Available online at www.sciencedirect.

com
ScienceDirect
ScienceDirect
Procedia
Available Computer
online Science 00 (2022) 000–000
at www.sciencedirect.com
Procedia Computer Science 00 (2022) 000–000 www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
ScienceDirect
Procedia Computer Science 219 (2023) 60–67

CENTERIS - International Conference on ENTERprise Information Systems / ProjMAN -

CENTERIS
International - International
Conference Conference
on Project on ENTERprise
MANagement / HCist -Information Systems
International / ProjMAN
Conference -
on Health
International Conference on Care
and Social Project MANagement
Information / HCist
Systems and -Technologies
International 2022
Conference on Health
and Social Care Information Systems and Technologies 2022
Decision Theory and Risk Simulation Analysis for
Decision Theory and Risk Simulation Analysis for
Optimizing Profit in PayLater Services
Optimizing Profit in PayLater Services
Nicholas Dominica,a,*, Bens Pardameana,b
Nicholas Dominic *, Bens Pardameana,b
a
Bioinformatics and Data Science Research Center, Bina Nusantara University, Jakarta, Indonesia 11480
b
Computer Science
a Department,and
Bioinformatics BINUS
DataGraduate Program,Center,
Science Research MasterBina
of Computer Science
Nusantara Program,
University, BinaIndonesia
Jakarta, Nusantara11480
University, Jakarta,
b
Computer Science Department, BINUS Graduate Program,Indonesia,
Master of 11480
Computer Science Program, Bina Nusantara University, Jakarta,
Indonesia, 11480

Abstract
Abstract
The projected increase in PayLater utilization reaches up to five million people by 2025. To optimize the yearly profit from their
The projected
PayLater increase
service, fintechincompanies
PayLater utilization
must examinereaches up to five
all possible million
risks people
before by 2025.decision
a unanimous To optimize the Therefore,
is taken. yearly profitwefrom their
proposed
PayLater service, fintech
a unified decision companies
framework derivedmust
fromexamine
decisionalltheory
possible
andrisks beforeCarlo
the Monte a unanimous decision
simulation is taken.
technique. TwoTherefore,
schemes werewe proposed
coined:
a(1)unified
a decision-making scheme,
decision framework and (2)
derived a risk
from simulation
decision theoryscheme.
and the Throughout
Monte Carloexperiments, the framework
simulation technique. was ablewere
Two schemes to estimate
coined:
(1) a decision-making
several scheme,
alternative decisions andand
their(2) a risk simulation
impacts, analyze thescheme.
causes ofThroughout experiments,
failure and delays the framework
in the development wasPayLater
of the able to estimate
service,
several
and alternative
execute Montedecisions and their impacts,
Carlo simulations in up toanalyze
10,000the causes
trials. of failure
Outputs and study
of this delayswill
in the development
benefit of the PayLater
decision-makers in the service,
fintech
initiative
and executebefore launching
Monte Carlo their PayLater
simulations in products.
up to 10,000 trials. Outputs of this study will benefit decision-makers in the fintech
initiative before launching their PayLater products.
© 2022 The Authors. Published by Elsevier B.V.
© 2022
This
© 2023 The
is an Authors.
open
The Published
accessPublished
Authors. by Elsevier
article under
by Elsevier B.V.
the CC BY-NC-ND
B.V. license (https://creativecommons.org/licenses/by-nc-nd/4.0)
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
This is an open
Peer-review access
under article underofthethe
responsibility CCscientific
BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
committee of the CENTERIS - International Conference on ENTERprise
Peer-review under responsibility of the scientific committee of the CENTERIS – International Conference on ENTERprise Information
Peer-review
Information under
Systems responsibility
/ ProjMAN - of the scientific
International committee
Conference
Systems / ProjMAN - International Conference on Project MANagement on of the MANagement
Project /CENTERIS -/ International
HCist Conference
HCist - International Conference
- International onandENTERprise
Conference
on Health on Health
Social Care
and Social Care
Information
Information Information
Systems
Systems / ProjMAN
and Systems and
2022Technologies
- International
Technologies Conference 2022on Project MANagement / HCist - International Conference on Health
and Social Care Information Systems and Technologies 2022
Keywords: PayLater; decision-making; Bayesian analysis; fault tree analysis; critical path method; Monte Carlo simulation.
Keywords: PayLater; decision-making; Bayesian analysis; fault tree analysis; critical path method; Monte Carlo simulation.

* Corresponding author. Tel.: +62-21-534-5830 ext. 2456; fax: +62-21-530-0244.

* E-mail [email protected]
address:author.
Corresponding Tel.: +62-21-534-5830 ext. 2456; fax: +62-21-530-0244.
E-mail address: [email protected]
1877-0509 © 2022 The Authors. Published by Elsevier B.V.
This is an open
1877-0509 access
© 2022 Thearticle under
Authors. the CC BY-NC-ND
Published license (https://creativecommons.org/licenses/by-nc-nd/4.0)
by Elsevier B.V.
Peer-review
This under
is an open responsibility
access of the scientific
article under CC BY-NC-NDcommittee of the
license CENTERIS - International Conference on ENTERprise Information Systems /
(https://creativecommons.org/licenses/by-nc-nd/4.0)
ProjMAN
Peer-review- International Conference
under responsibility onscientific
of the Project MANagement / HCist
committee of the - International
CENTERIS Conference
- International on Health
Conference onand Social CareInformation
ENTERprise InformationSystems
Systems/
and Technologies
ProjMAN 2022 Conference on Project MANagement / HCist - International Conference on Health and Social Care Information Systems
- International
and Technologies 2022
1877-0509 © 2023 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of the CENTERIS – International Conference on ENTERprise
Information Systems / ProjMAN - International Conference on Project MANagement / HCist - International Conference
on Health and Social Care Information Systems and Technologies 2022
10.1016/j.procs.2023.01.264
 Nicholas Dominic et al. / Procedia Computer Science 219 (2023) 60–67 61
2 Nicholas Dominic et al. / Procedia Computer Science 00 (2022) 000–000

1. Introduction

PayLater services offer straightforward access to credit installment without any complicated banking process; it is
proven to have a strong correlation with the product (0.72 of Pearson Correlation Coefficient/PCC) and transaction
(0.41 of Matrix Coding/MC) variables [1], which means many users rely on PayLater for product transactions.
projected increase in the PayLater Card utilization reaches up to five million people by 2025 [2, 3], due to its efficiency
of use [4]. However, as a short-term debt option, this service brings negative habits and hedonic motives [5]. To
suppress the desire to use PayLater, a study claimed that financial literacy is necessary so users can evaluate each of
their financial decisions [6]. From the fintech (financial technology) companies’ viewpoint, the careful decision of the
users can be a drawback [7], since customers’ loyalty [8] in using their PayLater will be a huge advantage in terms of
profit [9]. To optimize this yearly profit, fintech companies must examine all possible risks before a unanimous
decision is taken [10]. In this study, some quantitative risk analysis techniques were presented to address the issue.
Consider an unknown fintech company with its PayLater service, the objective is to demonstrate algorithms for (1)
decisions under risk and uncertainty, (2) fault calculation, (3) critical path method (CPM), and (4) Monte Carlo
simulation. Note that in fact, the same technique can be used from a user perspective as part of financial literacy. The
unknown status of the name of the fintech is due to restricted and confidential access to the related financial data silos.
Assume that the fintech in question is similar to Gojek, OVO, Traveloka, Kredivo, or Shopee. The contributions in
this research are:
• propose a unified decision framework derived from decision theory under uncertainty and risk with several
criteria, supported by Bayesian analysis [11, 12] and graphical decision tree, and
• simulate quantitative fault tree analysis, critical path method, and Monte Carlo given the objective function of
fintech yearly net profit obtained from their respective PayLater services.
To the best of our knowledge, this is the first joint use of decision theory techniques and risk simulation to analyze
and optimize the profit of fintech companies. Outputs of this study will benefit decision-makers in the fintech initiative
before launching their PayLater products.

2. Literature Review

2.1. Decision Theory

All available data (or resources) and possible alternatives [13] should be considered before making a good decision
[14–16]. Decision-making under uncertainty (DMU) explains a circumstance where there are unmeasurable possible
outcomes for each alternative, since the decision-makers may not know the availability of their resources. The
variation is called one-criterion decision-making under uncertainty (1-DM/U), where each alternative is evaluated
based on one objective [17]. On the other hand, decision-making under risk (DMR) explains a circumstance where
there are measurable possible outcomes, calculated by the probability of occurrence, for each alternative [13, 18].

2.2. Monte Carlo Simulation

The main quantitative simulation technique to optimize the fintech profit is Monte Carlo simulation, an
international standard governed by ISO 31000 Risk Management Standard and ISO 31010 Risk Assessment
Techniques. It works by randomizing uncertain variables in the objective function of interest within 𝑛𝑛 simulation. The
Monte Carlo simulation technique has been proved to strengthen enterprise risk management by preparing the optimal
contingency plans for each alternative scenario for decision-making processes and thus maximizing profit [19].
In small-medium banks (SMBs), the Monte Carlo simulation coupled with some econometric approaches can create
a solid and compact financial planning framework, which is critical for the overall bank efficiency [20]. Moreover, by
combining the Additional Funds Needed (AFN) methodology, the Monte Carlo simulation demonstrated a
profitability control of SMBs and identified the best confidence interval more comprehensively and reliably [21]. An
optimal financing portfolio can also be created by utilizing Monte Carlo simulation and Markowitz’s Modern Portfolio
Theory [22]. With Geometric Brownian Motion (GBM), the simulation was enhanced to forecast the future price
62 Nicholas Dominic et al. / Procedia Computer Science 219 (2023) 60–67
Nicholas Dominic et al. / Procedia Computer Science 00 (2022) 000–000 3

movement of multinational companies and banks [23]. However, Monte Carlo simulations are sometimes very
exhaustive and hence cloud computing [24, 25] is required to execute them.

3. Methodology

3.1. Variables

For the decision-making scheme, the independent variable is the completion date for each project activity 𝑝𝑝𝑡𝑡 ∈ 𝑃𝑃
(𝑡𝑡 denotes time) and the dependent variable is expected profit under favorable and unfavorable markets. For the risk
simulation scheme, the independent variable is divided into certain and uncertain variables. Certain variables include
interest rate per month 𝑖𝑖𝑚𝑚 and corporate first-year budget 𝑏𝑏𝑡𝑡=1 , while uncertain variables include labor cost 𝑐𝑐𝑙𝑙 ,
additional cost 𝑐𝑐𝑎𝑎 , and first-year customer demand 𝑑𝑑𝑡𝑡=1 . In this scheme, the dependent variable is expected profit
𝑓𝑓(𝑖𝑖𝑚𝑚 , 𝑏𝑏𝑡𝑡 , 𝑐𝑐𝑙𝑙 ~𝐷𝐷, 𝑐𝑐𝑎𝑎 ~𝒰𝒰(𝛼𝛼, 𝛽𝛽), 𝑑𝑑𝑡𝑡 ~𝒩𝒩(𝜇𝜇, 𝜎𝜎)), where 𝒟𝒟, 𝒰𝒰(𝛼𝛼, 𝛽𝛽), and 𝒩𝒩(𝜇𝜇, 𝜎𝜎) refer to discrete, uniform, and normal
distributions, respectively.

3.2. Data Collection

The generated dataset was used to run all simulations. For the decision table, see Table 1. The symbol Ω denotes
the best-case scenario (favorable market) for each alternative, while 𝒪𝒪 denotes the worst-case scenario (unfavorable
market). Note that the coefficient of realism to calculate the Hurwicz criterion is set to 𝜌𝜌 = 0.8. For the Bayesian, the
positive conditional probability in a favorable market is set to 𝑃𝑃(+|Ω) = 0.7, and thus the reverse is 𝑃𝑃(−|Ω) = 1 −
𝑃𝑃(+|Ω) = 0.3. The positive conditional probability in an unfavorable market is 𝑃𝑃(+|𝒪𝒪) = 0.2, and thus the reverse
is𝑃𝑃(−|𝒪𝒪) = 1 − 𝑃𝑃(+|𝒪𝒪) = 0.8.
To simulate a simple quantitative Fault Tree Analysis, three possible threats are suggested, i.e., team
miscommunication (with a probability of 0.8), internal conflicts (with a probability of 0.5), and malware attacks (with
a probability of 0.3). In Table 2, eight project activities were registered to fulfill the deployment of the PayLater
service, each with the best (Ω), average (Θ), and worst (𝒪𝒪) case scenario (in days). Every activity is also marked with
1 or 0, which means it is within or without the critical path, respectively. This critical path marking is useful for
calculating the average and variance of PayLater project completion time.

Table 1. Decision table.

State of Nature (Decision table) State of Nature (Opportunity loss table)
Alternative [Alt]
Ω 𝒪𝒪 Ω 𝒪𝒪
Release in January 2024 [Alt-1] $ 300,000 $ (170,000) - $ 170,000
Release in February 2024 [Alt-2] $ 200,000 $ (200,000) $ 100,000 $ 200,000
Release in March 2024 [Alt-3] $ 150,000 $ (225,000) $ 150,000 $ 225,000
Cancel the PayLater project [Alt-4] - - $ 300,000 -

3.3. Research Methodology

As diagrammed in Fig. 1, there are two starting points. First, in the decision-making scheme, there are DMR and
DMU, supported by probability values obtained from Bayesian analysis. The scheme tells of a decision making for
the release of the PayLater project given several input variables, 𝑝𝑝𝑡𝑡 . Quantitative Fault Tree analysis and CPM are
used iff a delay occurs. Second, a Monte Carlo simulation was demonstrated to estimate the net profit of the fintech
company from releasing its PayLater service, given two certain and three uncertain variables. The dashed line
connecting the Monte Carlo simulation and DMR means both can be interchangeably used for considerations.
Decision-making under uncertainty. Suppose 𝑖𝑖 represents a row (each alternative), 𝑗𝑗 represents existing columns
(Ω and 𝒪𝒪), and 𝑗𝑗 + 1 represents an additional column for each criterion 𝛾𝛾𝑖𝑖 ∈ Γ. Five criterion formulas in the DMU
are defined as follows.
1. Optimistic criterion (𝜸𝜸𝟏𝟏 ): 𝑚𝑚𝑚𝑚𝑚𝑚(𝛾𝛾𝑗𝑗+1 ), where 𝛾𝛾𝑗𝑗+1 = 𝑚𝑚𝑚𝑚𝑚𝑚(Ω𝑖𝑖 , 𝒪𝒪𝑖𝑖 ) = 𝑀𝑀𝑀𝑀𝑀𝑀.
2. Pessimistic criterion (𝜸𝜸𝟐𝟐 ): 𝑚𝑚𝑚𝑚𝑚𝑚(𝛾𝛾𝑗𝑗+1 ), where 𝛾𝛾𝑗𝑗+1 = 𝑚𝑚𝑚𝑚𝑚𝑚(Ω𝑖𝑖 , 𝒪𝒪𝑖𝑖 ) = 𝑀𝑀𝑀𝑀𝑀𝑀.
 Nicholas Dominic et al. / Procedia Computer Science 219 (2023) 60–67 63
4 Nicholas Dominic et al. / Procedia Computer Science 00 (2022) 000–000

3. Hurwicz or realism criterion (𝜸𝜸𝟑𝟑 ): 𝑚𝑚𝑚𝑚𝑚𝑚(𝛾𝛾𝑗𝑗+1 ), where 𝛾𝛾𝑗𝑗+1 = 𝜌𝜌(𝑀𝑀𝑀𝑀𝑀𝑀) + (1 − 𝜌𝜌)(𝑀𝑀𝑀𝑀𝑀𝑀).

4. Laplace or average criterion (𝜸𝜸𝟒𝟒 ): 𝑚𝑚𝑚𝑚𝑚𝑚(𝛾𝛾𝑗𝑗+1 ), where 𝛾𝛾𝑗𝑗+1 = 0.5(𝑀𝑀𝑀𝑀𝑀𝑀 − 𝑀𝑀𝑀𝑀𝑀𝑀).
5. Minimax regret criterion (𝜸𝜸𝟓𝟓 ): 𝑚𝑚𝑚𝑚𝑚𝑚(𝛾𝛾𝑗𝑗+1 ), where 𝛾𝛾𝑗𝑗+1 = 𝑚𝑚𝑚𝑚𝑚𝑚(Ω𝑖𝑖 − 𝑚𝑚𝑚𝑚𝑚𝑚(Ω𝑖𝑖 ), 𝑚𝑚𝑚𝑚𝑚𝑚(𝒪𝒪𝑖𝑖 ) − 𝒪𝒪𝑖𝑖 ).
Decision-making under risk. Four criterion formulas in the DMR are defined as follows.
1. EMV (𝜸𝜸𝟔𝟔 ): 𝑚𝑚𝑚𝑚𝑚𝑚(𝛾𝛾𝑗𝑗+1 ), where 𝛾𝛾𝑗𝑗+1 = Ω𝑖𝑖 𝑃𝑃(Ω| +) + 𝒪𝒪𝑖𝑖 𝑃𝑃(𝒪𝒪| +).
2. EVwPI (𝜸𝜸𝟕𝟕 ): 𝑚𝑚𝑚𝑚𝑚𝑚(Ω𝑖𝑖 )𝑃𝑃(Ω| +) + 𝑚𝑚𝑚𝑚𝑚𝑚(𝒪𝒪𝑖𝑖 )𝑃𝑃(𝒪𝒪| +).
3. EVPI (𝜸𝜸𝟖𝟖 ): EVwPI − max (EMV).
4. EOL (𝜸𝜸𝟗𝟗 ): (Ω𝑖𝑖 − 𝑀𝑀𝑀𝑀𝑀𝑀)𝑃𝑃(Ω| +) + (𝑀𝑀𝑀𝑀𝑀𝑀 − 𝒪𝒪𝑖𝑖 )𝑃𝑃(𝒪𝒪| +).

Table 2. CPM table (in days).

Activity Ω Θ 𝒪𝒪 𝑡𝑡* 𝑐𝑐?** 𝜇𝜇 𝜎𝜎
Project initialization 5 10 15 10 1 10 2 2/ 7
Application design 7 14 28 16 1 16 121/4
Develop prototype 14 21 30 22 0 - -
Develop alpha version 7 14 21 14 0 - -
Alpha testing 3 7 14 8 1 8 3 1/ 3
Develop beta testing 7 21 30 21 0 - -
Beta testing 7 10 21 12 1 12 5 4/ 9
Application deployment 3 7 10 7 0 - -
Total - - - 110 - 46 235/6
* 𝑡𝑡 = expected time. ** 𝑐𝑐? = on the critical path? (0=False, 1=True)

Fig. 1. The research framework

Bayesian analysis. Without any real market condition survey, assume that the estimation of both favorable and
unfavorable markets is 𝑃𝑃(Ω) = 𝑃𝑃(𝒪𝒪) = 0.5. The Bayes’ Theorem can be reapplied for our positive data (+) as

𝑃𝑃(+|Ω)𝑃𝑃(Ω)
𝑃𝑃(Ω| +) = (1)
𝑃𝑃(+|Ω)𝑃𝑃(Ω) + 𝑃𝑃(+|𝒪𝒪)𝑃𝑃(𝒪𝒪)
 Nicholas Dominic et al. / Procedia Computer Science 00 (2022) 000–000 5
64 Nicholas Dominic et al. / Procedia Computer Science 219 (2023) 60–67

𝑃𝑃(+|𝒪𝒪)𝑃𝑃(𝒪𝒪)
𝑃𝑃(𝒪𝒪| +) = (2)
𝑃𝑃(+|𝒪𝒪)𝑃𝑃(𝒪𝒪) + 𝑃𝑃(+|Ω)𝑃𝑃(Ω)

Fault tree analysis. Recall the logic gate operation of AND and OR. In fault tree analysis, if an AND operation is
encountered, simply multiply the probabilities value of all events. Nevertheless, if an OR operation is encountered,
use this formula

𝑃𝑃(𝐸𝐸) = 1 − (1 − 𝑃𝑃(𝑒𝑒1 )) − (1 − 𝑃𝑃(𝑒𝑒2 )) − ⋯ − (1 − 𝑃𝑃(𝑒𝑒𝑛𝑛 )) (3)

where 𝑒𝑒𝑖𝑖 ∈ 𝐸𝐸 represents the event of interest.

Critical path method (CPM). In CPM, only the activity within the critical path that we should calculate its project
Ω+4𝜃𝜃+𝒪𝒪 𝒪𝒪−Ω 2
mean, 𝜇𝜇 = and variance, 𝜎𝜎 = ( ) [26]. Note that (1) Ω, Θ, and 𝒪𝒪 refer to the best, average, and worst-
6 6
case scenario for completing each activity (in days) and (2) the 𝜎𝜎 formula is based on ±3 standard deviations.
Monte Carlo simulation. The pseudocode for this simulation is depicted in Algorithm 1. By default, we set 𝑖𝑖𝑚𝑚 =
0.25, 𝑏𝑏𝑡𝑡=1 = 50,000 (in USD), 𝑙𝑙𝑙𝑙𝑙𝑙 = 1,000, ℎ𝑖𝑖𝑖𝑖ℎ = 1,200, 𝜇𝜇 = 30,000, 𝜎𝜎 = 1,500, and 𝑛𝑛 = 10,000.

Algorithm 1 Monte Carlo Simulation

Require: 0 < 𝑖𝑖𝑚𝑚 ≤ 1, 𝑏𝑏𝑡𝑡=1 > 0, and 𝑛𝑛 > 1
1: 𝑐𝑐𝑙𝑙 ← get_discrete_distribution(𝑐𝑐, 𝑃𝑃(𝑐𝑐))
2: 𝑐𝑐𝑎𝑎 ← get_uniform_disctribution(𝑙𝑙𝑙𝑙𝑙𝑙, ℎ𝑖𝑖𝑖𝑖ℎ)
3: 𝑑𝑑𝑡𝑡=1 ← get_normal_distribution(𝜇𝜇, 𝜎𝜎)
4: for 𝑛𝑛 in 𝑁𝑁 do // 𝑛𝑛 = number of iterations
5: return 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 ← (𝑏𝑏𝑡𝑡=1 − 𝑐𝑐𝑙𝑙 − 𝑐𝑐𝑎𝑎 ) − (𝑑𝑑𝑡𝑡=1 (1 + 𝑖𝑖𝑚𝑚 ))
6: end for

3.4. Tools and Files Availability

All calculations in the decision-making scheme were done with Microsoft Excel version 2112, from Microsoft 365.
The Monte Carlo Simulation was executed using Python version 3.7.1 as the main programming language and
interactive Jupyter Notebook version 6.0.1 as the environment. Excel file and Python code are available for access
through https://github.com/NicholasDominic/PayLater-Profit-Optimizer-Framework.

4. Results and Discussions

4.1. Decision-making scheme

The result for all criterions is presented in Table 3. EMV tells the long-term value of each decision taken. For
instance, if a fintech company decides to take [Alt-1], it is expected it will get $300,000.00. But in the long-term, the
decision is only worth up to $196,600.00. For clarity, scrutinize the decision tree in Fig. 2 (left).
The values 0.78 and 0.22 come from Table 4. Let the prior probability for both best- and worst-case scenarios be
𝑃𝑃(Ω) = 0.5 and 𝑃𝑃(𝒪𝒪) = 0.5. The joint probability is calculated by multiplying the conditional probability with the
(0.70)(0.50) 0.35
respective prior probability. Finally, we obtained 𝑃𝑃(Ω| +) = (0.70)(0.50)+(0.20)(0.50) = = 0.78 and 𝑃𝑃(𝒪𝒪| +) =
0.45
(0.20)(0.50) 0.10
(0.20)(0.50)+(0.70)(0.50)
= = 0.22Ǥ
0.45
To compute Minimax Regret and EOL, please firstly recheck Table 1 (opportunity loss). This is how the values in
the table are obtained. Based on this same table, the maximum value in the favorable market is 𝑚𝑚𝑚𝑚𝑚𝑚(Ω𝑖𝑖 ) = $300,000.
Therefore, the opportunity loss of Alt-1 (favorable market) is calculated as $300,000 – $300,000 = $0. For [Alt-2]
(favorable market), it is calculated as $300,000 – $200,000 = $100,000. On the other hand, the maximum value in the
unfavorable market is 𝑚𝑚𝑎𝑎𝑎𝑎(𝒪𝒪𝑖𝑖 ) = $0. Therefore, the opportunity loss of [Alt-1] (unfavorable market) is calculated as
 Nicholas Dominic et al. / Procedia Computer Science 219 (2023) 60–67 65
6 Nicholas Dominic et al. / Procedia Computer Science 00 (2022) 000–000

$0 – (– $170,000) = $170,000. For [Alt-4] (unfavorable market), it is calculated as $0 – $0 = $0. In one sentence, the
opportunity loss table talks about, “If the fintech company decides that the PayLater product will be released in
January 2024 [Alt-1] and the market conditions are good, then no losses will be achieved. On the other hand, if the
market conditions are bad, this decision will cost the company $170,000.”

Fig. 2. Decision tree diagram (left) and fault tree analysis diagram (right).

Table 3. DMU and DMR results.

DMU (𝛾𝛾1−5 ) DMR (𝛾𝛾6−9 )
Alt
𝛾𝛾1 𝛾𝛾2 𝛾𝛾3 𝛾𝛾4 𝛾𝛾5 𝛾𝛾6 𝛾𝛾7 𝛾𝛾8 𝛾𝛾9
[Alt-1] $ 300,000 $ (170,000) $ 235,000 $ 206,000 $ 170,000 $ 196,600 $ 37,400
[Alt-2] $ 200,000 $ (200,000) $ 200,000 $ 120,000 $ 200,000 $ 112,000 $ 122,000
$ 234,000 $ 37,400
[Alt-3] $ 150,000 $ (225,000) $ 187,500 $ 75,000 $ 225,000 $ 67,500 $ 166,500
[Alt-4] - - - - $ 300,000 - $ 234,000

Table 4. Bayesian analysis.

State of Conditional Prior Posterior probability
Nature Probability Probability Joint probability 𝑃𝑃(State of Nature|+)
Ω 0.70 0.50 0.35 0.78
𝒪𝒪 0.20 0.50 0.10 0.22
Total - - 0.45 1.00

Table 5. Monte Carlo simulation results.

Trial 𝑐𝑐𝑙𝑙 c𝑎𝑎 𝑑𝑑𝑡𝑡=1 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 (in USD) ET (in ms)
100 1,650 1,176.50 30,047 9,560.35 ± 1,795.25 71.97
250 1,550 1,173.64 31,137 9.706.23 ± 1,868.41 134.65
500 1,550 1,077.59 32,228 9.745.03 ± 1,859.97 240.75
1,000 1,650 1,176.52 31,347 9,740.61 ± 1,867.98 519.68
2,000 1,650 1,066.38 32,155 9.833.26 ± 1,832.86 1,180.93
5,000 1,600 1,078.30 27,571 9,801.05 ± 1,883.61 2,701.17
7,500 1,600 1,044.80 29,391 9,807.53 ± 1,877.40 4,199.79
10,000 1,650 1,071.77 29,404 9.814.99 ± 1,887.40 6,140.65

If the development of the PayLater project is delayed, the company needs to (1) find and calculate the cause by
using the fault tree analysis, as shown in Fig. 2 (right). Based on the fault tree, we know that miscommunication (A)
and internal conflicts (B) will probably lead to member resignation (C). The calculation is 𝑃𝑃(𝐶𝐶) = 𝑃𝑃(𝐴𝐴)𝑃𝑃(𝐵𝐵) =
(0.80)(0.50) = 0.40. Both member resignation (C) or possible malware attacks (D) will then probably lead to the
project delay. The chance is calculated as (𝐸𝐸) = 1 − (1 − 𝑃𝑃(𝐶𝐶))(1 − 𝑃𝑃(𝐷𝐷)) = 1 − (1 − 0.40)(1 − 0.30) = 0.58.
Therefore, there is more than a 50% chance the project will be delayed if this scenario is followed. Next, the
company needs to (2) calculate the probability of completing the project at the new specified time by using the CPM.
From Table 2, each activity within the critical path has both mean and variance values. The total mean is ∑ 𝜇𝜇𝑖𝑖 = 46
(in days), while the total variance is ∑ 𝜎𝜎𝑖𝑖 = 235/6 (in days). With the help of the chart tool in Excel, two-dimensional
66 Nicholas Dominic et al. / Procedia Computer Science 219 (2023) 60–67
Nicholas Dominic et al. / Procedia Computer Science 00 (2022) 000–000 7

probability density function (PDF) curves within (𝜇𝜇, 𝜎𝜎) and in ranges (20, 70] were established. Now assume that the
PayLater development is delayed to 50 days (four days late). The chance of the project being completed within 50
days can be calculated by 𝑃𝑃(𝑥𝑥 ≤ 50) = (𝑥𝑥 − 𝜇𝜇)/𝜎𝜎 = (50 − 46)/23.833 = 𝑃𝑃(𝑧𝑧 > 0.91) ≈ 81.859%.
Using the Area Under the Standard Normal Curve (AUC-SN), we know that the value of 0.91 is equivalent to an
81.859% chance. Thus, there is still an 81.859% chance that the PayLater project will be completed within 50 days.
On the contrary, if the fintech company wants the project to be completed four days faster, then the chance is
𝑃𝑃(𝑥𝑥 > 50) = 1.0 − 𝑃𝑃(𝑥𝑥 ≤ 50) = 18.141% since the AUC-SN is symmetric, 𝑃𝑃(𝑧𝑧 > 0.91) = 𝑃𝑃(𝑧𝑧 < −0.91).

4.2. Risk Simulation Scheme

Fig. 3. PDF from Monte Carlo simulation.

When we set default parameters to run the Monte Carlo simulation, in the execution time (ET) of about 6 seconds
we obtained 9,814.99 (in USD) and 1,887.40 (in USD) as the mean and variance of the expected net profit,
respectively. We also tried to produce the results under different 𝑛𝑛 of trials. As seen in Fig. 3 and Table 5, the higher
the number of 𝑛𝑛, the more confident the result is. In another word, the results presented will be more reliable when
the simulation is executed under a higher number of trials, as the Law of Large Numbers suggested.
Note that all uncertain variables were also randomized under different probability distributions. By this example,
given USD 50,000 as the initial corporate first-year budget and 0.25 as the initial monthly interest rate charged to each
PayLater transaction, the fintech company will be expected to profit USD 9,814.99 (by average) at the end of the year.

5. Conclusions and Future Works

In this study, we succeeded in simulating a novel framework for optimizing the profit of a PayLater service provider
company. Throughout multiple experiments, this framework was able to estimate several alternative decisions and
their impacts, analyze the causes of failure and delays in the development of the PayLater service, and execute Monte
Carlo simulations in up to 10,000 trials. The presented framework is the best fit for simulating alternative possibilities
while estimating the risks that come with them. Eventually, some forthcoming works are left, including (1) the
sensitivity analysis for each decision-making, (2) demonstrating the Decision Tree Learning algorithm for better
decision results representation, and (3) the usage of Utility Theory to assess the worth of each alternative taken.

Acknowledgements

This research is fully supported by Bina Nusantara University (BINUS).

References

[1] Pratika, Yeyen, Salahudin, Salahudin, Riyanto, Dicky Wisnu Usdek, and Ambarwati, Titiek. (2021) "Analysis of Pay Later Payment System
on Online Shopping in Indonesia." Journal of Economics, Business, & Accountancy Ventura 23 (3): 329–339.
[2] Adirinekso, Gidion P., and Assa, Adrie F. (2021) "Is There Any Customer Switching the Intention from Traveloka PayLater to Traveloka
 Nicholas Dominic et al. / Procedia Computer Science 219 (2023) 60–67 67
8 Nicholas Dominic et al. / Procedia Computer Science 00 (2022) 000–000

PayLater Card? A Preliminary Investigation on Worker in Jakarta." Proceedings of the Ninth International Conference on Entrepreneurship
and Business Management (ICEBM 2020) 174 (ICEBM 2020): 570–577.
[3] Adirinekso, Gidion P., Purba, John Tampil, and Budiono, Sidik. (2021) "Worker switching intention from pay later apps to card: An HCM
approach of traveloka customers in Jakarta." Proc Int Conf Ind Eng Oper Manag 5 (8): 1348–1359.
[4] Abdillah, Leon A. (2020) "FinTech E-Commerce Payment Application User Experience Analysis during COVID-19 Pandemic." Scientific
Journal of Informatics 7 (2): 265–278.
[5] Hilmi, Luqman Dzul, and Pratika, Yeyen. (2021) "PayLater Feature: Impulsive Buying Driver for E-commerce in Indonesia." International
Journal of Economics, Business and Accounting Research (IJEBAR) 5 (2): 63–74.
[6] Chernovita, Hanna Prillysca. (2020) "The Role of Financial Literacy to Suppress Temptation of Using PayLater Service." ISCAIE 2020 -
IEEE 10th Symposium on Computer Applications and Industrial Electronics : 113–117.
[7] Gerrans, Paul, Baur, Dirk G., and Lavagna-Slater, Shane. (2021) "Fintech and responsibility: Buy-now-pay-later arrangements." Aust J
Manag. https://doi.org/10.1177/03128962211032448.
[8] Cenggoro, Tjeng Wawan, Wirastari, Raditya Ayu, Rudianto, Edy, Mohadi, Mochamad Ilham, Ratj, Dinne, and Pardamean, Bens. (2021)
"Deep Learning as a Vector Embedding Model for Customer Churn." Procedia Computer Science 179 (2019): 624–631.
[9] Alcazar, By Julian, and Bradford, Terri. (2021) "The Rise of Buy Now, Pay Later: Bank and Payment Network Perspectives and Regulatory
Considerations." Payment System Research Briefing, Federal Reserve Bank of Kansas City : 1–6.
[10] Gio, Prana Ugiana, Caraka, Rezzy Eko, Rosmaini, Elly, Syahputra, Rizki, Lubis, Irna Triannur, Siregar, Devidayanty, Shara, Yuni, and
Pardamean, Bens. (2019) "Financial Data Statistics Programs." 1st 2018 Indonesian Association for Pattern Recognition International
Conference, INAPR 2018 - Proceedings (September): 333–337.
[11] Mcmahan, Christopher, Baurley, James, Bridges, William, Joyner, Chase, Kacamarga, Muhamad Fitra, Lund, Robert, Pardamean, Carissa,
and Pardamean, Bens. (2017) "A Bayesian hierarchical model for identifying significant polygenic effects while controlling for confounding
and repeated measures." Statistical Applications in Genetics and Molecular Biology 16 (5–6): 407–419.
[12] Joyner, Chase, McMahan, Christopher, Baurley, James, and Pardamean, Bens. (2020) "A two-phase Bayesian methodology for the analysis
of binary phenotypes in genome-wide association studies." Biometrical Journal 62 (1): 191–201.
[13] Render, Barry, Stair, Ralph M., Hanna, Michael E., and Hale, Trevor S. (2018) "Chapter 3: Decision Analysis." Pearson Education Limited
13th ed: 81–93.
[14] Atmojo, Robertus Nugroho Perwiro, Cahyani, Anggita Dian, Abbas, Bahtiar Saleh, Pardamean, Bens, Anindito, and Manulang, Imanuel
Didimus. (2014) "Design of single user decision support system model based on fuzzy simple additive weighting algorithm to reduce
consumer confusion problems in smartphone purchases." Applied Mathematical Sciences 8 (13–16): 717–732.
[15] Atmojo, Robertus Nugroho Perwiro, Anindito, Pardamean, Bens, Abbas, Bahtiar Saleh, Cahyani, Anggita Dian, and Manulang, Imanuel
Didimus. (2014) "Fuzzy simple additive weighting based, decision support system application for alternative confusion reduction strategy in
smartphone purchases." American Journal of Applied Sciences 11 (4): 666–680.
[16] Budiarto, Arif, Pardamean, Bens, and Caraka, Rezzy Eko. (2018) "Computer vision-based visitor study as a decision support system for
museum." Proceedings - 2017 International Conference on Innovative and Creative Information Technology: Computational Intelligence
and IoT, ICITech 2017 2018-Jan: 1–6.
[17] Gaspars-Wieloch, Helena. (2021) "On some analogies between one-criterion decision making under uncertainty and multi-criteria decision
making under certainty." Economics and Business Review 7 (2): 17–36.
[18] Langholtz, Harvey, Gettys, Charles, and Foote, Bobbie. (1993) "Resource-allocation behavior under certainty, risk, and uncertainty."
Organizational Behavior and Human Decision Processes 54 (2): 203–224.
[19] Silva, Saulo Andre, Abreu, Pedro Henrique Camargo De, Amorim, Fernando Rodrigues De, and Santos, David Ferreira Lopes. (2019)
"Application of Monte Carlo Simulation for Analysis of Costs and Economic Risks in a Banking Agency." IEEE Latin America Transactions
17 (3): 409–417.
[20] Konstantinos, J., and Sotirios, J. (2019) "The control of small medium banks profitability using financial modeling approach under certainty
and uncertainty." Financial Studies 23 (4(86)): 52–84.
[21] Trigkas, Sotirios, Liapis, Konstantinos, and Thalassinos, Eleftherios. (2021) "Administrative Accounting Information to Control Profitability
Under Certainty and Uncertainty of a Universal Bank." Contributions to Management Science : 53–78.
[22] Fajartama, Rian. (2021) "Modern Portfolio Theory and Forecast Implementation Using Monte Carlo Simulation in Automotive Financing
Company." Review of Integrative Business and Economics Research 10 (3): 2021.
[23] Kapadia, Joy, Thirunavukkarasu, Kaniskaran, Gulecha, Kashish, Kotak, Krish, and Parmar, Krish. (2021) "Operations Research Techniques
Applied to the Financial Markets." International Journal of Innovative Science and Research Technology 6 (10): 281–290.
[24] Dominic, Nicholas, Prayoga, Jonathan Sebastian, Kumala, Daniel, Surantha, Nico, and Soewito, Benfano. (2021) "The Comparative Study
of Algorithms in Building the Green Mobile Cloud Computing Environment." Springer Book: Lecture Notes in Networks and Systems 343:
43–54.
[25] Fang, Qianqian, and Yan, Shijie. (2022) "MCX Cloud—a modern, scalable, high-performance and in-browser Monte Carlo simulation
platform with cloud computing." Journal of Biomedical Optics 27 (08): 1–14.
[26] Jamilatuzzahro, Caraka, Rezzy Eko, Herliansyah, Riki, Asmawati, S., Sari, Dian Megah, and Pardamean, Bens. (2018) "Generalized Space
Time Autoregressive of Chili Prices." Proceedings of 2018 International Conference on Information Management and Technology, ICIMTech
2018 : 291–296.