Title Science and Technology Indonesia e-ISSN:2580-4391 p-ISSN:2580-4405 Vol. 6, No. 4, October 2021 Research Paper Analysis of Information Service Pricing Scheme Model Based on Customer Self-Selection Indrawati1, Fitri Maya Puspita1*, Resmadona1, Evi Yuliza1, Oki Dwipurwani1, Sisca Octarina1 1Mathematics Department, Faculty of Mathematics and Natural Sciences, Sriwijaya University, Palembang, 30662, Indonesia *Corresponding author: fitrimayapuspita@unsri.ac.id AbstractThis study attempts to analyze pricing schemes with monitoring cost and marginal cost for perfect substitute and quasi-linear utilityfunctions for achieving Internet service Provider (ISP) in gaining benefit. Two types of customers analyzed, namely customers whoare heterogeneous (both high-end and low-end) as well as heterogeneous customers (high-demand and low-demand) based onFlat-fee, usage-based, and two-part tariff are the three types of pricing methods employed. The results show that usage-basedpricing schemes gain maximum profit optimal for heterogeneous customers (high-end and low-end), while for heterogeneouscustomers (high-demand and low-demand) type of pricing scheme two-part tariff obtains maximum profit optimal. The results ofthis study are more directed to the lemma of the perfect substitute utility function which compares the lemma of heterogeneouscustomers. This model was solved using LINGO 13.0 software and ISP to get maximum profit. KeywordsUtility Function, Perfect Substitute, Marginal Cost, Monitoring Cost, Pricing Schemes, Heterogeneous Customers Received: 25 July 2021, Accepted: 14 October 2021 https://doi.org/10.26554/sti.2021.6.4.337-343 1. INTRODUCTION Internet is a tool for technology that can be used to reach the information needs of its customers. Internet service providers or ISPs try to share the best capacity to internet users and to achieve the highest revenue (Indrawati et al., 2015). Bundling is a strategy carried out by combining two or more specic products into a sales package (Gu et al., 2011; Kopczewski et al., 2018; Yassine et al., 2018; Ye et al., 2017). The utility function Sitepu et al. (2016) is usually related to the satisfaction level that users receive compensation for the use of information services, especially those related to maximizing prots in achieving certain goals and namely as A = i(m1, m2, . . . , mn̂) which means that m1, m2, . . . , mn̂ con- tribute user utility (Kuo and Liao, 2007; Merayo et al., 2017) indicating goal-satisfaction. Further research on current inter- net pricing schemes has involved other utility functions that are often used such as the original Cobb-Douglas utility func- tion (Puspita et al., 2020b; Sitepu et al., 2017b), quasi-linear, perfect substitute (Sitepu et al., 2017a), and bandwidth func- tion (Guan et al., 2008; Indrawati et al., 2015; Zu-Xin et al., 2009; Moriya et al., 2005) utilized in three types of infor- mation service pricing systems, namely at-fee, usage-based and two-part tari (Gizelis and Vergados, 2010; Puspita et al., 2020a; Puspita et al., 2021) both analytically and as MINLP (Mixed Integer Nonlinear Programming) (Barrios and Cruz, 2017; Giraldo, 2017) with the help of LINGO (Cunningham and Schrage, 2004; Schrage, 2009) application software. The perfect substituteutilityfunctionwasone importantutilityfunc- tion to be selected to measure the satisfaction of the customers due to its linearity. The utility function is measurements of customer satisfaction indirectly (Hitt and Chen, 2005). So far, past research focus on the pricing of information pricing schemes have been conducted (Indrawati et al., 2014; Sitepu et al., 2017b; Sitepu et al., 2017a; Wu and Banker, 2010) and also with the added parameters such as marginal dan monitoring costs. However, this research only focus on the pricing for information services without considering customer self-selection (Rabbani et al., 2017). Customer self-selection is based on packages or schemes (Varadarajan, 2020) to be oered to various customers (Rabbani et al., 2017). In this recent situation, customer self-selection through products or servicesoered iscriticalandneeds tobedeveloped, so it causes a gap that has to be explored more. The research needs to be critically explained in detail to show the relationship between the pricing scheme of information service and the ability of customers to select its service (Zhou et al., 2020). Then, our contribution will be exploring new formulations through lemmas to show which pricing methods provide the https://crossmark.crossref.org/dialog/?doi=10.26554/sti.2021.6.4.337-343&domain=pdf https://doi.org/10.26554/sti.2021.6.4.337-343 Indrawati et. al. Science and Technology Indonesia, 6 (2021) 337-343 most eective models that can be adopted by customers based on his/her preferences (Li et al., 2013). The customers can choose theschemes, due to itsheterogeneity(Caiati etal.,2020; Kopczewski et al., 2018; Zhang et al., 2018). 2. METHOD In this study, the customer self-model that the model was de- signed to have the eligibility of a customer to choose his/her preference pricing schemes tted with his/her needs and bud- get, was compiled and validated using a local data server with a perfect substitute utility function, then the optimal results could be compared. The steps are listed as follows: 1. Modeling a pricing structure for the information ser- vices based on a quasi-linear utility function with at-fee, usage-based, and two-part tari pricing types for hetero- geneous customer problems. a) For service pricing schemes on at-fee, b) For service pricing schemes on usage-based, c) For service pricing schemes on two-part tari 2. Applying the scheme for optimal pricing based on data on the local server in the form of trac data. Processing this data from a local server in the form of trac data on the LPSE application, concerning with tool equipment available for the institution. 3. Validating a quasi-linear utility function for diverse cus- tomer types based on three types of pricing schemes: at-fee, usage-based, and two-part tari, with the addi- tion of marginal expenses and monitoring costs. 4. Comparingthepricingschememodelsobtainedfromthe analysis in Step 3 to obtain the optimal pricing scheme for each type of customer, namely high-end and low- end and high-demand and low-demand heterogeneous customers. 5. Make conclusions and get the best information service pricing solutions. 3. RESULTS AND DISCUSSION 3.1 Perfect Substitute Utility Functions for High-end and Low-end Customers In this instance, the perfect substitute utility function with the form A(m, n) = xm + yn. Suppose there are high-end customers ( j = 1) and low-end customers ( j = 2) where 𝛼1 + 𝛼2 = 1 Customer Problem Optimization: Max M , N , O Op = xM+yN−RmM−RnN−RO−(M+N)k (1) with constraints: M ≤ M̄O (2) N ≤ N̄O (3) xM + yN − RmM − RnN − RO − (M + N)k ≤ 0 (4) O = 0 or 1 (5) Where Rm is ISP’ xed price in peak hour, Rn is ISP’ xed price in o peak hour, O is the customer decision to join the schemes or not, k is marginal and monitoring cost, M is the largest amount of data consumed during peak hours, measured in kilobytes and N is the maximum level of usage in kilobytes during o-peak hours. Service Provider Problem Optimization: Max R, RM , RN ∑︁ j (RmM∗ + RnN∗ + RO∗) (6) where (M∗, N∗, O∗) = argmax xM +yN −RmM −RnN −RO− (M + N)k subject to Equation (2)-(5). where R is customer fees for joining the service. The Objective Function (6) is used to maximize user over- load according to the price set by the service provider. This model does not consider the initial cost for customers to join. But through This model can consider the long-term relation- ship between providers services and customers but does not charge for the short term or a certain period of time. O=0 or 1 is determined by the costumer, where if the customer chooses not to join the program then will be 0 so that M ≤ M̄O and M ≤ M̄O will be worth 0. Meanwhile, if the customer chooses to join the program, then O will be worth. The value of M and N can not exceed the limit of M̄ and N̄. For usage-based and two-part tari pricing schemes: Customer Problem Optimization: Max M , N , O Op = xM+yN−RmM−RnN−RO−(k+l)M−(k+l)N (7) Subject to Equation (2)-(3), and (5), also xM +yN −RmM −RnN −RO− (k+l)M − (k+l)N ≥ 0 (8) Service Provider Problem Optimization: Max R, RM , RN ∑︁ j (RmM∗ + RnN∗ + RO∗) (9) where (M∗, N∗, O∗) = argmax xM +yN −RmM −RnN −RO− (k + l)M − (k + l)N subject to Equation (2), (3), (5), and (8). Case 1: Flat-fee pricing scheme for heterogeneous customers (high-end and low-end) based on a perfect substitute utility function with marginal costs and monitoring costs MaxR𝛼1(RO∗1) + 𝛼2(RO ∗ 2) = 𝛼1(x2M̄ + y2N̄ − (M + N)k) + 𝛼2 (x2M̄ + y2N̄ − (M + N)k) = (𝛼1 + 𝛼2)(x2M̄ + y2N̄− (M + N)k) © 2021 The Authors. Page 338 of 343 Indrawati et. al. Science and Technology Indonesia, 6 (2021) 337-343 The maximum benets for the service provider will be (𝛼1 + 𝛼2)(x2M̄ + y2N̄ − (M + N)k); where m is the number of high- end customers and n is the numberof low-end customers. This analysis is summed up in Lemma 1. Lemma 1: If the ISP uses a at-fee pricing scheme, the price charged to customers will be x2M̄ + y2N̄ − (M + N)k and the maximum prot earned is (𝛼1 + 𝛼2)(x2M̄ + y2N̄ − (M + N)k). Case 2: A usage-based pricing is method based on a perfect substitute utility function for heterogeneous users (high-end and low-end) with marginal costs and monitoring costs. If the ISP chooses to use a usage-based pricing scheme, then it is set to Rm > 0, Rn > 0, and R = 0, so that the optimization of customer problems becomes: Max Mj , Nj , Oj Op =xjMj + yjNj − RmMj − RnNj − R(0) − (k + l) Mj − (k + l)Nj Max Mj , Nj , Oj Op = xjMj+yjNj−RmMj−RnNj−(k+l)Mj−(k+l)Nj Optimization of high-end heterogeneous customer problems: To maximize Equation (6), a dierentiation is made to the low- end heterogeneous customer optimization problem as follows. To maximize the Equation, a dierentiation is made to M1 and N1; under the conditions of 𝜕Op 𝜕M1 = 0 and 𝜕Op 𝜕N1 = 0 ↔ 𝜕(x1M1 + y1N1 − RmM1 − RnN1 − (k + l)M1 − (k + l)N1) 𝜕M1 = 0 ↔ x1(k + l) = Rm and ↔ 𝜕(x1M1 + y1N1 − RmM1 − RnN1 − (k + l)M1 − (k + l)N1) 𝜕N1 = 0 ↔ y1(k + l) = Rm Service Provider Problem Optimization becomes: Max R, Rm, Rn 𝛼1(RMM∗1 + RNN ∗ 1) + 𝛼2(RMM ∗ 2 + RNN ∗ 2) = Max Rm, Rn 𝛼1((x1 − (k + l))M∗1 + (y1 − (k + l))N ∗ 1) +𝛼2((x2 − (k + l))M∗2 + (y2 − (k + l))N ∗ 2) = Max Rm, Rn 𝛼1(x1M∗1 + y1N ∗ 1 − (k + l)M ∗ 1 − (k + l)N ∗ 1) +𝛼1(x2M∗2 + y2N ∗ 2 − (k + l)M ∗ 2 − (k + l)N ∗ 2) In order to maximize the optimization equation for the producer problem, the ISP must minimize the value of Rm and Rn. Since M1, M2, N1, and N2 are constrained, M∗1 , M ∗ 2 , N ∗ 1, and N∗2 cannot exceed M̄ and N̄.To nd the maximum cost, analysis during peak hours has been done. This analysis is ap- plied to problems during peak and o-peak hours. Specically, Rm and Rn will be Rm = x2 − (k + l) and Rn = y2 − (k + l) with the maximum prot obtained is: (𝛼1 + 𝛼2)(x2M̄ + y2N̄ − (k + l)M̄ − (k + l)N̄). Based on this case, the following Lemma 2 is obtained. Lemma 2: If ISP utilizes the usage-based pricing scheme, the optimal prices will be Rm = x2 − (k + l) and Rn = y2 − (k + l) with the maximum prot gained is (𝛼1 + 𝛼2)(x2M̄ +y2N̄ − (k + l)M̄ − (k + l)N̄). Case 3: Two-part tari pricing scheme for heterogeneous cus- tomers (high-end and low-end) based on a perfect substitute utility function with marginal costs and monitoring costs. If ISP chooses to select a two-part tari scheme then it will be set Rm > 0, Rn > 0, and R > 0. Then, RM will be obtained that ranges between x1 − (k + l) and x2 − (k + l) or x2 − (k + l) ≤ Rm ≤ x1 − (k + l) and so does Rn. The best price for Rm should be between x1 and x2. When the prices are within this interval thenthehigh-endcustomerdemandswillbexedatM̄ andlow- end customer demands will be proportional to a price decrease. Meaning, Rm and Rn will be Rm = x2 − (k+ l), Rn = y2 − (k+ l), and R = 0. Assume that x1 < 𝛼1+𝛼2 𝛼1 x2 and y1 < 𝛼1+𝛼2 𝛼1 y2. Then, the ISP problem will be: Max Px , Py 𝛼1(RMM∗1 + RNN ∗ 1 + RO ∗ 1) + 𝛼2(RMM ∗ 2 + RNN ∗ 2 + RO ∗ 2) 𝛼1((x2 − (k + l))M∗1 + (y2 − (k + l))N ∗ + 0) +𝛼2((x2 − (k + l))M∗2 + (y2 − (k + l))N ∗ + 0) 𝛼1(x2M∗ − (k + l)M∗ + y2N∗ − (k + l)N∗) +𝛼2(x2M∗ − (k + l)M∗ + y2N∗ − (k + l)N∗) = (𝛼1 + 𝛼2)(x2M∗ + y2N∗ − (k + l)M∗ − (k + l)N∗) The maximum prot will be (𝛼1 + 𝛼2)(x2M̄ + y2N̄ − (k + l)M̄ − (k + l)N̄) Then, this analysis is to be summarized in Lemma 3. Lemma 3: If ISP uses a two-part tari scheme, the optimal price will beRm = x2−(k+l), Pn = y2−(k+l), andR=0 with the maximum prot of (𝛼1 +𝛼2)(x2M̄ +y2N̄ − (k+l)M̄ − (k+l)N̄). 3.2 Perfect Substitute Utility function for High-demand and Low-demand Heterogeneous Customers With a maximum degree of usage, it is assumed that there are two sorts of customers: high-demand customers (type 1) and low-demand customers (type 2) M̄1, and N̄1 for (type 1) and M̄2 and N̄2 for (type 2) where M̄1>M̄2 and N̄1>N̄2. Suppose there are m high-demand customers and n low-demand cus- tomers with x1 = x2 = x and y1 = y2 = y. Then, determination of the maximum prot on each pricingscheme used bythe ISP. Case 4: Flat-fee pricing scheme for heterogeneous customers (high-demand and low-demand) based on a perfect substitute utility function with marginal costs and monitoring costs. If using a at-fee pricing scheme, it is determined that Rm = 0, Rn = 0, and R > 0. This means that if customers © 2021 The Authors. Page 339 of 343 Indrawati et. al. Science and Technology Indonesia, 6 (2021) 337-343 choose to join the given program, then the maximum level of satisfaction is obtained by choosing the level of consumption with the maximum level of satisfaction obtained will be M1 = M̄1, N1 = N̄1, or M2 = M̄2, N2 = N̄2xM̄1 + yN̄1 − (M̄1 + N̄1)k or xM2 + yN̄2 − (M̄2 + N̄2)k. So, ISP cannot charge more than xM̄1 + yN̄1 − (M̄1 + M̄1)k to every high-demand customer and every low-demand customer aM̄2 + bN̄2 − (M̄2 + N̄2)k. By using a at-fee pricing scheme, ISPs cannot distinguish pricing between high-demand customers and low-demand customers, so ISPs must choose to charge a fee of xM̄1 +yN̄1 − (M̄1 +N̄1)k and only high-demand customers can use the service or charge a fee of xM̄2+yN̄2−(M̄2+N̄2)k where high-demand customers and low-demand customers can join the given program. If it is assumed that 𝛼1[xM̄1 +yN̄1 − (M̄1 +N̄1)k] < (𝛼1 + 𝛼2)[xM̄2 + yN̄2−(M̄2+N̄2)k], thenthebestprice thatcanbechargedbythe ISP is xM̄2 + yN̄2 − (xM̄2 + yN̄2)k for high-demand customers and low-demand customers. So, the maximum prot that the ISP obtained is: (𝛼1 + 𝛼2)(xM̄2 + yN̄2 − (M̄2 + N̄2)k) Based on the analysis, the following Lemma 4 is obtained. Lemma 4: If the ISP uses a at-fee pricing scheme, the fee paid becomes P = xM̄2 +yN̄2 − (M̄2 +bN̄2)k and the maximum prot obtained is: (𝛼1 + 𝛼2[xM̄2 + yN̄2 − (M̄2 + N̄2)k] Case 5: Based on a perfect substitute utility function, a usage- based pricing method for heterogeneous customers (high-de- mand and low-demand) is proposed with marginal costs and monitoring costs. If the ISP chooses to use a usage-based pricing scheme, then it is determined that Rm > 0, Rn > 0, and R = 0. For optimization of the high-demand/low-demand customer problem results in: Max M , N , O Op = xMj+yNj−RmMj−RnNj−(k+l)Mj−(k+l)Nj Optimization of high-demand heterogeneous customer prob- lems: To optimize Equation (7), dierentiate toward M1 and N1; with conditions as following. 𝜕Op 𝜕M1 = 0 and 𝜕Op 𝜕N1 = 0 ↔ 𝜕(xM1 + yN1 − RmM1 − RnN1 − (k + l)M1 − (k + l)N1) 𝜕M1 = 0 ↔ x(k + l) = Rm and ↔ 𝜕(xM1 + yN1 − RmM1 − RnN1 − (k + l)M1 − (k + l)N1) 𝜕N1 = 0 ↔ y(k + l) = Rn Low-demand heterogeneous customer problem optimization: In order to optimize Equation (7), a dierentiation is made to M2 and N2; with conditions 𝜕Op 𝜕M2 = 0 and 𝜕Op 𝜕N2 = 0 ↔ 𝜕(xM2 + yN2 − RmM2 − RnN2 − (k + l)M2 − (k + 2)N2) 𝜕M2 = 0 ↔ x(k + l) = Rm and ↔ 𝜕(xM2 + yN2 − RmM2 − RnN2 − (k + l)M2 − (k + 2)N2) 𝜕N2 = 0 ↔ y(k + l) = Rn Then, the Service provider problem can be derived as follows: Max P , Px, Py 𝛼1(RmM∗1 + RnN ∗ 1) + 𝛼2(RmM ∗ 2 + RnN ∗ 2) = 𝛼1(xM∗1 + yN ∗ 1 − (k + l)M ∗ 1 − (k + l)N ∗ 1) + 𝛼2(xM ∗ 2 + yN ∗ 2 − (k + l)M∗2 − (k + l)N ∗ 2) If M1, M2, M1, and M2 are limited, then M∗1 , M ∗ 2 , N ∗ 1, and N∗2 will be M̄1, M̄2, N̄1, and N̄2. Then, Rm, and Ry will be Rm = x − (k + l), and Rn = y − (k + l) with the maximum prot of 𝛼1(xM̄1 + yN̄1 − (k + l)M̄1 − (k + l)N̄1) + 𝛼2(xM̄2 + yM̄2 − (k + l)M̄2 − (k + l)N̄2). This analysis is concluded in the following lemma. Lemma 5: If the ISP uses a usage-based pricing scheme, the optimal price is Rm = x − (k + l), and Rn = b − (k + l) with the maximum prot obtained is: 𝛼1(xM̄1 + yN̄1 − (k + l)M̄1 − (k + l)N̄1) + 𝛼2(xM̄2 + yN̄2 − (k + l)M̄2 − (k + l)N̄2) Case 6: Two-part tari pricing scheme for heterogeneous cus- tomers (high-demand and low-demand) based on a perfect substitute utility function with marginal costs and monitoring costs. If the ISP chooses to use a two-part tari pricing scheme, it will be determined that Rm > 0, Rn > 0, and R > 0. For the optimization process on the optimization of the problem of high-demand customers and low-demand customers, then xM̄j + yN̄j − RmM̄j − RnN̄j − R(1) − (k + l)M̄j − (k + l)N̄j ≥ 0 ↔ xM̄j + yN̄j − (x − (k + l))M̄j − (y − (k + l))N̄j − R − (k + l) M̄j − (k + l)N̄j ≥ 0 ↔ xN̄j + yN̄j − xM̄j + (k + l)M̄j − yN̄j + (k + l)N̄j − R − (k + l) M̄j − (k + l)N̄j ≥ 0 ↔ −R ≥ 0 ↔ R ≤ 0 Because P cannot be negative or R < 0, then R = 0. © 2021 The Authors. Page 340 of 343 Indrawati et. al. Science and Technology Indonesia, 6 (2021) 337-343 Table 1. Recapitulation of Pricing Schemes for Dierent Customer Types Customer Type Pricing Scheme Prot Heterogeneous: High-end Flat fee (𝛼1 + 𝛼2)(x2M̄ + y2N̄ − (M̄ + N̄)k & Low-end Usage-based (𝛼1 + 𝛼2)(x2M̄ + y2N̄ − (k + l)M̄ − (k + l)N̄)Two-part tari Heterogeneous: High-demand Flat fee (𝛼1 + 𝛼2)(xM̄2 + yN̄2 − (M̄2 + N̄2)k Usage-Based 𝛼1(xM̄1 + yN̄1 − (k + l)M̄1 − (k + l)N̄1)+ & Low-demand Two-part tari 𝛼2(xM̄2 + yN̄2 − (k + l)M̄2 − (k + l)N̄2) Therefore, the Service Provider problem will be Max R, Rm, Rn 𝛼1(RmM∗1 + RnM ∗ 1 + R) + 𝛼2(RmM ∗ 2 + RnN ∗ 2 + R) = Max R, Rm, Rn 𝛼1((x − (k + l))M∗1 + (y − (k + l))N ∗ 1 + R) +𝛼2((x − (k + l))M∗2 + (y − (k + l))N ∗ 2 + R) = Max R, Rm, Rn 𝛼1(xM∗1 + yN ∗ 1 − (k + l)M ∗ 1 − (k + l)N ∗ 1) +𝛼2(xM∗2 + yN ∗ 2 − (k + l)M ∗ 2 − (k + l)N ∗ 2) If M1, M2, N1 and N2 are constrained, then M∗1 , M ∗ 2 , N ∗ 1 and N∗2 will be M̄1, M̄2, N̄1 and N̄2. In other words, Rm = x − (k + l), Rn = y − (k + l), and R = 0 with the maximum prot obtained is: 𝛼1(xM∗1 + yN ∗ 1 − (k + l)M ∗ 1 − (k + l)N ∗ 1) + 𝛼2(xM ∗ 2 + yN ∗ 2 − (k + l)M∗2 − (k + l)N ∗ 2) So, the following lemma is summarized. Lemma 6 : If the ISP uses a two-part tari pricing scheme, the optimal price is Rm = x − (k + l), Rn = y − (k + l), and R = 0 with the maximum prot obtained is: 𝛼1(xM∗1 + yN ∗ 1 − (k + l)M ∗ 1 − (k + l)N ∗ 1) + 𝛼2(xM ∗ 2 + yN ∗ 2 − (k + l)M∗2 − (k + l)N ∗ 2) Table1displays themaximumprotgainedforeachscheme. By setting the parameter values assigned for each parameter as stated in Table 2, the values for each prot can be determined. The value of M̄ and N̄ is collected from the average demand of bandwidth consumption of LPSE trac data from a local server in one of the institutions in Palembang for one month at the beginning of March 2021. The data is not used to build the model, but the data is for the validation model only. The data value is used to show or to validate the value obtained from the formulation for each lemma. Table 2 depicts the values. Where M̄1 = M̄ is the largest amount of data consumed during peak hours, measured in kilobytes. M̄2 without obtaining data is the maximum consumption rate during peak hours X̄1, so M̄1>M̄2. Table 2. The Values of M̄ and N̄ from LPSE Trac Data Notation Ipse M̄1 or M̄ (kilobyte) 0.00185874 M̄2 (kilobyte) 0.061958 N̄1 or N̄(kilobyte) 0.00181696 N̄2 (kilobyte) 0.060565 N̄1 or N̄ is the maximum level of usage in kilobytes during o-peak hours. N̄2 is the greatest rate of consumption during o-peak hours, excluding data retrieval Ȳ1 , so N̄1>N̄2. Based on Table 3, the maximum prot obtained is in the at-fee pricingscheme, which is equal to (𝛼1+𝛼2 )(0.00185874 x2 + (0.00181696)y2 − 0.0036757k) for High-end and Low- end Heterogeneous customers. Meanwhile, for high-demand and low-demand heterogeneous customers, the maximum prot is by utilizing usage-based and two-part tari schemes, that are 𝛼1(0.061958x + 0.060565y − 0.128525k(k + l)) + 𝛼2(0.061958x + 0.060565y − 0.128525k(k + l)) 4. CONCLUSIONS This paper focuses on how to design a model based on cus- tomer self-selection on deciding which pricing scheme suit for heterogeneous customers. The data of the local server of LPSE is used to show the validation of the model that can be solved. Based on the results of the analysis and discussion, it can be concluded as follows. The best pricing scheme for heteroge- neous customers (high-end and low-end) is found using a at fee pricing scheme, while pricing strategies for heterogeneous customer (high-demand and low-demand) is obtained using a pricing scheme based on the perfect substitute utility func- tion. the most optimal was usage-based with the customers can be satised by selecting the schemes on his/her preferences. For further research, the concept of bundling strategy is likely to be considered in the schemes due to its advantages more customers to choose the service. © 2021 The Authors. Page 341 of 343 Indrawati et. al. Science and Technology Indonesia, 6 (2021) 337-343 Table 3. Recapitulation of Pricing Schemes for Dierent Customer Types Customer Type Pricing Scheme Prot Heterogeneous: High-end Flat fee (𝛼1+𝛼2 )(0.00185874x2 + (0.00181696)y2 − 0.0036757k) & Low-end Usage-based (𝛼1+𝛼2 )(0.00185874x2 + (0.00181696)y2 − 0.0036757(k + l))Two-part tari Heterogeneous: High-demand Flat fee (𝛼1 + 𝛼2)(0.061958x + 0.060565y − 0.128525k) Usage-Based 𝛼1(0.061958x + 0.060565y − 0.128525k(k + l))+ & Low-demand Two-part tari 𝛼2(0.061958x + 0.060565y − 0.128525k(k + l)) 5. ACKNOWLEDGEMENT On November 23, 2020, DIPA of the Public Service Agency of Universitas Sriwijaya 2021, SP DIPA-023.17.2.677515 /2021, supported the publishing of this article. On April 28, 2021, the Rector’s Decree Number: 0010/ UN9/ SK.LP2M. PT/2021 was issued. REFERENCES Barrios, P. and D. Cruz (2017). A mixed integer programming optimization of bundling and pricing strategies for multiple product components with inventory allocation considera- tions. International Conference on Industrial Engineering and Engineering Management; 16–20 Caiati, V., S. Rasouli, and H. Timmermans (2020). Bundling, pricing schemes and extra features preferences for mobility as a service: Sequential portfolio choice experiment. Trans- portation Research Part A: Policy and Practice, 131; 123–148 Cunningham, K. and L. Schrage (2004). The LINGO Algebraic Modeling Language. Springer Giraldo, O. D. M. (2017). Solución de un problema de op- timización clásico usando el paquete optimizador GAMS: implementación del problema de despacho económico. In- geniería y Ciencia, 13(26); 39–64 Gizelis, C. A. and D. D. Vergados (2010). A survey of pricing schemes in wireless networks. IEEE Communications Surveys and Tutorials, 13(1); 126–145 Gu, C., S. Zhuang, and Y. Sun (2011). Pricing Incentive Mech- anism based on Multi-stages Trac Classication Methodol- ogy for QoS-enabled Networks. Journal of Networks, 6(1); 163–171 Guan, Y., W. Yang, H. Owen, and D. M. Blough (2008). A pricing approach for bandwidth allocation in dierentiated service networks. Computers and Operations Research, 35(12); 3769–3786 Hitt, L. M. and P.-y. Chen (2005). Bundling with customer self-selection: A simple approach to bundling low-marginal- cost goods. Management Science, 51(10); 1481–1493 Indrawati, Irmeilyana, F. M. Puspita, and O. Sanjaya (2015). Internet pricing on bandwidth function diminished with increasing bandwidth utility function. Telecommunication Computing Electronics and Control, 13(1); 299 Indrawati, I., F. M. Puspita, and M. P. Lestari (2014). Cobb- Douglass UtilityFunction in Optimizing the Internet Pricing Scheme Model. Telecommunication Computing Electronics and Control, 12(1); 227–240 Kopczewski, T., M. Sobolewski, and I. Miernik (2018). Bundling or unbundling? Integrated simulation model of optimal pricing strategies. International Journal of Production Economics, 204; 328–345 Kuo, W.-H. and W. Liao (2007). Utility-based resource allo- cation in wireless networks. Transactions on Wireless Commu- nications, 6(10); 3600–3606 Li, M., H. Feng, F. Chen, and J. Kou (2013). Numerical investigation on mixed bundling and pricing of informa- tion products. International Journal of Production Economics, 144(2); 560–571 Merayo, N., P. Pavon-Marino, J. Aguado, R. Duran, F. Burrull, and V. Bueno-Delgado (2017). Fair bandwidth allocation algorithm for PONS based on network utility maximization. JournalofOpticalCommunicationsandNetworking, 9(1); 75–86 Moriya, T., H. Ohnishi, T. Ogawa, and T. Ito (2005). Method of improving bandwidth and connectivitywith multiple links in nomadic network environment. 6thAsia-Pacic Symposium on Information and Telecommunication Technologies; 41–46 Puspita, F. M., A. Wulandari, E. Yuliza, R. Sitepu, and Y. Yunita (2021). End-to-End Delay QoS Attribute-Based Bundling Strategy of Wireless Improved Reverse Charging Network Pricing Model. Science and Technology Indonesia, 6(1); 30–38 Puspita, F. M., A. Wulandari, E. Yuliza, R. Sitepu, et al. (2020a). ModicationofWirelessReverseChargingScheme with Bundling Optimization Issues. 3rdInternationalSeminar on Research of Information Technology and Intelligent Systems; 556–561 Puspita, F. M., E. Yuliza, W. Herlina, Y. Yunita, and R. Ro- hania (2020b). Improved Multi Service-Reverse Charging Models for the Multi Link Internet wireless Using QOS Bit Error Rate QoS Attribute. Science and Technology Indonesia, 5(1); 6–13 Rabbani, M., R. Salehi, and A. Farshbaf-Geranmayeh (2017). Integrating assortment selection, pricing and mixed- bundling problems for multiple retail categories under cross- selling. Uncertain Supply Chain Management, 5(4); 315–326 Schrage, L. (2009). Optimization Modeling with LINGO (6th © 2021 The Authors. Page 342 of 343 Indrawati et. al. Science and Technology Indonesia, 6 (2021) 337-343 ed.). LINDO Systems, Inc Sitepu, R., F. M. Puspita, and S. Apriliyani (2017a). Util- ity function based-mixed integer nonlinear programming (MINLP) problem model of information service pricing schemes. International Conference on Data and Software Engi- neering; 1–6 Sitepu, R., F. M. Puspita, A. N. Pratiwi, and I. P. Novyasti (2017b). Utility Function-based Pricing Strategies in Max- imizing the Information Service Provider’s Revenue with Marginal and Monitoring Costs. International Journal of Elec- trical and Computer Engineering (2088-8708), 7(2); 877–887 Sitepu, R., F. M. Puspita, H. Tanuji, and I. P. Novyasti (2016). Cobb-Douglas UtilityFunction Of Information Service Pric- ing Scheme Based On Monitoring And Marginal Costs. 2nd International Conference on Education, Technology and Science; 602 Varadarajan, R. (2020). Customer information resources ad- vantage, marketing strategy and business performance: A market resources based view. Industrial Marketing Manage- ment, 89; 89–97 Wu, S. y. and R. Banker (2010). Best pricing strategy for information services. Journal of the Association for Information Systems, 11(6); 339–366 Yassine, N., A. AlSagheer, and N. Azzam (2018). A bundling strategyfor itemswithdierentqualitybasedon functions in- volving the minimum of two random variables. International Journal of Engineering Business Management, 10; 1–9 Ye, L., H. Xie, W. Wu, and J. C. Lui (2017). Mining customer valuations to optimize product bundling strategy. IEEE International Conference on Data Mining; 555–564 Zhang, Z., X. Luo, C. K. Kwong, J. Tang, and Y. Yu (2018). Impacts of service uncertainty in bundling strategies on het- erogeneous consumers. Electronic Commerce Research and Applications, 28; 230–243 Zhou, Y., T. Zhang, Y. Mo, and G. Huang (2020). Willingness to pay for economy class seat selection: From a Chinese air consumer perspective. Research in Transportation Business and Management, 37; 100486 Zu-Xin, L., W. Wan-Liang, and X.-M. CHENG (2009). Op- timal bandwidth scheduling for resource-constrained net- works. Acta Automatica Sinica, 35(4); 443–448 © 2021 The Authors. Page 343 of 343 INTRODUCTION METHOD RESULTS AND DISCUSSION Perfect Substitute Utility Functions for High-end and Low-end Customers Perfect Substitute Utility function for High-demand and Low-demand Heterogeneous Customers CONCLUSIONS ACKNOWLEDGEMENT