VC should do.
(1) Interview Analysis
From our interviews we obtain this VCs espoused investment decision rankings for its
14 criteria by order of importance, as follows:
High Importance: (1) Management, (2) Scalability, (3) Barriers to Entry, (4) Exit
Potential.
Medium Importance: (1) Financials, (2) Valuation (Attractiveness of Valuation),
(3) Innovation, (4) Potential for Partnerships, (5) Market potential.
Low Importance: (1) Risk, (2) Competition, (3) Potential Customers, (4) Product
development, (5) Existing Customer Base.
We also learn that this VC explicitly considers as key investment criteria: a) how, at the
pre-screen/sourcing stage, the deal fits with its investment focus in terms of Country/Location,
Industry, and Financing Stage; and then b) how the deal scored on its 14 quantitative evaluation
scales, at all stages. Other information recorded in the database such as the age of the firm, the
source of the deal, the entrepreneur profile, and the investment round were not spontaneously cited
as investment criteria. We observe however that this VCs 14 investment criteria fall within 4 main
categories already identified in many previous publications (i.e. Vinig and de Haan 2001,
Zacharakis 1995).
Criteria analysis - Differences between 3 criteria types
Hypothesis 1: Espoused criteria data are different from enacted criteria data and a-
posteriori reported criteria data, which leads to significant differences in analyses results
depending on the data sample type used.
As described in Section 3.3, to test Hypothesis 1, we perform a comparative analysis of the
ex-ante/ex-post correspondences in order to evaluate the extent of potential biases. As presented in
22
Table 7, the low correlation between ex-ante and ex-post evaluation criteria based on their ability to
predict the VCs investment, deal survival and/or success shows that the data contained in the
ex-post questionnaire answers does not accurately reflect the VCs written deal evaluation at the
time of due diligence.
The scales that deviate the most between the ex-ante and the ex-post data are for the more
subjective criteria: Management, Financials, Exit Potential, Valuation, Potential Customers, and
Existing Customers. This observation is consistent with Dubini (1989), Goslin and Barge (1986),
and Gorman and Sahlman (1989) that assert that the evaluation of subjective criteria such as
Managerial Capabilities is challenging (see also Rah et al., 1994).
Looking at the potential causes for these deviations, from Table 6b, we observe that ex-post
knowledge does not help this VC to predict other VCs investment decision, suggesting that the
bias is specific to this VCs ex-post knowledge of its own decision, not of other VCs decisions. In
Table 7b, we test two potential reasons for this bias:
(1) Ex-post knowledge
Ex-post evaluation could be influenced by additional knowledge about the financial
outcomes of the deals. If so, lasting interaction with financed ventures should result in higher extra-
knowledge, and thus in more bias. Accordingly, Table 7a shows that the discrepancy between
ex-post and ex-ante ratings significantly12 decreases when only considering the rejected deals. This
tends to confirm that ex-post knowledge did bias retrospective questionnaire responses.
(2) Forgetting
Ex-post data could explain an increasing fraction of investment decision variance at later
due diligence stages. This suggests that ex-post information underestimates what happens at earlier
stages, probably because this VC better remembers deals it evaluated more in detail. Also, this VC
might be inclined to forget information about older deals and better recollect information on more
recent investment opportunities. Indeed, a majority of correlations get significantly13 better when
only recent deals are considered (Table 7a). In addition, the ex-post knowledge bias mentioned
above may not have taken affect yet since the outcome is still unknown for these younger deals.
Overall, the discrepancies between VCs ex-ante and ex-post evaluations seem to have
several causes, including forgetting and a posteriori knowledge. Although conclusions have to be
drawn carefully because our sample size is modest, it tends to support as Zacharakis and Meyer
(2001), which observes that when more or newer information becomes available to a particular
decision, the VC's ability to introspect about that decision process diminishes.
As Hypothesis 1 is supported, we decide arbitrarily to pursue our data analyses using
primarily our ex-ante data to reduce ex-post knowledge related biases.
12 p
23
(2) Enacted Criteria Analysis
Predictors of Investment
Hypothesis 2: The 14 selection criteria Management, Innovation Potential, Stage
of Product Development, Resistance to Risks, Scalability, Barriers to Entry,
Existing Customer Base, Customer Potential, Potential for Partnerships, Market
Potential, Competition, Financials, Exit Potential, and Valuation
Attractiveness are statistically significant predictors of investment.
Before presenting our data analyses results a word of caution is necessary. Indeed, our
results could be also biased by the VC specific investment motivation as it may not care about
ventures survival, but only about spectacular successes. In this case, for the VC, it may be optimal
to invest in deals with lower survival probability if the return (e.g. Internal Rate of Return or IRR)
probability is high. However, during our interviews, this VC did not specifically express this as a
strategic objective, although it was aware of its effect. The VC explained that at the time of
investment its return expectations were high for all deals and it was equally enthusiastic about all
initial investments without knowing which ones would be spectacularly successful.
As described in Section 3.3, via univariate analysis using our ex-ante data, we determine
the criteria that were both decisive (e.g. used by this VC for its investment decision), predictive of
survival, and/or of success, and encounter the following four situations:
Decisive and Predictive
As shown in Table 8, both the quality of the Management and the Financials of the
company are Decisive and Predictive since they are significantly correlated to this VCs decision to
invest, the subsequent investment by any VC, and the deals survival. In addition, they are also
decisive for other VCs, which suggests their universal value, in line with published literature (i.e.
Kaplan and Strmberg, 2004; Kaplan et al., 2005a). Finally, they correlate positively with our ex-
post data, suggesting that this VC also perceives them as a decisive investment criteria ex-post.
Accordingly, it is cited as a moderately important espoused factor by this VC, congruent with the
fact that they are decisive factors also for others (ex-ante and ex-post), in line with studies on VCs
espoused criteria (i.e. Hall, 1989, and Fried and Hisrich, 1994).
Management is decisive, cited as highly-important by this VC, and almost all
espoused criteria studies (i.e. Tyebjee and Bruno, 1984; Hall, 1989; Fried and Hisrich, 1994; and
Boocock and Woods, 1987). This would tend to show that VCs may attribute deal
successes/failures to management rather than to other deal characteristics. If so, one would expect
the VC to upgrade ex-post its ratings of the management of successful ventures. This assumption
could not be tested on this VCs deals due to the limited sample size and was not verified on all
deals since ex-post management ratings also do not correlate to survival or success.
24
Another explanation could be that, although financial success is important for VCs, it may
not be their only objective. Highly-rated management would be decisive because it predicts easier
work partners, fewer critics from investors, or even learning from entrepreneurs. To solve this
mystery, we would need more cases of paired ex-ante/ex-post ratings of portfolio companies
(successes and failures) and alternative measures of success, beyond financial success, such as
VCs intellectual and social satisfaction from working with each company.
Decisive, but not predictive
Financials, Potential Customers are decisive criteria but seem not to be highly
predictive of venture survival, especially in comparison to other criteria. However, from this VCs
ex-ante data, it is not predictive at all, whatever the model and the period considered. It is even
negatively correlated with success, which is consistent with Kaplan et al. (2006).
Predictive, but not decisive
Although it was cited as a highly-important criterion by this VC in our interviews and is
also often cited by other VCs, Barriers-to-Entry was not significantly used by this VC for its
investment decision, although it is predictive for venture survival.
How could this VC know it is a highly important criterion and not apply it? One
explanation could be that this VC only realized the importance of Barriers-to-Entry later on. This
VC might have used this information to update its espoused criteria only at the time of interview.
Such an update would be possible since Barriers-to-Entry correlates to survival and success ex-
post (see Table 7a).
Not predictive, not decisive
The criterion Scalability seems neither predictive nor decisive; but considered as a
highly important espoused criterion by this VC (see Table 9). Indeed, this VC does not use it
significantly for its decision, it is not useful as it is not predictive, but initially in our interviews the
VC claims that it is! When only looking at the ex-post ratings, Scalability appears predictive of
survival and success. Worse, ex-post, this VC has the right impression that it does not use this
criterion. During our feedback interviews, this VC could not give any explanation to this finding.
Therefore for Hypothesis 2, criteria Deal Value, Deal Source, and Follow-up
Financing are statistically significant predictors of VC investment, whereas Scalability is not
(even though this VC claims ex-post that it is).
Table 8 shows that the most significant predictors of the VCs investment are the quality of
the Management Team and the Financials outlook of the company. Next in line are the companys
Market Potential, Potential Customers, and Resistance to Risk. Therefore Hypothesis 2a could only
be partially supported since not all criteria were predictive of this VCs investments.
Not surprisingly, this VC preferentially selected deals that were: not from Eastern Europe;
in Financial Services and Telecom, but not in Telecom Services and Consulting; at a Public/Spin-
25
Off financing stage, but not at seed/start-up (even though the former were much rarer in the deal
flow); not raising their 1st round; sourced actively or from its network, not passively sourced; lead
by other VCs; in which other investors already had committed to invest; with a known, serial or
successful entrepreneur; and well rated on almost all 14 evaluation scales and factors.
Predictors of Success
Hypothesis 3a: The 14 selection criteria Management, Innovation Potential, Stage
of Product Development, Resistance to Risks, Scalability, Barriers to Entry,
Existing Customer Base, Customer Potential, Potential for Partnerships, Market
Potential, Competition, Financials, Exit Potential, and Valuation
Attractiveness are statistically significant predictors of venture success.
The univariate analysis in Table 5a shows some correlation between some deals
characteristics and success: Media industry; Pre-IPO stage; ratings of Risk, Competition (scale and
factor), and Partnerships. Among this VCs investments, we could also correlate Return on
Investments (e.g. ROIs) to deals; not from USA; not in Retail industry; Pre-IPO stage; not 2-3
years of age; and not with successful entrepreneurs. ROI was also correlated with ratings on
Financials, Competition, and Barriers scales, but only within the 80% confidence interval.
Therefore, Hypothesis 3a is supported, but only for a few criteria and with a lot of caution
due to the low sample size.
Although we suspect that Market Conditions may be a significant predictor of venture
success, we could not fully clarify our suspicion. Unfortunately, our 198 ex-ante rated deals only
contain 7 successes so the statistical significance is low, especially across periods.
While this VC decreased its investment size per deal with time, it kept its investment rate
per quarter almost constant. This was beneficial, since Market Condition was the best predictor of
success. For this VC, we cannot assess however if its good investment timing was the result of an
intelligent investment strategy or the automatic consequence of its funds inception dates.
First, VC-backed ventures achieve a higher survival rate than non-VC backed businesses
(Kunkel an Hofer, 1990, Sandberg, 1986, Timmons, 1994). Tyebjee and Bruno (1984) asks VCs to
evaluate previously examined plans on 23 criteria using a four-point scale and reduces these criteria
via factor analysis to five underlying dimensions namely (1) Market Attractiveness (size, growth,
and access to customers), (2) Product Differentiation (uniqueness, patents, technical edge, profit
margin), (3) Managerial Capabilities (skills in marketing, management, finance and the references
of the entrepreneur), (4) Environmental Threat Resistance (technology life cycle, barriers to
competitive entry, insensitivity to business cycles and down-side risk protection), (5) Cash-Out
Potential (future opportunities to realize capital gains by merger, acquisition or public offering). It
finds that investment decisions can be predicted from the perceptions of risk (e.g. new venture
failure) and return (e.g. financial performance).
26
Predictors of Survival
Hypothesis 4: The 14 selection criteria Management, Innovation Potential, Stage
of Product Development, Resistance to Risks, Scalability, Barriers to Entry,
Existing Customer Base, Customer Potential, Potential for Partnerships, Market
Potential, Competition, Financials, Exit Potential, and Valuation
Attractiveness are statistically significant predictors of venture survival.
Table 8 shows that Barriers-to-Entry, Resistance to Risk, Management, and Innovation
were the most predictive criteria, followed by Financials. Most of these criteria are related to risk
management so it could be logically expected to find them as greater determinants of venture
survival. Investors concentrating mostly on Resistance to Risk and Barriers- to-Entry
Company survival does not mean financial success, although they correlate across quarters
(coefficient .58, p
27
Therefore, we note that this VC seems to have different investment patterns than the
average other VCs and may have been more successful at adapting early to changes in market
conditions, a significant predictor of investment success. More research could be undertaken in this
area as well.
4.5. Decision models comparison
Zacharakis and Meyer (2000) focus on the task of screening venture proposals without
unduly rejecting high potential investments. Participants capacity to select the right investments is
compared to the predictions of (1) a bootstrap actuarial mode, structured to capture the cognitive
system of a decision-maker, which uses information factors previously identified by VCs as being
most important to making good investments decisions and (2) a second actuarial model based on
the findings of Roure and Keeley (1990), who identified predictors of success for technology
ventures. On predictions of success and failure, the bootstrap model outperformed all but one
venture VC (who achieved the same hit rate as the bootstrap model), and the Roure and Keeley
model outperformed over half of the VCs.
Zopounidis (1994) presents an overview of published VC investment decision models and
their relative performances. Mainprize et al. (2002) also presents an overview of the performance a
few models developed via conjoint analysis or plain bootstrapping. These two studies allow us to
compare the performance of this VCs actual decision model to other published models. The
outcome of this analysis is presented Table 9. It shows that this papers decision model fends
favorably to the decision models developed by researchers in laboratory contexts via various forms
of a-posteriori data collections and analyses.
Table 9 also shows that this VC, on average, made 52% of Good Investment Decisions
(GID), significantly better than other VCs (29%) active during the same period, which confirms its
selection skills. In addition, after formally implementing its structured investment process in
August 1999, it showed a relative resistance to worsening market conditions: its success rate
decreased to 36% (14/35) from 69% (18/26) while, in parallel, that of the average VC decreased
steeper to 3% from 54% (35/65).
A number of studies across a variety of decision contexts have found that such models often
outperform actual decision makers (see Camerer, 1981; Dawes, 1971; Osherson et al., 1997;
Zacharakis and Meyer, 2000). For additional detailed discussions on VC investment decision
modeling, please refer to Riquelme and Rickards (1994), Zacharakis and Meyer (2000), Shepherd
and Zacharakis (2003), Rider (2005), etc.
28
V. Conclusion and Future Research
This study analyses the investment process, investment criteria, and investment
performance of a single VC while attempting to limit known biases that have influenced the results
of previous studies. This study also attempts to tie its findings to the published theoretical
framework on VC investment criteria.
We show that, with access to ex-ante data, it is possible to explore and compare empirically
what a VC says it does, what it effectively does, what it should have done, and how it may differ
from what other VCs do.
Using our ex-ante data, we identify investment criteria that are both decisive for VC
investment, and/or predictive of deal future survival and/or venture subsequent success.
The comparison of the ex-ante data and the ex-post data highlights significant differences,
especially on the more subjective criteria. The differences confirm that our studys findings would
be substantially altered if they were solely based on the analysis of ex-post questionnaire answers.
This supports recent studies that highlight the poor introspection skills of VCs. Also, in support of
Zacharakis and Shepherd (2001), our findings confirm that in its ex-post questionnaire responses,
the VC displays overconfidence in its decision process and its predictive capabilities.
With regards to predictors of deal success, the criteria Business Potential and Barriers to
Entry were both significantly predictive. However, a deals financial success could have been
mostly determined by market conditions and/or country and industry, while the unique contribution
of the Deal Value factor to success prediction is nil. However, we feel that to support these findings
more research is needed with a larger dataset possibly from multiple VCs.
We note that most significant selection criteria of the VC were Management and
Financials and that the two criteria determining success were exactly the same. It proves that the
VC had correctly identified the right criteria to invest if it wanted to follow a strategy for success.
Although it did not directly admit it, it was obviously following such a strategy of success. This is
evidenced by its track record where few deals were outstanding success that more than
compensated for the numerous total write-offs in the portfolio.
With respect to predictors of deal survival, this study shows that a companys
characteristics could predict in part its survival 3 to 5 years later, irrespective of it receiving further
financing rounds and market conditions. Furthermore, Barriers-to-Entry was the most predictive
factor of company survival, followed by Business Potential (Market, Competition and
Innovation) and Deal Value.
The comparison of our investment decision model with those in published studies shows
that our model rates favorably with respect to investment successes, the main driver of a VCs
financial performance.
29
The use of checklists, scorecards, and templates for rating potential investments could be an
important initial step into formalizing an investment process. This area would benefit from further
definition of robust scales based on empirical results, as suggested by our findings in Section IV.
According to our results, scales should be constructed from ex-ante findings, since ex-post findings
do not give similar information structures. While ex-post introspection may help VCs improve in
the long term their performance, systematic recording of investment decision justifications live
(ex-ante) may bring more accurate results.
Our analyses also confirm that VCs could probably benefit from being able to
retrospectively analyze its deal flow to refine its investment criteria, as we have done, thereby
improving its effectiveness at selecting successful investments more systematically. However, a
cross-sectional study on many VCs would be necessary to get more information on the potential
added value of each element. Unfortunately, access to similar ex-ante cross-referenced data from
multiple VCs is difficult to obtain.
Indeed, it is logical that VCs update their criteria on the basis of the information they can
remember on deals at the time they learn about the outcome, typically 2 to 5 years after investing. If
this ex-post information differed from the ex-ante information used at the time of the investment
decision, VCs would not be able to improve their criteria. If we suppose that VCs can learn from
the outcomes of their decisions, we also expect that VCs increase their predictive power with
experience. Finally, if we suppose that VCs have strategic intelligence, they should adapt their
investment criteria to changing market conditions. However, until now, little research evidence
exists that demonstrates that specific deal selection skills impact performance.
Further research is also needed on VCs investment strategies with respect to the fact that a
fund has a defined number of years within its lifespan when it can make investments. The time
window of a funds investment opportunity should influence investment managers investment
appetite and investment strategy. Intuitively, one could expect that VCs invest in earlier stage
ventures at the beginning of a funds life and then gradually invest more in later stage transactions
(including follow-on investments for early deals) as the investment period ends. VCs would be
motivated to do this because during the second part of their funds life, the harvesting period, they
would want to exit all their investments. If they had invested at the end of their investment phase in
early stage deals, VCs would probably not be able to do follow-on investments with the same fund,
due to investment guideline restrictions. Unfortunately, the data set used for this study did not allow
for extensive or conclusive analysis of this phenomenon.
Unfortunately, this study does not show the interactions between selection criteria and
market conditions. Future research using a large enough sample size per period and covering a
whole investment cycle could be very revealing.
It is important to mention that while the actual criteria included the criteria that were
significant for the survival and/or the success of venture, the selection of the investment was only
30
part of the beginning of the value creation process aimed at generating outstanding returns for its
investors. This value creation, while crucial, falls outside the scope of this study, but could certainly
benefit from further in-depth research.
Although many questions still remain open for future study, researchers are making
progress towards unraveling the VC investment process and investment success/survival predictors.
These efforts should assist investors in avoiding poor investment decisions in the future; especially
in booming markets when irrational behavior may occurs (see Kaplan et al. 2005a).
Acknowledgements
No paper is solely the effort of its author. For their valuable inputs and support, I am also
grateful to all VCs who participated in the study and provided essential insights. Finally, I want to
thank Professor Didier Cossin, Professor Autio Erkko, Professor Benoit Leleux, Professor Per
Stromberg, and Professor Steve Kaplan for their academic support.
31
References
Abell, D.F., 1978. Strategic windows. Journal of Marketing, 42(3): pp. 21-26.
Almus, M. and Nerlinger, E.A., 1999. Growth of new technology-based firms: which factors
matter?, Small Business Economics 13(2), pp. 141-154.
Audretsch, D.B. and T. Mahmood, 1995. New Firm Survival: New Results Using a Hazard
Function. Review of Economics and Statistics, 77: pp. 97-103.
Autio, E., 1997. Early growth and external relations in new technology-based firms. Paper
presented at the ICSB conference, San Francisco, California, June 1997.
Autio, E. and Lumme, A., 1998. Does the innovator role affect the perceived potential for growth?
Analysis of four types of new, technology-based firms. Technology Analysis and Strategic
Management, 10(1), pp. 41-54.
Barr, P. S., Stimpert, J.L., & Huff, A.S., 1992. Cognitive change, strategic action, and
organizational renewal. Strategic Management Journal, 13: pp. 15-36.
Barry, Christopher B., 1994. New Directions in Research on Venture Capital Finance. Financial
Read Management, Vol. 23, Nr. 3, pp. 3-15.
Baum, Joel A.C. and Brian S. Silverman, 2004. Picking winners or building them? Alliance,
intellectual, and human capital as selection criteria in venture financing and performance of
biotechnology startups. Journal of Business Venturing, May2004, Vol. 19 Issue 3, pp. 411-
437.
Baum, Christopher F., Andreas Stephan, and Oleksandr Talavera, 2004. Macroeconomic
Uncertainty and Firm Leverage. Boston College Working Papers in Economics 602, Boston
College Department of Economics.
Beim, Gina and Moren Levesque, 2004. Selecting Projects for Venture Capital Funding: A Multiple
Criteria Decision Approach. Weatherhead School of Management.
Birley, S. and Westhead, P., 1994. A comparison of new businesses established by novice and
habitual founders in Great Britain. International Small Business Journal, 12: pp. 38-60.
Birrer, Alexander, 2002. Evaluation Criteria and Deal Breakers at Venture Capital Investments:
Evidence from a Swiss Venture Capitalist. Term Paper, Swiss Banking Institute of the
University of Zurich.
Boocock, Grahame and Margaret Woods, 1997. The evaluation criteria used by venture capitalists:
Evidence from a UK venture fund. International Small Business Journal, London, pp. 36-57.
Brander, James A., Amit, Raphael, and Werner Antweiler, 2002. Venture-Capital Syndication:
Improved Venture Selection vs. the Value-Added Hypothesis, Journal of Economics and
Management Strategy, Nr. 11 (3): pp. 423-452.
Brettel, Malte, 2001. Decision Criteria of Venture Capitalists: Empirical Analysis in International
Comparison (in German), WHU-Research Paper, Nr. 82.
32
Brderl, Josef; Preisendrfer, Peter; and Rolf Ziegler, 1992. Survival chances of newly founded
business organizations, American Sociological Review, Vol. 57 Issue 2, pp. 227-242.
Burch, J., 1986, Entrepreneurship, John Wiley and Sons, New York.
Camerer, C. F., 1981. General conditions for the success of bootstrapping models. Organizational Behavior and Human Performance, 27: pp. 411-422.
Carter, N. M., Williams, M. and Reynolds, P. D., 1997, Discontinuance among new firms in retail:
the influence of initial resources, strategy and gender. Journal of Business Venturing, 12, pp.
125-146.
Carroll, Glenn R., Hannan, Michael T, 1989. Density Delay in the Evolution of Organizational
Populations: A Model and Five Empirical Tests. Administrative Science Quarterly, Sep89,
Vol. 34 Issue 3, pp. 411-431.
Chorev, Schaul, and Alistair R. Anderson, 2006. Success in Israeli high-tech start-ups; Critical
factors and process. Technovation, Vol. 26 Issue 2, pp. 162-174.
Clarysse, Bart, Mirjam Knockaert, and Andy Lockett, 2005. How do Early Stage High Technology
Investors Select Their Investments? Working Paper, University of Ghent.
Cochrane, John H., 2005. The Risk and Return of Venture Capital, University of Chicago, Journal
of Financial Economics, Jan. 2005, Vol. 75 Issue 1, pp. 3-52.
Cooper, A. C., G. E. Willard, and C. Y. Woo, 1986. Strategies of high-