An Innovative Cost-Benefit Model for Analyzing Solution Architectures Considering Uncertainty

In the dynamic environment nowadays, companies require continuously to implement and improve their technology solutions to respond to new technology tendencies, business requirements, and enterprise architecture initiatives. This transformation is supported by solution architectures, which contain solution components that are implemented to respond to company’s requirements. As a consequence, the ability to analyze solution architectures financially becomes significantly essential to determine their costs, benefits, profitability, and the impact of the unpredictable events that can derive positive or negative consequences in technology investments. However, since most current financial models address overall IT investments, it is difficult to calculate the detailed and precise financial indicators required for solution architectures and their components. Consequently, financial variables can be omitted, and financial indicators can be limited and insufficient to analyze advantages, disadvantages, and risks associated with solution architectures and their components. With this in mind, our study presents an innovative cost-benefit model specifically for solution architectures. To define, design and evaluate our model, we adopted the Design Science Research paradigm. As a result of the model’s evaluation, we determined that our model calculates specific indicators related to alternatives and components which can reduce the risk omitting of variables while increasing flexibility for comparing alternatives and provision of support for technology investment decisions.

solution's design considers solution architectures from which are derived one or more alternatives that represent different options for the technology solution (Pena et al., 2017a). An alternative has a set of solution components that represent technology functionalities such as information systems (Arango Serna et al., 2015; Pena et al., 2017b), which in turn can have particular variables such as benefits and costs associated with them and that must be considered to calculate the alternative's financial indicators using cost-benefit models (Brealey et al., 2011;Irani et al., 2006;Pena et al., 2017a). Financial indicators are indispensable for determining the feasibility of the technology solution, analyzing its possible risks, and for justifying any investment (Irani et al., 2006;Neufville and Scholtes, 2011). These financial indicators must consider uncertainties derived from unpredictable events such as changes in requirements to identify positive or adverse consequences of IT investments (Neufville and Scholtes, 2011;Pena et al., 2017a).

Fig. 1: Solution architecture approach
Once alternatives are defined, a company has the crucial task of selecting the one that is most convenient for implementing the technology solution (Pena et al., 2017a). To choose an alternative, companies commonly compare the alternatives' financial indicators (Brealey et al., 2011;Pena et al., 2017a). Although comparison of alternatives is valuable, it is important to consider the financial indicators at the component level because they include particular financial variables, economic factors, and unpredictable events (Pena et al., 2017a) that support analyses such as comparison of the profitability between components of various alternative solutions (Pena et al., 2017a).
Although there is an extensive literature about cost-benefit models addressed to technology solutions (Apfel and Smith, 2003;Cormier, 2010;Kazman et al., 2002), most of these models are directed at overall IT investments and have a high level of granularity i (Mead et al., 2009;Pena et al., 2017b) that limits the detailed cost-benefit analysis required for solution architectures. Consequently, financial indicators may not accurately indicate valuation due to the omission of specific variables (Mead et al., 2009;Pena et al., 2017b), it may be difficult to identify financial indicators at the component level which could limit comparisons of components (Pena et al., 2017a), and uncertainty at the component level could impact financial indicators. This study defines FINFLEX-CBMC, an innovative cost-benefit model that supports financial analysis of alternatives and solution components in solution architectures while considering uncertainty. To define and evaluate FINFLEX-CBMC, we adopted the Design Science Research paradigm (Hevner et al., 2004). The results from the evaluation of FINFLEX-CBMC indicate that it supports financial analysis of solution architectures and their alternatives and components, offers detailed data considering the uncertainty of financial variables, and supports the comparison between and among components and alternatives. This paper is structured as follows. Section 0 describes the theoretical background. Section 0 presents related work and the objective of this research. Section 0 describes the research methodology. Section Error! Reference source not found. describes FINFLEX-CBMC. Section 0 presents the evaluation of FINFLEX-CBMC, and Section 0 presents a discussion, future work, and conclusions from this research.

Theoretical Background
This section briefly describes the conceptual framework of FINFLEX-CBMC.

Financial variables
A solution component can incur specific costs (Irani et al., 2006), support materialization of different expected benefits (Irani et al., 2006), be associated with specific economic factors such as taxes and depreciation (Brealey et al., 2011), and be impacted by unpredictable events that imply different uncertainties (Neufville and Scholtes, 2011). Due to these conditions, we can identify the following financial variables at the component level (Pena et al., 2017a). a. Initial Investment consists of acquisition costs of assets required for a component at time zero (Irani et al., 2006). b. Benefits are consequences of an action that improve or promote the comfort of a company (Irani et al., 2006) and which are represented as quantified monetary profit in this study (Pena et al., 2017a). c. Costs consist of the purchase or rental price of a service or asset implied in a solution component (Irani et al., 2006;Smit, 2012) and are represented as a quantified monetary value in this study (Pena et al., 2017a). d.
Economic factors are relevant market and economic data that can influence the value of an IT investment (Irani et al., 2006;Pena et al., 2017a). This study includes taxes, the inflation rate, depreciation, and salvage or rescue value.

Unpredictable Events
Unpredictable events such as changes in providers and new technologies (Vélez Pareja, 2003) create uncertainty levels that must be considered in the variables of the cost-benefit analysis to determine whether the consequences for a company will be positive or adverse ( • Estimating the ranges (low, more probable, very probable) of the cost and benefit variables using expert criteria, subjective probability technique, or historical data (Hubbard, 2010;Vélez Pareja, 2003). • Quantifying the uncertainty using probability distributions (Cormier, 2010;Hubbard, 2010 • Measuring the uncertainty supported by Monte Carlo simulations (Hubbard, 2010).

Other Factors
To perform the cost-benefit analysis, the life cycle of a solution component must be considered and therefore costs and benefits must be forecast (Pena et al., 2017a;Smit, 2012). The life cycle comprises activities from conception (tc) until retirement (td) where duration of the lifecycle is given by td -tc ( Fig. 2) (Smit, 2012).

Methods of Cost-benefit Analysis
Cost-benefit analysis assesses the effectiveness of budgeted investment as well as the efficiency and profitability of a technology solution (Brealey et

Related Work
In this subsection, we briefly describe the cost-benefit models we identified, comparisons between and among them, and the objective of this research.

Description of the Cost-Benefit Models Identified
A cost-benefit model is a mathematical representation that comprises one or several equations that are used to analyze how a business reacts to different economic situations and to calculate the outcomes of financial decisions before investments are made (Brealey et al., 2011). We identified the following costbenefit models in the literature: decisions that result in a system's quality attributes (Kazman et al., 2002).
Total Economic Impact (TEI) is a model that aims to identify cost supported on TCO, determine the benefit related to the business value and strategic contribution, and analyze uncertainty considering the Black-Scholes model (Cormier, 2010;Mead et al., 2009).
Rapid Economic Justification (REJ) provides a pragmatic expression of the economic justification for an IT investment considering the alignment among its associated costs, benefits, and risks (Mead et al., 2009).

An Integrated Real Options Framework for Model-based Identification and Valuation of
Options under Uncertainty (Mikaelian, 2009) by Mikaelian presents an integrated real options framework (IRF) for holistic analysis of enterprise architecture projects (Mikaelian, 2009). IRF allows the evaluation of enterprise architecture alternatives through modeling of uncertainty, flexibility, benefits, and costs.
A method for valuing Architecture based Business Transformation and measuring the value of Solutions Architecture by Slot (Slot, 2010) presents a method to quantify the values of enterprise architecture that is financially based on business transformation (Slot, 2010). The method considers costs, benefits, and management of uncertainty based on real options (Slot, 2010). Table 1 compares the cost-benefit models identified.  (Slot, 2010) and was designed considering the methodology proposed by Avon (Avon, 2013). We followed the DSR evaluation patterns proposed by Sonnenberg and Brocke (Sonnenberg and Brocke, 2012), and we adapted those patterns to FINFLEX-CBMC as shown in Fig. 4. The use of patterns allowed us to demonstrate, design and evaluate FINFLEX-CBMC. We explain below the three evaluation activities used for FINFLEX-CBMC.  To execute EVAL3, we first defined the cost-benefit model (Sections Error! Reference source not found.). Then, we created a prototype of FINFLEX-CBMC to capture the required data, perform the cost-benefit analysis and present the results (Section 0).

An Innovative Cost-Benefit Model for Analyzing Solution Architectures Considering Uncertainty
To describe FINFLEX-CBMC, we used the generic process presented in Fig. 5. It includes the essential activities required for calculating the financial indicators of an alternative and its solution components.

Identifying the alternatives and their components
This activity defines the solution architecture alternatives of a technology solution and identifies the solution components of an alternative.

Calculating the financial indicators by component
The process considers the activities explained below in order to calculate the financial indicators of a component.

Fig. 6: Inputs required for FINFLEX-CBMC
As Fig. 6 presents, inputs have two sources: alternatives and components. The alternatives' inputs are used to perform the cost-benefit analysis for all of the alternatives' components while the components' inputs are particular to each component.

Input identification considers two steps:
Step Step 2) Identifying inputs of component i a. Lifecycle. Determine (Section 0). b. Initial investment (II). Identify the II (Section 0). c. Benefit. Determine the benefit (Section 0) and its profile (Section 0).

Performance Of Cost-Benefit Analysis
To perform the cost-benefit analysis of a component, FINFLEX-CBMC executes the following activities:  Fig. 7 and explained below

Fig. 8: Approach to calculation of NPV (Brealey et al., 2011)
Utility before taxes is calculated as: 1,2,…,n is the period in which variables are considered.
Next, the net profit is calculated as: Then, the cash flows are calculated as: Finally, the NPV is calculated as: Once the 200,000 simulations have been run by FINFLEX-CBMC, the ENPV is represented as shown in the example illustrated in Fig. 9.  3. Cash flow.

Evaluation of FINFLEX-CBMC
FINFLEX-CBMC was evaluated to justify its utility based on the scenarios method (Hevner et al., 2004). This method was selected because we evaluated FINFLEX-CBMC on the basis of six scenarios that were performed by 32 IT professionals distributed into six teams. EVAL3 was conducted between August 2016 and December 2016 as part of the Solution Architecture course offered by the master's degree in Architecture of Information Technology of Los Andes University (https://sistemas.uniandes.edu.co/en/mati -en). Each scenario represented between one and three solution architecture alternatives that responded to the requirements of a hypothetical company called CCT. CCT offers technology services based on Business Process Outsourcing within Colombia and internationally. CCT currently has difficulty managing customer data related to information traceability and consolidation. Consequent high levels of customer dissatisfaction have increased risks of losing customers and sales opportunities. Error! Reference source not found. presents an example of one alternative defined by one of the teams.
The evaluation methodology is presented in Error! Reference source not found. Due to space limitation, we have only briefly described the quantitative analysis. As Fig. 13 shows, we first designed a prototype and a guide for FINFLEX-CBMC. Then, we identified research questions to determine how FINFLEX-CBMC supports financial analysis of solution architectures. Next, we presented the prototype and the guide to the six teams in a face-to-face session. Subsequently, the teams evaluated their alternatives using FINFLEX-CBMC. Finally, we collected and analyzed the instances of the prototype with scenarios defined by the teams (A total of 12 solution architecture alternatives because some of the teams only defined one alternative for each scenario).
The objective of the quantitative analysis was to compare traditional financial indicators with the financial indicators generated by FINFLEX-CBMC. Specifically, we compared NPV and ENPV and the cash flows of these approaches.   It can be seen from the results that, while the NPV offers a unique value for identification of the profitability of a solution, the ENPV is a rich source of data for analyzing risks and possibilities of profitability.
For example, in scenario 1, the NPV is $ 41,630.50 which could be a good basis for making a decision since it is positive. However, the NPV's value is lower than the mean (μ) of ENPV which is $204,031.30 and, since the standard deviation (σ) of ENPV is $130,437.38, profitability is likely to be nearly 50% higher than that indicated by NPV alone. ENPV can give experts more information for each percentile plus the standard deviation measures possible risks of an investment. It is also interesting to analyze the minimum and maximum profits of a solution. For example in scenario 2, the minimum profit is -$322.467, the maximum profit is $298.911, and the average profit is -$70.840. In this scenario, experts can identify the maximum and minimum possible profits as well as the risks thereby providing more robust information to support investment decisions.
The scenarios reveal some differences regarding cash flows between the traditional NPV and the ENPV which could affect the identification of outflows and inflows that in turn could result in notable differences regarding investment in a particular solution. The ENPV's cash flows are the results of simulations which may provide more certainty to the analysis.
Finally, on the basis of the results of the quantitative evaluation, we concluded that FINFLEX-CBMC offers more detailed and richer financial indicators than does the traditional NPV for making investment decisions regarding solution architectures.

Discussion and Future Work
This paper has presented FINFLEX-CBMC, a cost-benefit model that calculates financial indicators of solution architectures on the basis of their components and alternatives. FINFLEX-CBMC contributes to the determination of unpredictable events and measurement of uncertainties to determine positive and adverse events related to an investment. FINFLEX-CBMC has a number of advantages: • It calculates financial indicators at two levels of detail for solution architectures: alternatives and components.
• It permits comparison of financial indicators for components and alternatives and measures uncertainty associated with financial variables which can increase the accuracy of the indicators calculated.
• It consolidates financial variables and financial indicators related to solution architectures. FINFLEX-CBMC's flexibility allows performance of sensitivity analysis on the basis of the variation of the financial variables, and it supports risk analysis of IT investments on the basis of ENPV and its metrics.
Although FINFLEX-CBMC provides a complete set of calculations and supports a cost-benefit analysis of solution architectures, it does have some limitations. These include the amount of data required for the use of FINFLEX-CBMC, the performance of the FINFLEX-CBM's prototype, and absence of the experience of the architects in the interpretation of financial indicators of stochastic models like FINFLEX-CBMC. Consequently, results could be omitted or misunderstood by an interdisciplinary team.
We are considering the following activities to continue the development of FINFLEX-CBMC: 1. Implementation of the prototype on another platform that improves performance and user experience.
2. Definition of a methodology to support FINFLEX-CBMC that will offset the   ii AIE is a rigorous quantitative methodology that presents an explicit method for measuring uncertainty at low levels of granularity and which supports more accurate valuation (Mead et al., 2009).
iii To calculate the number of simulations, we considered a 95% confidence level, a probability of failure of 0.001, and a 15% error of estimation to provide us with a high level of confidence in the results (Missouri University of Science and Technology, 2016).