604 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 65, NO.
4, NOVEMBER 2018
How Data-Driven Entrepreneur Analyzes Imperfect
Information for Business Opportunity Evaluation
Ethem Çanakoğlu, S. Sinan Erzurumlu , and Yaman Omer Erzurumlu
Abstract—High market uncertainty impedes an entrepreneur’s
ability to evaluate the state of the market for a business opportunity.
For many entrepreneurial ventures, data collection and analysis
techniques and technologies are becoming an important source
to manage uncertainty. This trend is often referred to as “data-
driven entrepreneurship.” We consider a dynamic approach using
data to overcome market uncertainty for business opportunity-
related evaluations. In particular, we examine the entrepreneur’s
investment portfolio in which each investment generates expected
returns and some information about a specific aspect of the market
for a single business opportunity. We develop a model that analyzes
imperfect market data (e.g., financial, social, regulatory), while
factoring in the entrepreneur’s risk preference and operational
shortages of resources, routines, reputation, and regulations. Our
numerical findings show that, rather than pursuing the highest
expected returns, an entrepreneur may choose perfect information,
risk hedging, or market-controlling investments based on his/her
cash level and risk preference. Hence, the entrepreneur, fueled by
the availability of data analysis, could overcome uncertainties and
obtain better insights for business opportunity decisions.
Index Terms—Business opportunity evaluation, decision making
under uncertainty, data-driven entrepreneurship, entrepreneur-
ship, operational entrepreneurship, operations management (OM),
partially observed Markov decision process (POMDP).
I. INTRODUCTION
NDERSTANDING the nature and sources of uncertainty
U that underlie entrepreneurial decision making have been
a primary focus of entrepreneurship research [1], [2]. The infu-
sion of data analysis techniques (i.e., inspecting, transforming,
and modeling data with the goal of supporting decision-making)
and technologies (e.g., data analytics) in entrepreneurship has
generated new ways of dealing with uncertainty [3], [4]. For
example, the constant flow of “big data” acquired through so-
cial media apps (e.g., Twitter) has been analyzed to overcome
Manuscript received September 1, 2016; revised April 11, 2017, September
29, 2017, January 24, 2018, and March 2, 2018; accepted April 2, 2018. Date of
publication June 3, 2018; date of current version October 27, 2018. The work
of S. S. Erzurumlu was supported in part by the Babson Faculty Research Fund.
Review of this manuscript was arranged by Department Editor N. Jolkelar.
(Corresponding author: S. Sinan Erzurumlu.)
E. Çanakoğlu is with the Department of Industrial Engineering, Bahçeşehir
University, İstanbul 34353, Turkey (e-mail:,ethem.canakoglu@eng.bau.edu.tr).
S. S. Erzurumlu is with the Technology, Operations, and Information Man-
agement Division, Babson College, Babson Park, MA 02457 USA (e-mail:,
serzurumlu@babson.edu).
Y. O. Erzurumlu is with the Department of Engineering Manage-
ment, Bahçeşehir University, İstanbul 34353, Turkey (e-mail:, yamanomer.
erzurumlu@eng.bau.edu.tr).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TEM.2018.2826983
0018-9391 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
opportunity-related uncertainties in healthcare [5]. A growing
number of venture capitalists have employed automated data
analysis techniques to evaluate business investments (e.g., [6],
[7]). We refer to the trend of data-driven techniques and tech-
nologies in shaping activities of the entrepreneurial process (i.e.,
opportunity recognition, development, and evaluation) as “data-
driven entrepreneurship.”
Nevertheless, business opportunity evaluation with a data-
driven technique may not be an easy or direct process. The
success of the business opportunity is subject to external market
factors, including market conditions for entrepreneurs in gen-
eral [8] and regulatory frameworks affecting access to consumer
and labor markets and finance [9], [10]. The entrepreneur may
have a very limited or no control over such external conditions
[11]. Therefore, the information flow necessary to conclude the
economic outlook of the market—positive or negative—for the
business opportunity may not be available (i.e., the market infor-
mation may be imperfect). Furthermore, the “true market” may
be hidden when that information is not observable [12]. For
example, in the developing clean energy market of Turkey, the
government’s mixed signals on a feed-in-tariff (a government
policy mechanism aimed to accelerate investment in clean en-
ergy) made it difficult for entrepreneurs faced with the shortage
of resources to evaluate the prospects of the opportunity [13].
In this paper, we study the business opportunity evaluation
from the data-driven perspective of entrepreneurship, and
ask: How could the entrepreneur analyze imperfect market
information in order to evaluate the business opportunity?
Furthermore, when the entrepreneur’s resources are lacking,
routines are nonexistent, reputation has not been established,
or operating regulations are inadequate [14], [15], these short-
ages of resources, routines, reputation, and regulations pose
operational constraints—we refer to as operational shortages
of the 4Rs—on overcoming market uncertainty. In addition,
the entrepreneur, in general, evaluates an opportunity based on
his/her individual risk preferences (e.g., high, medium, or low
risk aversion) [2], [16].
From the data-driven perspective of entrepreneurship,
entrepreneurs with different risk preferences and operational
shortages could better deal with imperfect market informa-
tion, so long as they could continuously monitor sources of
uncertainty [4] and analyze accumulated data to obtain insights
for decision making [17] (e.g., [18]). Importantly, data-driven
entrepreneurship demands dynamic and algorithmic data
analysis techniques to evaluate the business opportunity. In
this paper, we explore how to analyze real-time and imperfect
ÇANAKOĞLU et al.: HOW DATA-DRIVEN ENTREPRENEUR ANALYZES IMPERFECT INFORMATION
data (e.g., data or statistics from external factors forming an
observable process) in order to make an optimal portfolio of
investments.
In particular, information from each external factor, as deter-
mined by the investment portfolio, recognizes a different degree
of impact on the entrepreneur’s evaluation of the state of the mar-
ket for the business opportunity (i.e., to what degree the market
supports the business opportunity). Hence, the entrepreneur’s
investment portfolio in our study has a dual role: it not only
generates economic returns, but also provides information about
market observability and control over time. To understand the
impact of entrepreneur’s risk preference and operational short-
ages on the investment portfolio, we examine the tradeoff be-
tween expected returns and the level of market information
(e.g., by choosing investments with high returns, but a lower-
level of information): How would the entrepreneur’s risk pref-
erence and operational shortages affect the investment portfolio
that balances the tradeoff between expected returns and market
information?
We analyze external factors in complex environments of “true
uncertainty” where no basis exists for classifying the states of
the market [1]. That is, the market will not reveal its state di-
rectly, but may provide related information via another pro-
cess observable by the entrepreneur. Models of this type, where
the random market environment is represented by a Markov
chain but the state of this Markov chain cannot be observed di-
rectly (another process can provide partial information about the
state), are called “partially observed Markov decision process”
(POMDP) [19]–[21]. Our POMDP model sets up a probabilis-
tic relationship between an observable and the hidden market
process via the entrepreneur’s investment portfolio. We then
develop a probabilistic information measure to quantify the
connection between the observable and hidden process in or-
der to better understand the state of the market for the business
opportunity. With the probabilistic information measure, the en-
trepreneur could construct and test novel investment portfolios
that are best suited for the market with respect to her/his risk
preference and operational shortages.
Our contributions in this paper are as follows. Under the
data-driven perspective of entrepreneurship, we develop an al-
gorithmic data analysis technique for entrepreneurs to analyze
imperfect data from external market factors for an investment
portfolio. The optimal investment portfolio balances the trade-
off between expected returns and market information, and is
further moderated by the entrepreneur’s risk preference and
operational shortages. Our numerical analysis shows how a
data-driven investment portfolio can be used to manage the
key tradeoff. We observe that a portfolio of POMDP-infused
investments (e.g., to hedge risk or control market) might bal-
ance the tradeoff better than with one of the investments pre-
dominantly involving full information (FI) or high expected
returns. In practice, a data-driven analysis could enable en-
trepreneurs and managers to customize investment portfolios
based on risk preferences and operational shortages so as to over-
come market uncertainty and make better opportunity-related
decisions.
Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
605
II. LITERATURE REVIEW
Opportunity evaluation is the core of entrepreneurial decision
making. Entrepreneurship scholars have extensively investi-
gated how entrepreneurs make opportunity evaluation decisions
based on individual factors (e.g., cognition and aspirations)
combined with external factors (e.g., valuation of the market)
(see [8], [22] for reviews). McKelvie et al. [2] found that an
increase in uncertainty decreases the entrepreneur’s willingness
to act on an opportunity in the face of uncertain environmental
conditions. Entrepreneurs must manage entrepreneurial risk,
rationality, and high levels of uncertainty about markets when
evaluating opportunities [23]–[25]. Entrepreneurs are also
advised to develop strategies to hedge adverse outcomes
according to their risk preferences [26], [27].
While research on operations management (OM) has studied
the exploitation process of opportunities, subject to operational
shortages of 4Rs (see [15] for a review), OM scholars have not
yet explored “a deeper strategic understanding of evaluations
of a recognized opportunity to determine if it represents an op-
portunity for the specific entrepreneur” [28, p. 1416]. In the
evaluation stage, entrepreneurs operate under significant uncer-
tainty about the true value of an opportunity, and information is
needed to assess that value [29]. Shepherd and Patzelt [30] note
this issue in emerging operational entrepreneurship research,
and call for processes to effectively capture and utilize informa-
tion and improve entrepreneurs’ ability to refine the potential
opportunities and to act upon subsequent potential opportunities.
Prior studies on innovation and entrepreneurship have
mainly recognized a stable and fixed entrepreneurial process for
evaluating a new product/service idea that underlies a market
opportunity (e.g., [31], [32]). With the infusion of data-driven
technologies, the entrepreneurial process has become less
bounded (predefined) by structural boundaries of product scope
and market search and temporal boundaries of entrepreneurial
activities [3], [33]. The data-driven framework of Miller and
Mork [17] constitutes a process for data collection, transfor-
mation, and application of analysis techniques that underlie
insights required for decision making.
Two notable research streams in entrepreneurship, opportu-
nity creation [34] and effectuation [35], [36] perspective of en-
trepreneurship, have studied the continuous re-evaluation and
evolution of opportunity in less bounded entrepreneurial pro-
cesses, as well as the accompanying uncertainty. However, ow-
ing to the complex and interrelated nature of less bounded pro-
cesses, conventional research methods have been of limited use
in capturing the dynamics of this phenomenon [37]. Hence, there
is a gap in the literature on operational and data-driven perspec-
tives of entrepreneurship for novel data-driven techniques and
technologies that shape the entrepreneurial process, including
opportunity evaluation.
Studies on decision making under uncertainty have identi-
fied methods to address situations when the outcome proba-
bilities are not, or cannot, be known precisely [1], [24], [38].
Specifying mathematically tractable risky choice problems de-
velop precise rules for rational decision making [24], [39]. Ex-
pected utility theory provides the backdrop for the normative
 606
status of rational decision making involving risky situations
and risk preferences, as in investment decision making when
outcomes are probabilistic. In contrast, the entrepreneurial
decision maker may not know all possible outcomes of
his/her actions, and does not know the probability distribu-
tion of those outcomes [40]. In such cases, decision mak-
ers may be better off pursuing favorable outcomes utiliz-
ing decision technologies that allow for enhanced information
flow [41].
Our study contributes to the operational and data-driven per-
spectives of entrepreneurship by refining decision making using
a data-driven technique for the opportunity evaluation in a dy-
namic entrepreneurial process. Furthermore, we contribute to
entrepreneurial decision making under uncertainty by identi-
fying the tradeoff between expected returns and market infor-
mation, and providing insights into how data-driven investment
portfolio can be used to manage this tradeoff, while accounting
for the entrepreneur’s risk preference and operational shortages
of the 4Rs. In particular, we consider a Markovian model to
address the changes in business opportunity-related evaluations
in uncertain and complex environments, such as in a hidden
market, where the decision maker cannot directly estimate the
outcome probabilities.
Modeling the highly uncertain nature of the market using
a Markovian model has long been recognized as a reasonable
approach [42], and Markov models have been used primarily
to describe the dynamic behavior of financial markets (e.g.,
[43]). Although POMDPs are used in various fields, includ-
ing financial analysis and engineering (e.g., [44], [45]), they
have not been used extensively in entrepreneurial decision mak-
ing [39], with the notable exception of Lévesque and Maillart
[18]. Lévesque and Maillart [18] analyze a POMDP in an opti-
mal stopping problem, which requires “the consideration of not
only the decision maker’s uncertainty about the viability of the
opportunity, but also the environmental conditions that influ-
ence the context in which the opportunity assessment is made”
[18, p. 279]. They obtain probability thresholds based on the
information cost, the payoff structure, and information quality.
Their results indicate the ideal time for the decision maker to
stop accumulating information and to either accept or reject an
opportunity.
In a similar vein, we analyze a POMDP model in a portfolio
optimization problem that considers the probabilistic relation
between the hidden and observable processes. Our study differs
from that of Lévesque and Maillart [18] in two ways. First,
the entrepreneur decision maker in our study evaluates external
factors of uncertainty(e.g., consumer market and regulation),
while factoring in operational shortages, and then determines
an investment portfolio for market observability. Second, we
examine whether the entrepreneur can exert partial control over
the market state changes in a Markov chain. Hence, our data-
driven technique, a PODMP model, manages “true uncertainty”
cases of high complexity and uncertainty [1], which contributes
to a better understanding of entrepreneurial decision making
under uncertainty.
Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 65, NO. 4, NOVEMBER 2018
III. MODEL
Consider an entrepreneur who is evaluating a business op-
portunity for a market in a multiperiod setting. The state of
the market—whether the market has a positive or negative out-
look for the entrepreneur’s opportunity—depends on a variety
of external market factors such as economic, governmental, so-
cial, and regulatory [46] that may not be directly observable
and may change over time. Furthermore, the entrepreneur’s risk
preference and operational shortages of one or more of 4Rs
(i.e., internal constraints) influence the entrepreneur’s ability to
observe the state of the market and take control over market
changes.
To exemplify the impact of an external factor on opportunity
assessment, we offer a simulated example of an entrepreneurial
venture in a developing industry (e.g., clean energy). The
entrepreneur may not have established know-how of external
market regulations and lobbying practices for the technology
(i.e., shortage of external regulations). Although the new
venture’s investors may provide some policy and regulatory
assistance, the regulations for an emerging technology may
be transient, which is likely to result in a hidden market.
Therefore, he or she might not be able to fully evaluate
the true economic outlook—positive or negative—without
understanding the regulatory conditions, particularly among the
rapidly changing laws surrounding energy. Information about
the state of regulations and policy is needed to evaluate the
valuation of the market for the clean energy innovation through
hiring legal services, lobbying practices and active participation
in discussions about pending regulations [47].
The entrepreneur in our model gathers information about the
market by allocating her/his total funds X across a portfolio of
m investments in f independent external market factors over T -
periods. To maximize potential returns of a business opportunity,
a resource-constrained entrepreneur could invest small amounts
of his/her resources, while minimizing risk exposure [23], [48],
[49]. For example, prior to Turkey’s passage of a renewable
energy law in 2005, most clean energy entrepreneurs made rel-
atively marginal investments to reflect the market’s appetite
for solar. The level of information, as denoted by νkj ∈ [0, 1],
about the market factor j ∈ {1, . . . , f } depends on investment
k ∈ {1, . . . , m}, whose return provides information about fac-
tor j. Subsequently, the investment returns form the state values
of an observable process as characterized by the observed mar-
ket factors.
In our example above, a specific investment that is associ-
ated with the external factor of energy regulations might help
the entrepreneur to gain information about how regulations and
possible future policies might influence the technological op-
portunity’s economic value. In this way, the investment reveals
partial information about the state of the market. Furthermore,
the entrepreneur gains no information (NI) related to a factor
unless she/he specifically invests in a source associated with
that factor, but may obtain similar information about the market
from other factors. Hence, the objective of the entrepreneur is
to choose an investment portfolio U that provides information
ÇANAKOĞLU et al.: HOW DATA-DRIVEN ENTREPRENEUR ANALYZES IMPERFECT INFORMATION
about the state of the market and maximizes returns gained from
the investment at a terminal time T .
Regarding our formal model, we make the following three
assumptions. First, the change in the valuation of investments
is exogenous and independent of funds invested at the previous
periods. Thus, the investment return is a random variable that
depends solely on the given state of the market. Second, we
do not allow for risk free borrowing to avoid unbounded solu-
tions. The omission of risk-free borrowing also underlines the
entrepreneurial decision making amidst shortages of resources,
routines, and reputation with no immediate resource infusion.
Finally, the entrepreneur does not experience any decrease in
efficiency by investing in multiple investments.
In the remainder of this section, we develop a POMDP
model with sufficient statistics to examine imperfect informa-
tion flow. We then formulate a data analysis algorithm for the
entrepreneur’s opportunity evaluation problem with dynamic
programming (DP).
A. POMDP Model
We let Zt denote the state of the market at period t. The
Markov chain Z = {Zt ; t = 0, 1, . . . , } is a hidden process with
the transition matrix
Qt (a, b, Ut ) = P (Zt+1 = b |Zt = a, Ut )
over a state space F = {a, b, c, . . . , s}.1 Ut = (u1t , u2t , . . . , um
t )
denotes the portfolio for m investments at period t, where ukt (≥
0) represents the allocated investment amount for investment k.2
Ut influences the transition states because Qt depends on the
investments of the portfolio. The states of the market are not
observable; thus, the entrepreneur’s portfolio of m investments
generate random net returns
 
Rt (s) = Rt1 (s), Rt2 (s), . . . , Rtm (s) (2)
when the hidden state of the market is s at period t. The en-
trepreneur has a total cash of Xt at the beginning of period
t. We follow the stochastic evolution of the capital process
X = {Xt ; t = 0, 1, 2, . . . , } with a cash dynamics equation:
 
Xt+1 (Ut ) = Rt (s)Ut , denotes the transpose operator.
While Z is a hidden process, the entrepreneur can obtain
imperfect information about the state of the market with an ob-
servable process Y = {Yt ; t = 0, 1, 2, . . .} over a state space
E = {i, j, . . . , z}, where Yt is the information available at pe-
riod t. The observable process Y is not necessarily a Markov
chain and the states of the market Z are dependent on previous
observations of Y. We develop a probabilistic information mea-
sure, referred to as the emission matrix E, that represents the
probabilistic relationship between the true market state and the
observation of external market factors. In particular, the proba-
bilistic evolution of Y depends on the state of Z at period t and
1 It is possible to have a countably infinite number of states in theory. We use a
limited number of states for the observed and hidden market processes because
of the estimation required by the model. In practice, some of these states would
be less likely to occur. Thus, we do not lose any insights from our analysis and
can easily include further states in our modeling framework.
2 Throughout this paper, unless otherwise stated, subscript denotes time and
superscript denotes the index of investment.
Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
607
is independent of all previous states of Z and Y. The emission
matrix at period t, Et , is
Et (i, a, u) = P (Yt = i | Zt = a, Ut ) . (3)
Note that E is characterized by the investments in the
portfolio Ut .
Each investment provides information about specific exter-
nal factors. Specifically, an identity matrix (or any other matrix
that consists of a single 1 in each column) as the emission
matrix, Et = It , represents FI where every observation corre-
sponds to a single state in a one-to-one relationship. On the
other hand, an emission matrix with identical rows for every
state gives NI about the states of the market since the observa-
tion is independent of the underlying Markov model. Any matrix
in between provides partial information. A legitimate emission
matrix should have positive elements where the summation of
elements at each row is equal to 1.
In our example, this emission matrix corresponds to the
probability of acceptance of regulations favorable to the en-
trepreneur’s business. Suppose that the observable process con-
sists of two states of the market for the business opportunity
(i.e., positive and negative outlook) as classified by the regula-
tions. A high probability is more likely (i.e., a positive outlook)
to be observed if favorable legislation is adopted, and less likely
(1)
(i.e., a negative outlook) if it is rejected. In practice, experienced
energy and clean tech attorneys can serve as trusted legal advis-
ers to best estimate these probabilities for specific sectors and
regions. By investing in regulatory and policy making efforts,
the entrepreneur gains information regarding the state of the
market from the energy regulatory perspective. Recall that the
entrepreneur gains NI without any investment.
In another study closely relevant to ours, Canakoglu and
Ozekici [19] examine a multiperiod portfolio-selection problem
and examine a hidden market model (HMM) with imperfect
information flow. Our model differs from their formulation in
two important ways. First, in our POMDP-based model, the
entrepreneur observes not only the Markov market but also
controls it. Second, their formulation focuses on an investor
with no cash reserves constraint who invested in a fixed num-
ber of investments (i.e., the firm’s entire portfolio). Hence, the
modeling structure is predetermined. Our model hinges on the
entrepreneur’s investments, which can be less than available
investment options. In this way, we extend their work on port-
folio optimization in HMMs to a case with a dynamic transition
matrix in a POMDP.
B. Sufficient Statistics
The entrepreneur relies on information from past periods in
our multiperiod model. In particular, since energy regulations
are shaping over time, the entrepreneur may need to invest in
multiple periods. A new observation at period t will increase
the dimension of the information vector with a new observation
Yt , therefore, the probabilistic structure of the information flow
increases the dimension of the problem with each period. To
monitor the increasing dimension, we utilize sufficient statis-
tics, a common approach used under these circumstances, that
 608 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 65, NO. 4, NOVEMBER 2018
represents the probabilistic structure of the information flow
[19], [50].
Let Y¯t = [Y0 , Y1 , Y2 , . . . , Yt ] and Z̄t = [Z0 , Z1 , Z2 , . . . , Zt ]
denote the cumulative observed information and past his-
tory about the states of the market until period t, respec-
tively. The size of Y¯t further increases with each period t;
Ȳt+1 = (Y¯t , Yt+1 ). Furthermore, let Πst = P (Zt = s | Y¯t , Ut )
be the probability that the market states at period t is s, given
previous observations Y¯t , and investment portfolio at this period
Ut . We specify the conditional  distribution of Zt given Y¯t by
1 2
Πt = [Πt , Πt , . . . , Πt ], where s∈F Πst = 1 and Πst ≥ 0 for
s
all s ∈ F.
Let the state of the market at time t + 1 be s . Using the
conditional distribution of the state of the market at period t,
Πt , a new observation on the state of the market at period t + 1,
Yt+1 , and the investment portfolio at period t, Ut , we formulate
the conditional distribution of the state of the market at the
consecutive period t + 1 as follows:
   . 
Πst+1 = P Zt+1 = s | Ȳt+1 , Ut = Gts (Πt , Ut , Yt+1 ) (4)
which we can further formulate with Bayesian updating. Given
that the probability distribution about the state of the market is π
 We now analyze the entrepreneur’s expected returns maxi-
s π s Qt (s, s , u)Et+1 (s , u, j)
Gt (π, u, j) =  s∈F s  
(5)
s,s  ∈F π Qt (s, s , u)Et+1 (s , u, j)
for t ≥ 0. The initial condition Π0 can be determined with an
initial observation Y0 . In practice, the initial conditions can be
estimated via a preliminary analysis of the unobserved market.
Then, the entrepreneur can update initial conditions with the
emission matrix at period t = 0, E0 . In our example, updating
the distribution over the core states of the market in this manner
can be viewed as gathering information about the positive or
negative outlook of the market for the business opportunity.
Πt in (4) summarizes the past history of the state of the market
and provides sufficient statistics for the history of all observa-
tions on the state of the market until period t, Ȳt , depending on
the portfolio of investments at period t, Ut . Π = {Πt ; t ≥ 0} is
deemed to be a Markov chain defined on a continuous space with
a set of probability distributions on F [20]. Consequently, we
model our problem as an observable Markov chain Π, where Πt
is a state of the Markov chain at period t. Equation (5) represents
 from Πt to Πt+1 inthis Markov chain, where
the transformation
Gt (π, u, j) = Gts (π, u, j) ; s ∈ F is a probability distribu-
   function expresses the entrepreneur’s returns in terms of avail-
tion with s∈F Gts (π, u, j) = 1 and Gts (π, u, j) ≥ 0 for all

s ∈ F.
C. DP Model
In our model, the entrepreneur maximizes the expected re-
turns at a terminal time T by selecting the investment portfolio.
To develop the entrepreneur’s dynamic problem formally, we let
gt (π, x, u) denote the returns gained from an investment port-
folio u constrained by available cash funds x in period t and
optimal portfolios u∗ from t + 1 to T . Also, the probability of
the information flow in the market (i.e., sufficient statistics) is
denoted by π. Then we define the optimal expected returns at
Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
period t from choosing the investment portfolio u as
vt (π, x) = max gt (π, x, u). (6)
u
According to the DP principle
gt (π, x, u) = E [vt+1 (π  (ut ), xt+1 (ut ))] (7)
where π  (ut ) is the updated sufficient statistics for the next pe-
riod. The entrepreneur’s optimal expected returns maximization
problem in (6) becomes a DP problem with the following value
function:
vt (π, x) = max E [vt+1 (π  (ut ), xt+1 (ut ))] . (8)
u
This can further be extended using (5) as
vt (π, x) = max E [vt+1 (Gt (π, u, Yt+1 ) , R(Zt )u)] (9)
u
for t = 0, 1, . . . , T − 1. The solution for DP in (9) can be found
with a recursive solution procedure and a complete emission
matrix.
IV. ANALYSIS
A. Optimal Investment Portfolio
mization problem in consideration of the evaluation of a busi-
ness opportunity with the DP solution procedure outlined in the
previous section. To obtain explicit characterizations of the op-
timal investment portfolio, we define an emission matrix and
power utility function. For the emission matrix, we consider the
following matrix for a portfolio of m investments and external
factor f , E f :
m
Ef = dk (1 − ν f ,k )E0f + ν f ,k I (10)
k =1
where d = u /x is the proportion of the total funds x allocated
k k
to investment k. Recall that ν f ,k ∈ [0, 1] is the level of infor-
mation provided by investment k about market factor f . E0f
represents the emission matrix with base level of information
about market factor f , and I is the identity matrix. Investment
k characterizes the emission matrix with the added information
(1 − ν f ,k )E0f + ν f ,k I.
Furthermore, we define the entrepreneur’s individual risk
preference with the power utility function.3 The power utility
able cash investment X based on her/his risk preference with a
relative risk aversion of 1 − γ, formally U (X) = X γ /γ. Note
that the entrepreneur’s returns are endogenous and independent
of the market factors. The power utility function with total cash
investment XT at a terminal time T sets the boundary condi-
tion at vT (π, XT ) = (XT )γ /γ for all possibilities of π and the
firm’s expected returns at E [(XT )γ /γ].
3 The power utility function has the same characteristics with a class of utility
functions, e.g., exponential and logarithmic. This family of utility functions
possess a constant relative risk aversion type under imperfect market conditions.
Thus, the explicit characterizations of the optimal solution inherits similar linear
structure with different utility functions [19].
ÇANAKOĞLU et al.: HOW DATA-DRIVEN ENTREPRENEUR ANALYZES IMPERFECT INFORMATION 609

Proposition 1 characterizes the optimal investment portfolio market, as they may gain similar information from legal counsel-
for the emission matrix in (10) and power utility function. ing on understanding of energy trends. Indeed, if the availability
Proposition 1: Let the entrepreneur’s return function be the of information is not excessively dependent on the amount of
power function (xγ /γ) and the emission matrix be (10). Then, investment, then the entrepreneur can decide whether or not to
the optimal solution to the DP in (9) is invest in a particular factor with its portfolio configuration.
xγ This may not be the case because the entrepreneur can get
vt (π, x) = Ct (π) (11) better or more information by increasing his/her investment
γ
amount. For instance, the entrepreneur could invest in partici-
and the optimal return-maximizing investment portfolio is de- pating in policy making conversations on state, national, or inter-
termined by national levels (e.g., [47]). We thus consider the entrepreneur’s
u∗t (π, x) = dt (π)x (12) optimal portfolio in (12) to be dependent on the amount of
cash allocated to each investment rather than the proportion of
where her portfolio. We replacedk with uk in (10); that is, the new

Ct (π) = πa Qt (a, b, dt (π)) E(j, b, dt (π))Ct+1 emission matrix is E f = m k =1 u ((1 − ν
k f ,k
)E0f + ν f ,k I). Al-
a∈F b∈F j ∈E
though this problem does not lead to a closed-form solution, it

   γ  is possible to analyze the backward dynamic equation where the
× Gts (π, dt (π) , j) · E Rt (b) dn (π) value function is reflected by the state variable π and the total
(13) cash x numerically.
In addition, a numerical analysis may be appropriate due to
and dt (π) satisfies for all t = 0, 1, . . . , T − 2 the “three curses of dimensionality” in the above DP problem.
First, a continuous-state space consists of all possible distribu-
πa Qt (a, b, dt (π)) E (j, b, dt (π))Ct+1
tions of π. Second, the actions including all possible investment
a∈F b∈F j ∈E

  strategies may consist of continuous weight distribution of cur-
 γ −1

× Gts (π, dt (π) , j) · E Rtk (b) Rt (b) dt (π)

=0 rent funds. Accordingly, the third curse, the size of the state
space and the action space inflate the size of the transition func-
(14) tion. Therefore, we develop an approximation algorithm that
and for t = T − 1 and all k = 1, 2, . . . , m mimics the DP solution procedure in Proposition 1 with the new
emission matrix in discrete states. This algorithm is outlined in
πa QT −1 (a, b, dT −1 (π)) E (j, b, dT −1 (π)) Algorithm 1 in Appendix A. We utilize Algorithm 1 to inves-
a∈F b∈F j ∈E tigate the optimal investment portfolio with a comprehensive
  γ −1
numerical analysis in the next section.
× E RTk −1 (b) RT −1 (b) dT −1 (π) =0
m
where k =1 dkt (π) = 1, with boundary condition
CT −1 (π) = πa QT −1 (a, b, dT −1 (π))
a∈F b∈F

× E (j, b, dT −1 (π))E RT −1 (b) dT −1 (π)
The proof of Proposition 1 is in Appendix A. Proposition 1
shows that the optimal proportion of the portfolio dt (π) is
independent of the current cash level but dependent on the
entrepreneur’s risk aversion level in (12). Hence, the optimal
portfolio configuration depends more on the entrepreneur’s risk
preference than the level of remaining cash. Our result holds
when the emission matrix in (10) is determined by the propor-
tion of the funds (i.e., regardless of the amount) allocated to
investments.
The closed form in Proposition 1 may be applicable in settings
where the entrepreneur and her/his competitors have a similar
configuration of financial resources or routines to access to in-
formation. For example, in innovation contests, all participants
are considered equally capable when the performance of a so-
lution is uncertain for the solver (i.e., what constitutes a good
solution has not yet been established) [51]. In our example, all
participants in the clean energy market may obtain similar emis-
sion matrices on the probability of the positive outlook of the
Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
(15)
B. Numerical Analysis
Our numerical analysis characterizes the optimal investment
j ∈E portfolio for a return-maximizing entrepreneur after T = 5 peri-
γ  ods based on expected returns and information about market ob-
.
servability and control. We particularly focus on an entrepreneur
(16)
faced with operational shortages of financial resources (“funds”
hereafter) and external regulatory knowledge (“regulations”
hereafter). The shortage of funds is the first and foremost re-
source limitation in creating a business [52]. For the shortage of
regulations, the entrepreneur is likely to have insufficient knowl-
edge of the external regulations and lobbying practices due to
his/her potential lack of understanding of the regulatory frame-
works (e.g., [47]), or the regulations for an emerging technology
may be transient [9], which may in turn result in a hidden mar-
ket. For instance, for clean energy startups in Turkey, financial
resources and regulations have been two most leading factors in
developing the renewable energy market [13].
Besides that, the entrepreneur is faced with the shortages of
funds and regulations, according to Ernst and Young’s 2013 G20
Entrepreneurship Barometer, the fiscal and regulatory system is
one of the five main categories of potential external barriers to
pursuing a business opportunity. In addition, an entrepreneur
would be reluctant to invest in an opportunity if the returns,
which depend on the final demand patterns, are unknown or
 610
unclear [53]. In line with Smith et al. [53], we add the pre-
dictability of consumer market. We therefore examine the re-
turns of the investment portfolio with respect to two external fac-
tors (regulations and consumer market) in evaluating the market
state.
We define an observable process Y associated with two inde-
pendent external factors: 1) regulatory environment; (i.e., factor
1; f = 1); and 2) consumer market (i.e., factor 2; f = 2).
The observable process Y depends on the state of the mar-
ket s = {P, N }–positive outlook (P) or negative outlook (N)–
with respect to each external factor. Y then forms four pos-
sible states of the hidden market process Z, {P P, P N, N P,
N N }. The first (second) letter represents the state of the mar-
ket with respect to information about the regulations (consumer
market). In our example, the positive outlook with the regu-
latory environment (consumer market) may represent the en-
trepreneur’s relative acceptance of pending regulations (high
adoption rate) with the entrepreneur’s technology. We set the
transition matrices of two independent state variables for regu-
latory environment and consumer market, respectively, as
   
0.7 0.3 0.8 0.2
Q1 = , Q2 = .
0.4 0.6 0.3 0.7
This results in a combined transition matrix
⎡ ⎤
0.56 0.14 0.24 0.06
⎢ 0.21 0.49 0.09 0.21 ⎥
⎢ ⎥
Q= ⎢ ⎥.
⎣ 0.32 0.08 0.48 0.12 ⎦
0.12 0.28 0.18 0.42
Accordingly, we compute the steady-state probabilities for  the
hidden market process Z, as 0.34 0.23 0.26 0.17 . We
tested and confirmed similar results with a wider range of pa-
rameter values. Herein, we present a representative set of our
results.
Throughout our numerical analysis, we consider three lev-
els of the entrepreneur’s initial cash, X0 = {0.1, 0.5, 1}—
low, medium, and high—and identify cash levels with a cash-
strapped (i.e., low cash level), limited-funded (i.e., medium
cash level) and well-funded (i.e., high cash level) entrepreneur,
respectively. We set the investment portfolio size to m =
{2, 4, 6}. In addition to the shortages of funds and regula-
tions, we consider the entrepreneur’s relative risk aversion of
1 − γ. The entrepreneur can be classified as risk neutral when
γ = 1, and risk averse when γ < 1. We examine a range of val-
ues for risk aversion (γ ∈ [−1.5, 1.5]), which covers a sufficient
experimental range of risk preferences [54].
The range of values for X0 , m and γ allows us to conduct a
comprehensive examination of the entrepreneur’s optimal port-
folio with a sensitivity analysis. We represent the state variable
as π = {π1 , π2 , x}, where πf is the probability of a positive
outlook on external factor f and x is the level of cash on hand.
To control the problem’s dimensionality, we limit the size of the
market state space by computing and storing the values of π in
multiples of 0.1, and we set 10 possible values for x. Therefore,
our numerical computation has a state space with 1210 separate
market states. The initial probability distribution is equal to the
steady-state distribution for the two market factors.
Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 65, NO. 4, NOVEMBER 2018
In our numerical study, we define investment k ∈ [1, 6] in
the portfolio with expected returns of E[Rk ]. Given the steady-
state probabilities, Table I summarizes returns based on state s,
E[Rk | s], and the level of the information νkf gained from in-
vestment kabout market factor f . Each investment has different
characteristics regarding information about market observabil-
ity and control. Investment 1 (E[R1 ] = 12%) represents a NI
model, and is a risk-free investment where no extra informa-
tion is obtained; basically, the entrepreneur observes nothing
about the market. Investment 2 (E[R2 ] = 7%) represents the
FI model where the entrepreneur gains FI through the emission
matrix and observes the state of the market. A relatively lower
expected return on Investment 2 with FI is more likely to occur
due to the amassed, costly, imperfect market information [18].
The next two investments (k = 3, 4) represent a POMDP,
where partial information is obtained about one of the two exter-
nal market factors. Investment 3 in the portfolio (E[R3 ] = 10%)
provides the partial information about the state of the regula-
tory environment and is independent of the consumer market.
Similarly, Investment 4 (E[R4 ] = 11%) presents information
only about the consumer market. We further define two specific
POMDP-infused investments, Investment 5 (E[R5 ] = 8%) and
Investment 6 (E[R6 ] = 9%). Investments 5 and 6 serve spe-
cific purposes, risk hedging and market growing, respectively.
Investment 5 presents a hedging characteristic in the presence
of increasing risk because it provides an option, that has higher
returns if both factors are negative; E[R5 | s = N N ] = 1.4. In-
vestment 6, on the other hand, increases the probability that
the consumer market will have a positive outlook: E[R6 | s =
P P ] = 1.4 and E[R6 | s = N P ] = 1.2.
We consider two types of investments for Investment 6: with
and without the power to control and grow the market. In
other words, Investment 6 without market control is designed
with the transition matrix Q2 . To examine the entrepreneur’s
control over the market, we allow market-shaping Investment
6 to alter the transition matrix Q2 . The transition matrix is
6
updated as Q(u) = min(1, u6 )Q2 + max(1, 1 − u6 )Q2 , where
0.84 0.16
Q62 = . Technically, if u6 = 1, then the limiting
0.44 0.56
probability of factor 2 is positive, increasing the positive outlook
of the market from 60% to 73%. Otherwise, this investment does
not influence Q2 . We later utilize these investments to overcome
risks associated with the shortages of funds and regulations.
Next, we compute the optimal portfolio configuration with
Algorithm 1. We derived the numerical solutions of equations
using MATLAB. Fig. 1 explores the optimal investment portfo-
lio under three levels of initial cash; high, medium, and low for
different available investments (m = 2, 4, 6).
1) Base Scenario: We start our numerical analysis with a
base scenario [see Fig. 1(a)] that focuses on the portfolio with
two investments (m = 2): Investment 1 with NI and Invest-
ment 2 with FI. Fig. 1(a) represents higher expected returns
with Investment 1 in the portfolio and shows that expected re-
turns is a greater concern when the entrepreneur has low risk
aversion. Furthermore, when X0 is high, the well-funded en-
trepreneur with low risk aversion considers Investment 2 in
his/her portfolio to obtain FI. However, this is not the case
as the shortage of funds becomes significant (i.e., X0 is re-
ÇANAKOĞLU et al.: HOW DATA-DRIVEN ENTREPRENEUR ANALYZES IMPERFECT INFORMATION
TABLE I
EXPECTED RETURNS AND LEVELS OF INFORMATION ABOUT EXTERNAL MARKET FACTORS (s = 4; f = 2; m = 6)
duced). The limited-funded or cash-strapped entrepreneur with
low risk aversion could generate higher returns with Investment
1, reducing the proportion of Investment 2 in the portfolio, be-
cause the information gain is dependent on the investment dollar
amount.
Hence, the shortage of funds could be a significant driver of
the investment portfolio in a hidden market, but its impact on
the portfolio is reduced with increased risk exposure. Fig. 1(a)
illustrates that when the entrepreneur becomes more risk averse,
the proportion of FI Investment 2 increases for all three types of
entrepreneurs because the lack of market information becomes a
greater concern than returns for that entrepreneur. Furthermore,
that increase is greater for well-funded entrepreneurs with low
shortage of funds.4 Hence, the base scenario represents the trade-
off between expected returns and market information: while the
shortage of funds motivates the entrepreneur to pursue high ex-
pected returns, and the shortage of external regulations (i.e.,
increased risk exposure due to lack of information) motivates
the entrepreneur with high risk aversion to choose an investment
portfolio with high level of market information.
2) Market Observability: In the base scenario, we found
numerical evidence demonstrating the tradeoff between ex-
pected returns and market information, while considering the
entrepreneur’s cash level and risk preference. This insight is
also valid with a larger portfolio of four investments (m = 4) in-
cluding POMDP-infused Investments 3 and 4. Each investment
facilitates the flow of relatively high level of information about
a single and different market factor, while generating returns.
Fig. 1(b) (left figure) shows that the well-funded entrepreneur
would consider information over return. As the shortage of funds
becomes significant, the limited-funded entrepreneur with low
risk aversion considers POMDP-infused Investments 3 and 4,
which generate some returns as well as information, as dis-
played in the center of Fig. 1(b). Interestingly, the cash-strapped
entrepreneur who focuses on generating financial returns more
than information invests a portion of his/her very little cash in
Investment 3, as illustrated by the portfolio for low-initial cash
in Fig. 1(b) (right). Furthermore, each entrepreneur with dif-
4 For a practical validation that better-funded entrepreneurs would spend more
when faced with the shortage of regulations, consider the report by the Center
for Responsive Politics, which compiled lobbying expenditures from the Sen-
ate Office of Public Records up until October 2016. Facebook, Alphabet, and
Amazon are top three spenders on lobbying significantly more than any close
competitors.
Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
611
ferent level of cash choose information generating investment
(e.g., Investment 2) rather than expected returns with higher risk
aversion, shown in Fig. 1(b).
These numerical examples in Fig. 1(a) and (b) send an in-
teresting message to the entrepreneur: rather than pursuing
highest expected returns, an entrepreneur can choose an in-
vestment portfolio based on his/her cash level and risk pref-
erence to gain information about the hidden market. Fig. 1(c)
shows further support for these findings with a portfolio of
six investments (m = 6), including risk hedging Investment
5. In particular, while the well-funded entrepreneur with rel-
atively low risk aversion continues pursuing FI (e.g., Invest-
ment 2), the limited-funded or cash-strapped entrepreneur with
similar risk aversion chooses investments that balance the ex-
pected returns and information (e.g., Investments 3 and 4), as
shown in Fig. 1(c). Although the cash-strapped entrepreneur
does not include Investment 2 in his/her portfolio, she/he al-
locates some funds to generating some information via other
investments.
Hence, our numerical analysis shows how the entrepreneur’s
risk aversion and shortages may impact the investment portfo-
lio. Hence, the entrepreneur may benefit from constructing and
testing portfolios customized to his/her internal contraints. As
Fig. 1(c) illustrates, the entrepreneur considers the impact of an
investment (i.e., Investment 5) specifically designed to handle
risk hedging on the entrepreneur’s portfolio. Entrepreneurs with
relatively high risk aversion and different cash levels, all hedge,
via Investment 5, up to 50% of their total investment. In this
vein, our findings suggest that the POMDP-infused investments
in a portfolio might help the entrepreneur faced with the oper-
ational shortages to evaluate a business opportunity better than
with a portfolio of investments predominantly involving FI or
high expected returns.
It is important to note that our numerical results are based
on how we define expected returns of each investment in our
numerical analysis. For instance, the proportion of an FI Invest-
ment 2 will not decrease if this investment generates higher re-
turns. In effect, we tested the portfolio with six investments with
E[R2 ] = 11% and ostensibly observed a significant increase in
the amount of funds allocated to Investment 2. However, this ob-
servation did not change our findings about the tradeoff between
expected returns and level of gathered information because such
a portfolio ensures higher expected returns and maximizes the
entrepreneur’s returns.
 612 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 65, NO. 4, NOVEMBER 2018

Fig. 1. Optimal investment portfolio with respect to risk aversion 1 − γ for three levels of initial cash: high (Left, X 0 = 1), medium (Center, X 0 = 0.5), and
low (Right, X 0 = 0.1). (a) m = 2. (b) m = 4. (c) m = 6. (d) m = 6, market control.

3) Market Control: Thus far, we have assumed no con-
trol over the market state changes. However, it is very likely
that the entrepreneur could shape some external factors, e.g.,
with reputation, lobbying, and marketing policies. As such,
the entrepreneur might be able to control the marketplace
through direct investments in the consumer market. To en-
able such control, we include an investment that can at-
tract new consumers and grow the consumer market. Techni-
cally, we let Investment 6 improve the likelihood of a pos-
itive outlook in transition matrix Q2 with investment. As
shown in Fig. 1(c), Investment 6 was significantly absent
from the portfolios when market control was not consid-
ered. We now note an important difference between Fig. 1(c)
and (d).

Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
ÇANAKOĞLU et al.: HOW DATA-DRIVEN ENTREPRENEUR ANALYZES IMPERFECT INFORMATION
Fig. 2. Entrepreneur’s portfolio at different risk aversion and cash levels.
The low risk-averse, well-funded, and limited-funded en-
trepreneurs move from choosing the FI Investment 2 to market-
controlling Investment 6. Thus, the availability of market control
could benefit the entrepreneur with relatively minimal opera-
tional shortages more so than gaining FI or high expected re-
turns. Nevertheless, when concern over operational, shortages
becomes significant due to reduced funds and increased risk
aversion, the entrepreneur, regardless of cash level, will be less
interested in controlling the market. In particular, with increased
risk aversion, the proportion of Investment 6 first diminishes and
then disappears from the entrepreneur’s portfolio, as shown in
Fig. 1(d). The cash-strapped entrepreneur who operates with
significant shortages excludes market controlling Investment 6
from his/her portfolio completely, as shown in Fig. 1(d) (right).
4) Summary of Findings: Fig. 2 summarizes our findings
and shows how the individual risk preference and operational
shortages of funds and regulations affect the entrepreneur’s in-
vestment portfolio, which in turn, affects his/her ability to eval-
uate an opportunity. Overall, our numerical results suggest that
the value of high expected returns or FI may be secondary to the
availability of investments aimed to hedge risk and control the
market. That is, the entrepreneur can reduce his/her exposure to
uncertainty by changing his/her investment portfolio.
Relevant to our results, Tanrisever et al. [55] show that debt-
financed startups can also behave conservatively and respond to
the business risks by creating an operational hedge with process
investments rather than investing in production quantity. Our
numerical analysis finds, as shown in the “risk hedging” region
in Fig. 2, that the risk-averse entrepreneur with an operational
shortage (e.g., of external regulations) may choose risk-hedging
investments for his/her portfolio to balance the business risk
while growing a business opportunity.
Fig. 2 also shows with the “expected returns” region that the
low risk-averse entrepreneur, in particular, one with a shortage
of funds (i.e., low cash level), considers investments with high
expected returns because the accumulated information is de-
pendent on the amount of investment. The shortage of funds
limits that entrepreneur’s gain from picking investments with
information return. More interestingly, when the impact of the
Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
613
shortage of funds is mitigated (e.g., with a high level of cash),
the low risk-averse entrepreneur invests in ways to control the
market. Unless that type of investment exists, that entrepreneur
chooses investments with a high level of information gain over
high expected returns. Hence, a more effective way of dealing
with operational shortages and market uncertainty could be to
develop algorithmic models, such as the POMDP that analyze
imperfect data generated through investments and based on dif-
ferent outcomes of the Markovian model.
V. CONCLUSION
The entrepreneurial environment is characterized by high lev-
els of uncertainty about the markets that entrepreneurs wish to
enter [25]. We develop a dynamic data analysis technique based
on a POMDP model to answer our research question about how
to analyze imperfect market data for business opportunity eval-
uation, while accounting for the entrepreneur’s individual risk
preference and operational shortages. Specifically, we obtain a
probabilistic information measure in the form of an emission
matrix. That measure enables insights from an observable pro-
cess related to external factors, which, in turn, helps assess the
state of the hidden market. Owing to Markovian modulation of
the POMDP model, the findings of our dynamic model are more
realistic than standard static models.
Whereas one can derive a closed-form solution for certain
probabilistic measures using a POMDP, closed-form analytical
expressions cannot be obtained for certain cases, such as situa-
tions where the investment dollar amount determines the level of
information gain. Therefore, our algorithm numerically mimics
the POMDP-based model. We offer insights from our numerical
analysis in response to our research question on the impact of
the entrepreneur’s risk preference and operational shortages on
the data-driven investment portfolio. Rather than pursuing the
highest expected returns, an entrepreneur may choose perfect
information, risk hedging, or market controlling investments,
based on his/her cash level and risk preference, in order to max-
imize the venture’s prospects.
Thus, the value of high expected returns or perfect informa-
tion may be secondary to the availability of investments aimed
at hedging risk and/or controlling the markets. For instance, in
our numerical analysis, we show that when the shortage of reg-
ulations or the entrepreneur’s risk aversion creates a higher risk
exposure, he/she may be more successful making a risk-hedging
investment decision to mitigate that exposure. This outcome of
risk hedging may be somewhat counter-intuitive to previous
findings, which argue that an entrepreneur is more likely to in-
vest in activities that generate an FI and a path for opportunity
change [30]. Akin to the data-driven perspective of entrepreneur-
ship, the data analysis tied to an investment strategy can lead to
new or revised activities that further reduce uncertainty.
In particular, our research shows that the data-driven decision
maker might be more successful by managing an investment
portfolio specifically tailored to an observable Markovian mar-
ket, and then analyzing real-time and imperfect data before
making a decision. As such, the entrepreneur’s creativity, fueled
by the availability of data and analytical models, can play a
 614 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 65, NO. 4, NOVEMBER 2018

crucial role in mitigating market uncertainty. We next describe finding in [41] that, in environments with high uncertainty, de-
the theoretical and managerial implications of our study, as well cision makers may fare better by utilizing decision technologies
as future research directions related to our modeling assump- that enable better information flow. Recent infusion of data anal-
tions and limitations. ysis techniques and technologies have provided entrepreneurs
with information for decision making in the entrepreneurial pro-
A. Theoretical Implications cess [3], [30]. In this vein, our POMDP-based model presents
an algorithmic data analysis technique for entrepreneurs and
Our research contributes to studies on the interface of the
managers to utilize real-time data and make better business
OM and entrepreneurship involving decision making under
opportunity-related evaluations and decisions.
uncertainty for business opportunity evaluation for operational
Second, using the proposed data analysis technique, en-
and data-driven entrepreneurship. In cases of statistical prob-
trepreneurs can factor in their risk preferences and operational
ability, we can rely on the fact that the observed probabilities
shortages, and construct and evaluate business opportunities
approach the true probabilities (i.e., the probabilities inherent
with a diversified investment portfolio. Such a diversified port-
in the case) with the sample size. Accordingly, the expected
folio could better position the entrepreneur against potential
utility theory can provide an evaluation of risky options relative
outcomes of the uncertain market. For example, Solazyme, a
to financial resources. However, this is not possible for true
clean energy startup that was operating under a shortage of es-
uncertainty faced by entrepreneurs [1] (e.g., within hidden
tablished public policy, diversified its investments to explore
markets), because the perceived states preclude estimating
the various possible outcomes of the evolving regulatory mar-
the basis of empirical data. In the absence of such statistical
ket, and thus, was able to change successfully [11]. This result is
input, the entrepreneur may gather any information iteratively
also consistent with the findings of Erzurumlu et al. [56] in the
[34], [35], and the entrepreneurial process may become less
context of clean-tech startups constrained by operational short-
bounded in terms of structural and temporal boundaries of the
ages. Their study of government-funded high-risk/high-reward
entrepreneurial activities [3].
projects shows that startups, which evaluated a wide range of
With the data-driven perspective of entrepreneurship, we ex-
diverse markets, mitigated investment uncertainty and received
plore data collection and analysis within a less bounded en-
higher valuations from the U.S. Department of Energy.
trepreneurial process in an environment of true uncertainty. In
Finally, the data-infused portfolio selection with our POMDP
consequence, new methods need to follow new theory [37].
model could inform the entrepreneur on how to best use his/her
We contribute to the entrepreneurial decision-making litera-
resources and reduce the adverse effect of shortages. Janney
ture [8] by developing a data analysis model that examines
and Folta [57] point out that “a portfolio approach can dissipate
real-time and imperfect data in a high-uncertainty environment,
limited financial resources much more quickly than a single
where the business opportunity-market fit is hidden. In particu-
investment, [and, in turn, impair] an entrepreneur’s ability to
lar, a DP formulation incorporates changes in the entrepreneur’s
survive... A better assessment of risk is not dependent on the
business opportunity-related evaluations and investment port-
amount of different projects a firm has underway, but the amount
folio decisions over time. Hence, our study contributes to the
of resources available to see those products through to a state of
entrepreneurial decision making and data-driven entrepreneur-
self-financing.” [57, p. 394]. In particular, a dynamic POMDP
ship by incorporating a dynamic entrepreneurial process into
model could enable the entrepreneur to utilize his/her financial
opportunity-related evaluations under uncertainty.
resources with an investment portfolio by analyzing data itera-
OM scholars have extensively studied efficient exploitation
tively, and therefore, better evaluate the business opportunity in
of opportunities in terms of operational shortages [15], but have
the market.
not sufficiently addressed the evaluation of a recognized oppor-
tunity in entrepreneurial settings [28]. Our research contributes
to the operational entrepreneurship by examining a dynamic en- C. Study Limitations and Future Work
trepreneurial process for the business opportunity evaluation and
Although our model has enabled us to examine a new method
characterizing not only external factors (e.g., regulations), but
of evaluating a hidden market process, several assumptions,
also factoring in internal constraints (e.g., risk preferences and
limitations, and related extensions to this research need to be
operational shortages). The presence of operational shortages
acknowledged. First, our key assumptions pose inherent lim-
extends the research study regarding the entrepreneur’s operat-
itations on our model. For example, although our assumption
ing framework of the 4Rs (e.g., see [14], [15] for reviews) by
about an exogenous and independent change in the valuation
showing how the entrepreneur’s risk preference and operational
of investments does not necessarily change our insights, the
shortages influence the tradeoff between expected returns and
relaxation of this assumption could lead to more profound in-
the level of market information. Our study of the key tradeoff
sights into the market. Second, a DP allows for nonlinearity,
has led us to better understand the impact of the operational
path-dependence, and unpredictability. These properties are im-
shortages of the 4Rs in entrepreneurial settings.
portant, assuming that a Markov model is a close representation
of decision-making in real-world entrepreneurial contexts.
B. Managerial Implications
Third, we did not account for dependence between market
From a managerial perspective, our research offers three factors (i.e., spill-over effects within the entrepreneur’s accumu-
key implications. First, our results relate to March’s previous lated information) and selection bias for market factors. Nor did

Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
ÇANAKOĞLU et al.: HOW DATA-DRIVEN ENTREPRENEUR ANALYZES IMPERFECT INFORMATION 615

we check the validity of the received information, which raises = πa Qt (a, b) E (j, b, dt (π))Ct+1
the question of how an entrepreneur can ensure that he/she is in- a∈F b∈F j ∈E
putting the right information. It would be fruitful for researchers
   
to examine both the selection and validation of market factors × Gts (π, dt (π) , j) E (Rt (b) u)γ /γ .
(e.g., financial) that we investigated, as well as factors that we
overlooked (e.g., political and regulatory). Finally, it would be Let u∗t (π, x) be the optimal policy, such that
helpful to examine how entrepreneurs adapt to market realities vt (π, x) = max gt (π, x, u) = gt (π, x, u∗ ) .
while their internal processes and technologies evolve. These ar- u

eas, if explored, could provide important insights for the fields Taking the gradient of gt (π, x, u) with respect to u, the opti-
of strategy, OM, and entrepreneurship. mality condition is
∂gt−1
APPENDIX k
= πa Qt (a, b) E (j, b, dt (π))Ct+1
∂u
a∈F b∈F j ∈E
A. Proof of Proposition 1
   
Proof: To show that the recursion is true by induction, con- × Gts (π, dt (π) , j) E Rtk (b) (Rt (b) ut (π, x))γ −1 = 0
sider the boundary condition U (xT ) = (xT )γ /γ. Note that the m k∗
return function at time T − 1 is
subject to k =1 ut = x for all k = 1, 2, . . . , m. Since
∗
  dt (π, x) = ut (π, x) /x, the optimality condition becomes

 
gT −1 (π, x, u) = E π a
Qt (a, b)U (RT −1 (b) u) πa Qt (a, b) E (j, b, dt (π))Ct+1 Gts (π, dt (π) , j)
a∈F b∈F   a∈F b∈F j ∈E
= πa Qt (a, b)E (RT −1 (b) u)γ /γ .  
a∈F b∈F
× E Rtk (b) (Rt (b) dt (π, x))γ −1 = 0

For the optimal investments u∗ = (u∗1 , u∗2 , . . . , u∗m ), vT −1 (π, x) subject to m k =1 dt (π, x) = 1, for all k = 1, 2, . . . , m. Since
k

= maxu gT −1 (π, x, u) = gT −1 (π, x, u∗ ) . The convexity of there is no dependence on x in the optimality condition, dt (π, x)
this function is shown in [19]. does not depend on x and dt (π, x) = dt (π) satisfies (14). When
To find the optimal portfolio, we set the gradient of the return the value function at time t − 1 is rewritten for the optimal
function to zero policy, we obtain
∂gT −1 (x)γ
= πa Qt (a, b)E vt (π, i, x) = Ct−1 (π) .
∂uk γ
a∈F b∈F
 γ −1
× RT −1,k (b) RT −1 (b) u∗ (π, x) =0 This completes the proof. 
 k∗
for all k = 1, 2, . . . , m, where m k =1 ut = x. When we sub- B. Algorithm for DP
stitute dT −1 (π, x) = uT −1 (π, x) /x, we obtain u∗T −1 (π, x) =
∗

dT −1 (π, x) x. The optimality condition can be rewritten as

Algorithm 1: Dynamic Programming.
∂gT −1
= πa Qt (a, b)E Initialization
∂uk Initialize the set of aggregate states Π, and wealth

a∈F b∈F
 γ −1
× RT −1,k (b) RT −1 (b) dT −1 (π, x) =0 levels X
Set t = T − 1
 while t ≥ 0 do
for all k = 1, 2, . . . , m, where mk =1 dT −1 = 1. Since there is
k

no dependence on x in the gradient function, and the summation for x ∈ X do

constraint over dT −1 (π, x) does not depend on x, we conclude for π ∈ Π find optimal policy u∗T −1 (π, x) using
that dT −1 (π, x) = dT −1 (π) satisfies (15). Writing the value (9) do
function at time T − 1 for the optimal policy, we obtain calculate vt (π, x) for optimal policy u∗T −1 (π, x)
using (9)
vT −1 (π, x) = xγ πa Qt (a, b) Set t ← t − 1
 a∈F b∈F   end for
RT −1 (b) dT −1 (π)
γ
×E /γ = CT −1 (π)xγ /γ. end for
end while
To prove by induction, suppose first that the hypothesis holds
for periods t + 1, . . . , T − 2, T − 1, T. Then, at time t

ACKNOWLEDGMENT
gt (π, x, u) = E πa Qt (a, b) E (j, b, dt (π))vn
a∈F b∈F j ∈E The authors would like to thank special issue editors and two

anonymous reviewers for their comments. The second author
× (Gt (π, dt (π) , j) , Rt (b) u) wishes to thank S. Fixson, V. Seidel, and Z. Zacharakis for their
comments on the earlier version of this paper.

Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
 616 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 65, NO. 4, NOVEMBER 2018
REFERENCES
[1] F. H. Knight, Risk, Uncertainty and Profit. New York, NY, USA: Houghton
Mifflin, 1921.
[2] A. McKelvie, J. M. Haynie, and V. Gustavsson, “Unpacking the uncer-
tainty construct: Implications for entrepreneurial action,” J. Bus. Ventur-
ing, vol. 26, pp. 273–292, 2011.
[3] S. Nambisan, “Digital entrepreneurship: Toward a digital technology per-
spective of entrepreneurship,” Entrepreneurship Theory Pract., vol. 41,
no. 6, pp. 1029–1055, 2017.
[4] Y. Yoo, O. Henfridsson, and K. Lyytinen, “The new organizing logic of
digital innovation: An agenda for information systems research,” Inf. Syst.
Res., vol. 21, no. 4, pp. 724–735, 2010.
[5] G. Manogaran, D. Lopez, C. Thota, K. M. Abbas, S. Pyne, and R.
Sundarasekar, Big Data Analytics in Healthcare Internet of Things. New
York, NY, USA: Springer, 2017.
[6] A. Davis, “Venture capital firms use big data,” Wall Street J., Apr. 25,
2017.
[7] V. Wu, “A machine-learning approach to venture capital,” McKinsey
Quart., Jun. 27, 2017.
[8] D. A. Shepherd, T. A. Williams, and H. Patzelt, “Thinking about en-
trepreneurial decision making: Review and research agenda,” J. Manage.,
vol. 41, no. 1, pp. 11–46, 2015.
[9] S. S. Erzurumlu and Y. O. Erzurumlu, “Development and deployment
drivers of clean technology innovations,” J. High Technol. Manage. Res.,
vol. 24, no. 2, pp. 100–108, 2013.
[10] “Data-driven innovation for growth and well-being,” OECD, Paris, France,
Interim Synthesis Rep., p. 10, 2014.
[11] K. Fehrenbacher, “Solazyme ditches biofuels (& name) in a world of
cheap oil,” Fortune, Mar. 16, 2016.
[12] R. J. Elliott, L. Aggoun, and J. B. Moore, Hidden Markov Models: Esti-
mation and Control. New York, NY, USA: Springer, 1994.
[13] B. G. Baykan, “Yenilebilir enerji haberlerinde regulasyon ve fi-
nansman one cikiyor,” Retrieved on May 17, 2018. [Online]. Availab-
le: http://www.sirtcantam.com.tr/yenilenebilir-enerji-haberlerinde-regulas
yon-ve-finansman-one-cikiyor/
[14] N. Joglekar and M. Levesque, “The role of operations management across
the entrepreneurial value chain,” Prod. Oper. Manage., vol. 22, no. 6,
pp. 1321–1335, 2013.
[15] N. Joglekar, M. Levesque, and S. S. Erzurumlu, “Business startup oper-
ations,” in Business Startup Operations. Evanston, IL, USA: Routledge,
2017, pp. 255–275.
[16] J. Mullins and D. Forlani, “Missing the boat or sinking the boat: A study
of new venture decision making,” J. Bus. Venturing, vol. 20, pp. 47–69,
2005.
[17] H. G. Miller and P. Mork, “From data to decisions: A value chain for big
data,” IT Prof., vol. 15, no. 1, pp. 57–59, 2013.
[18] M. Levesque and L. M. Maillart, “Business opportunity assessment with
costly, imperfect information,” IEEE Trans. Eng. Manage., vol. 55, no. 2,
pp. 279–291, May 2008.
[19] E. Canakoglu and S. Ozekici, “Portfolio selection with imperfect informa-
tion: A hidden Markov model,” Appl. Stochastic Models Bus. Ind., vol. 27,
pp. 95–114, 2011.
[20] G. E. Monahan, “A survey of partially observable Markov decision pro-
cesses: Theory, models, and algorithms,” Manage. Sci., vol. 28, pp. 1–16,
1982.
[21] L. R. Rabiner and B. H. Juang, “An introduction to hidden Markov mod-
els,” IEEE ASSP Mag., vol. 3, no. 1, pp. 4–16, Jan. 1986.
[22] J. C. Short, J. J. Ketchen, C. L. Shook, and R. D. Ireland, “The concept
of “opportunity” in entrepreneurship research: Past accomplishments and
future challenges,” J. Manage., vol. 36, no. 1, pp. 40–65, 2010.
[23] T. E. Brown, P. Davidsson, and J. Wiklund, “An operationalization of
stevensons conceptualization of entrepreneurship as opportunity-based
firm behavior,” Strategic Manage. J., vol. 22, no. 10, pp. 953–968,
2001.
[24] K. D. Miller, “Risk and rationality in entrepreneurial processes,” Strategic
Entrepreneurship J., vol. 1, pp. 57–74, 2007.
[25] B. Wu and A. M. Knott, “Entrepreneurial risk and market entry,” Manage.
Sci., vol. 52, no. 9, pp. 1315–1330, 2006.
[26] L. W. Busenitz, “Entrepreneurial risk and strategic decision making: It’s
a matter of perspective,” J. Appl. Behav. Sci., vol. 35, no. 3, pp. 325–340,
1999.
[27] D. Forlani and J. W. Mullins, “Perceived risks and choices in en-
trepreneurs’ new venture decisions,” J. Bus. Venturing, vol. 15, pp. 305–
322, 2000.
Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
[28] D. A. Shepherd and H. Patzelt, “Operational entrepreneurship: How oper-
ations management can advance entrepreneurship,” Prod. Oper. Manage.,
vol. 22, no. 6, pp. 1416–1422, 2013.
[29] C. Gaimon and J. Bailey, “Knowledge management for the entrepreneurial
venture,” Prod. Oper. Manage., vol. 22, no. 6, pp. 1429–1438, 2013.
[30] D. Shepherd and H. Patzelt, Trailblazing in Entrepreneurship: Creating
new Paths for Understanding the Field. Cham, Switzerland: Palgrave
Macmillan, 2017.
[31] P. Davidsson, “Entrepreneurial opportunities and the entrepreneurship
nexus: A re-conceptualization,” J. Bus. Venturing, vol. 30, pp. 674–695,
2015.
[32] K. Ulrich and S. Eppinger, Product Design and Development. Maiden-
head, U.K.: McGraw-Hill, 2011.
[33] E. von Hippel and G. von Krogh, “Identifying viable ’need-solution pairs’:
Problem solving without problem formulation,” Org. Sci., vol. 27, no. 1,
pp. 207–221, 2016.
[34] S. A. Alvarez, J. B. Barney, and P. Anderson, “Forming and exploiting
opportunities: The implications of discovery and creation processes for
entrepreneurial and organizational research,” Org. Sci., vol. 21, no. 1,
pp. 301–317, 2013.
[35] N. Dew, S. Read, S. D. Sarasvathy, and R. Wiltbank, “Effectual versus
predictive logics in entrepreneurial decision-making: Differences between
experts and novices,” J. Bus. Venturing, vol. 24, pp. 287–309, 2009.
[36] S. Sarasvathy, “Causation and effectuation: Toward a theoretical shift from
economic inevitability to entrepreneurial contingency,” Acad. Manage.
Rev., vol. 26, pp. 243–263, 2001.
[37] E. S. C. Berger and A. Kuckertz, “The challenge of dealing with com-
plexity in entrepreneurship, innovation and technology research: An in-
troduction,” in Complexity in Entrepreneurship, Innovation and Technol-
ogy Research: Applications of Emergent and Neglected Methods. Cham,
Switzerland: Springer, 2016, pp. 1–9.
[38] B. Kogut and N. Kulatilaka, “Capabilities as real options,” Org. Sci.,
vol. 12, no. 6, pp. 744–758, 2001.
[39] M. Levesque, “Mathematics, theory, and entrepreneurship,” J. Bus. Ven-
turing, vol. 19, no. 5, pp. 743–765, 2004.
[40] D. A. Shepherd, J. S. McMullen, and P. D. Jennings, “The formation of
opportunity beliefs: Overcoming ignorance and reducing doubt,” Strategic
Entrepreneurship J., vol. 1, no. 1/2, pp. 75–95, 2007.
[41] J. G. March, “Rationality, foolishness and adaptive intelligence,” Strategic
Manage. J., vol. 27, pp. 201–214, 2006.
[42] G. Pye, “A Markov model of the term structure,” Quart. J. Econ., vol. 80,
pp. 60–72, 1966.
[43] O. L. V. Costa and M. V. Araujo, “A generalized multi-period mean-
variance portfolio optimization with Markov switching parameters,” Au-
tomatica, vol. 44, pp. 2487–2497, 2008.
[44] R. J. Elliott and J. Hinz, “Portfolio optimization, hidden Markov models,
and technical analysis of point and figure charts,” Int. J. Theoretical Appl.
Finance, vol. 5, pp. 385–399, 2002.
[45] J. Sass and U. G. Hausmann, “Optimizing the terminal wealth under
partial information: The drift process as a continuous time Markov chain,”
Finance Stochastics, vol. 8, pp. 553–577, 2004.
[46] G. R. Madey and B. V. Dean, “Strategic planning for investment in R&D
using decision analysis and mathematical programming,” IEEE Trans.
Eng. Manage., vol. EM-32, no. 2, pp. 84–90, May 1985.
[47] A. J. Hoffman and J. G. Woody, Climate Change-What’s Your Business
Strategy? Brighton, MA, USA: Harvard Business Press, 2008.
[48] R. M. Bakker and D. A. Shepherd, “Pull the plug or take the plunge: Mul-
tiple opportunities and the speed of venturing decisions in the australian
mining industry,” Acad. Manage. J., vol. 60, no. 1, pp. 1–26, 2017.
[49] R. G. McGarth, “Falling forward: Real options reasoning and en-
trepreneurial failure,” Acad. Manage. Rev., vol. 24, no. 1, pp. 13–30,
1999.
[50] G. D. Forney, “The Viterbi algorithm,” Proc. IEEE, vol. 61, no. 3, pp. 268–
278, Mar. 1973.
[51] C. Terwiesch and X. Yu, “Innovation contests, open innovation, and mul-
tiagent problem solving,” Manage. Sci., vol. 54, no. 9, pp. 1529–1543,
2008.
[52] C. G. Brush, P. G. Greene, M. M. Hart, and H. S. Haller, “From initial
idea to unique advantage: The entrepreneurial challenge of constructing
a resource base,” Acad. Manage. Executive, vol. 15, no. 1, pp. 64–78,
2001.
[53] A. Smith, A. Stirling, and F. Berkhout, “The governance of sustainable
socio-technical transitions,” Res. Policy, vol. 34, pp. 1491–1510, 2005.
[54] G. Charness, U. Gneexy, and A. Imas, “Experimental methods: Eliciting
risk preferences,” J. Econ. Behav. Org., vol. 87, pp. 43–51, 2013.
ÇANAKOĞLU et al.: HOW DATA-DRIVEN ENTREPRENEUR ANALYZES IMPERFECT INFORMATION
[55] F. Tanrisever, S. S. Erzurumlu, and N. Joglekar, “Production, process
investment, and the survival of debt-financed startup firms,” Prod. Oper.
Manage., vol. 21, no. 4, pp. 637–652, 2012.
[56] S. S. Erzurumlu, J. Davies, and N. Joglekar, “Managing highly innovative
projects: The influence of design characteristics on project valuation,”
IEEE Trans. Eng. Manage., vol. 61, no. 2, pp. 349–361, May 2014.
[57] J. J. Janney and T. B. Folta, “Moderating effects of investor experience
on the signaling value of private equity placements,” J. Bus. Venturing,
vol. 21, pp. 27–44, 2006.
Ethem Çanakoğlu received the B.S. degree in elec-
trical and electronics engineering, and the M.S. de-
gree in industrial engineering from Bogaziçi Univer-
sity, Istanbul, Turkey, in 2000 and 2004, respectively,
and the Ph.D. degree in industrial engineering and
operations management from Koç University, Istan-
bul, under the supervision of Prof. Özekici.
He is currently an Assistant Professor with
Bahçesehir University, Istanbul. After completing the
Ph.D. degree, he was at Warwick Business School
for three years as a Research Fellow. He utilized his
expertise in stochastic modeling and optimization to different areas such as
portfolio optimization, risk management in energy markets, or asset liabil-
ity management. He has published several international papers in high-quality
journals and supervised numerous theses.
Authorized licensed use limited to: Vignan Institute of Technology & Science. Downloaded on February 17,2023 at 08:45:04 UTC from IEEE Xplore. Restrictions apply.
617
S. Sinan Erzurumlu received the B.S. degree in elec-
trical and electronics engineering from Bogazici Uni-
versity, Istanbul, Turkey, in 1999, and the M.S. degree
in operations research/industrial engineering and the
doctoral degree in management science and informa-
tion systems from the University of Texas at Austin,
Austin, TX, USA, in 2002 and 2007, respectively.
His work experience includes technology and op-
erations consulting at technology organizations. He
is currently a Faculty Member with the Technology,
Operations, and Information Management Division,
Babson College, Babson Park, MA, USA. His research interests include the
confluence of technology and operations management, entrepreneurship, and
innovation strategy.
Dr. Erzurumlu is also with the Production and Operations Management
Society (POMS), the Institute for Operations Research and Management Sci-
ence (INFORMS) Technology Innovation Management and Entrepreneurship
Section (TIMES), and the Manufacturing and Service Operations Management
Society (M&SOM), and the Academy of Management (AoM) Entrepreneurship
Division (ENT) and Technology and Innovation Management Division (TIM).
Yaman Omer Erzurumlu received the B.S. degree
in engineering management from Istanbul Techni-
cal University, Istanbul, Turkey, in 1998, and the
Ph.D. degree in finance from the University of Cen-
tral Florida, Orlando, FL, USA, in 2006.
His work experience includes financial consulting
with different small and medium enterprises. He is
currently with the faculty of the Engineering Manage-
ment Department, Bahcesehir University, Istanbul.
His research interests include mutual/pension fund
management and governance, risk management, en-
trepreneurial finance, and innovation strategy.
Dr. Erzurumlu is also with the Financial Management Association (FMA),
the Eurasia Business and Economics Society (EBES), and the Society for Fi-
nancial Analysts (SFA).