Two-step model for performance evaluation and improvement of New Service Development process based on fuzzy logics and genetic algorithm

Abstract

The problem of assessment, selection and improvement of key performance indicators in the New Service Development process is one of the most important tasks of process managers, and it has a critical effect on the considered process effectiveness which is further propagated on the competitive advantage of each service small and medium enterprises. The relative importance of the introduced key performance indicators and their values are assessed by decision makers in selected enterprises (total of 187 persons). The assessment of decision makers are described by pre-defined linguistic expressions which are modelled by using fuzzy sets theory. Aggregated relative importance is determined according to approach developed in this paper. The ranking and improvement of key performance indicators is stated as multi-criteria decision making problem that could be solved by the genetic algorithm. Priority of management initiatives that should lead to the improvement of selected key performance indicator is based on fuzzy if-then rules and single-objective genetic algorithm. In this way, more appropriate improvement strategy, which demands lower costs, may be defined. By applying the proposed model it is possible to identify weak points in organizations, to provide corrective measures, and to enhance the effectiveness of new service development process. The model presents a suitable solution for reengineering and improvement of the process performance. The application of this model could be introduced in other industrial branches.

Keywords

Performance improvement fuzzy data genetic algorithm fuzzy logics

1 Introduction

Due to the rapid growth of service economy and increased number of customers, stakeholders, and regulatory requirements, service production has become much more complex. Generally, large complexity is generated by a low level of service quality, the lack of customer involvement, insufficient knowledge of customer desires and needs [1]. In such circumstances, decision makers have to incorporateNew Service Development (NSD) process with appropriate performance assessment and improvement measures [2]. So that new services or service procedures may affect the achievement of business effectiveness [3]. In order to achieve this, NSD process performance has to be assessed and improved by some performance improvement approach, which are presented in the literature [4, 5]. However, these approaches have been developed and based on the decision makers’ opinion, so they are mostly subjective and not mutually comparable. This gives an opportunity to define a new approach for NSDprocess performance assessment and improvement.

Motivation for the research is derived from the fact that there are no developed models for NSD process performance improvement presented in literature. Improvement of NSD process performances, as well as overall business effectiveness of the organizations, may be achieved by application of appropriate improvement strategies which are defined by decision makers, who have to make decisions based on data derived from uncertain environment. To overcome uncertain environment influence, it is possible to use some approaches such as fuzzy set theory [6], and genetic algorithm (GA) [7], since these methods are seen as powerful tools that could enable use of data from uncertain environment and their transformation into information necessary for assistance in business decision making.

The goal of this paper is to present newly developed model for NSD process performance assessment and improvement under uncertainties based on Key Performance Indicators (KPIs) similar to Pan et al. [2]. The considered problem has become critical area of interest in service and non-service industry for additional value creation [8]. Modelling of all uncertainties is performed by using the fuzzy set theory. The novelty aspects of the proposed method may be presented in the following manner: 1) the determination of the each identified KPI relative importance is stated as the fuzzy group decision making problem and it is obtained by applying novel proposed procedure; 2) the rank of identified KPIs is obtained by using multi-objective GA; and 3) by applying fuzzy if-then rules the improvement level of treated KPIs is given. Model testing is performed on a sample of 100 small and medium enterprises (SMEs) of the service oriented industry. By applying developed model, optimal improvement of the NSD process effectiveness may be determined on an exact way. It can help decision makers to define the appropriate improvement strategy, with lower costs and time consumption.

In order to define assessment and improvement approach, the analysis of the NSD process in accordance with the requirements of the relevant ISO standards and process approach, as one of eight basic principles of quality management, has been done. Systematic management and effective monitoring of NSD processes performance are enabled in business organizations which use process approach defined by application of ISO 9001 standards [9]. Assessment is based on the NSD process KPI weights and values definition and measurement. The starting assumption was that the weights and values of the KPIs are hard to be described by the precise numbers [10]. Consequently, in order to represent KPIs weights and values linguistic expressions are used, as they are closer to human thinking. Modelling of linguistic expressions is based on a fuzzy set theory.

The problem of process performance improvement could be seen as a multiple-criteria optimization problem [11], which may be solved by genetic algorithm (GA), as it has been performed in this paper. By applying this approach, it is possible to identify optimal KPI ranking coefficients and KPIs modifications values. KPIs modifications values obtained by GA method are calculated with respect to the different factors that may be mutually conflicting. Optimal improved KPI values could have very significant or almost insignificant influence on further process performance enhancement. Impact significance is obtained by using fuzzy if-then rules. Fuzzy if-then rules have already been used alongside with GA to determine if KPIs modification values aresufficient for improvement [7, 12].

Paper is organized in the following manner: in the first section basics of the proposed methods and performance assessment and improvement are stated; in the second section the review of the literature is presented; in third section proposed model is defined, while algorithm of the proposed model is presented in fourth section, and its application is introduced in fifth section. Finally, conclusions are presented in sixth section.

2 Literature review

A significant number of researchers have worked with the issues related to the NSD process in the recent time and they focused on a frontline employees’ creativity [13], different types of relationships between new product development and NSD process [14, 15], logistics service quality [16], customer involvement in service development [17], service supply chains design [18, 19], but few of them were oriented to the application of data decision making tools and methods for overall process performance assessment and improvement.

The use of these tools and methods could be crucial to improve and achieve the organization’s competitive position in the market [20]. A well-defined improvement methodology may be important for decision makers and managers, if it is used for: identification of weak points, effective reactions and weak points solving, and for the business process quality improvement [21]. It could be stated that improvement may include activities which enable continual effective and efficient strategic goal achievement, and the ability to coordinate resources, systems and employees in accordance with organisational overall strategy [22]. Successful service oriented organizations have recognized that it is necessary to focus attention on the services customers. So that focus on customers has become an integral part of performance assessment and managing. Through performance improvement, the service oriented organization may promote the creation of value for the customers, whether it is in terms of quality or price of deliveries, with the result of the economic added value for stakeholders and owners [22]. One of possibilities is that performance improvement could be based on KPIs, focused on aspects of the most critical organizational performance for the current and future success of the organizations [23]. These KPIs have to be defined and selected by organization top managers [1]. In this paper, defined KPIs have been used to determine overall performance of NSD process, and based on the literature review. The authors believe that the selection of KPIs, obtained according to the literature is less burdened by the subjective assessment of decision makers.

For effective and rapid adaptation to global market, service organizations have to assess KPIs related to critical activities of their processes [24]. Literature indicates that the most papers provide recommendations for the definition of new product development process’ KPIs [25]. On the other hand, research in the field of NSD is increasingly important [26], not only for service organizations, but also for other organizations, who wish to provide appropriate service. Shortening NSD time cycle and service performance improvement in uncertain environment have become strategic goals of many service oriented organizations, these are the reasons why fuzzy data and heuristic methods based models may beimplemented.

Sometimes performance are not measurable and determination of their values is based on various methods that could be found in literature. For instance: a new self-assessment method [27], the use of fuzzy linguistic scale [28 –30], and an interview method [31, 32]. Decision makers are faced with many uncertainties, which can often occur in areas where the reasoning, evaluation and decision making are very important. In many papers that can be found in literature uncertainties into KPIs values may be well described by using fuzzy set theory[24 , 34]. In this paper, modelling of existing uncertainties is based on fuzzy set theory, so that they are precisely and rigorously described [6]. Also, flexibility of fuzzy sets makes them ideal for representation of vague concepts, in which the availability of information is often insufficient for the usage of probability theory [35]. Relative importance of KPIs may be assessed by decision makers which make a decision by consensus [33]. Many authors suggest that the problem of determining KPIs relative importance should be observed as fuzzy group decision making problem [29 , 24]. The aggregation of individual options in group consensus can be obtained by applying different operators, for example max operator [30], geometric mean [29], and fuzzy Analytical Hierarchy Process (FAHP) [24, 34]. In this paper calculation of aggregated importances of KPIs is based on proposed method. Authors believe that given the nature and magnitude of the problem, the proposed method is more appropriate in relation to the methods used in the literature.

The ranking of KPI may be performed by using FAHP [32, 33], the fuzzy Technique for Order of Preference by Similarity to Ideal Solution [34], and GA [24]. GAs may help solving problems in cases in which traditional optimization techniques require detailed preparation of input data and the use of complex mathematical modelling. It may be stated that GAs managed to find their application in many scientific and industrial fields. The reason for this is that GAs don’t have starting search point. They are able to find the global optimum in a search field with a lot of local optimal values; also, they can handle a larger number of objective functions, and linear or non-linear constraints. In this paper, with the respect to the type and the size of the problem, the multi-objective functions of GA model are defined.

In the literature, determination of the business process improvement strategies is mainly based on the standard requirements, the results of good practice, etc. Application of fuzzy algebra may provide accurate business information that can be used to assists the decision making processes [36, 37], and thus may simplify performance improvement activities of NSD process. There are a few papers in which the problem of determining the priority of the performance to be improved is being addressed [24 , 32].However, by applying these methods, it cannot be answered how much selected performance and their KPIs should be improved. In this paper, fuzzy if-then rules were used to determine level of selected KPIs improvement. There are a certain number of papers regarding application of fuzzy if-then rules for decision making in different fields [34].

In this way, the main differences between analyzed models and proposed model are described, and the scientific contribution of the proposed model is presented in detail.

3 The problem framework

Performance assessment and improvement represent complex management issues that may be realized in two integrated steps described in this section. The first step is used of assessment and selection of performance that should be improved, while in the second step the optimal improvement level for each selected performance is determined. Application of the developed approach may lead to a highest possible effectiveness of business processes and at the same time that to lowest resources consumption. The proposed model framework is presented, as follows.

NSD performance and their KPIs in SMEs from service domain are defined based on the literature review [26, 38], and the results of good practice. They are introduced in Table 1.

Table 1
New service development KPIs

NSD performance KPI Indicators Description

Idea Generation i = 1 Idea verification Percentage of reviewed and approved new services proposals for further development by leaders and managers

i = 2 Feedback time The average time elapsed from the submission of ideas to the decision by the leaders and managers to initiate the development

Concept development i = 3 Customer engagement Yes/no metric that indicates whether users are involved in the development process, i.e. if there is feedback received from users regarding new ideas

i = 4 Time-to-market The time between the creation of the service specification to the determination of its validity

Concept implementation i = 5 Staff training effectiveness Rate of success of employees training to provide new services

i = 6 New services testing success Success rate of the new services final test

Full Prototype Verification i = 7 Development costs allocation success Ratio of costs allocated to the development of new services and the total income of the SME in the same period expressed in percentages

i = 8 Process capability Fulfilment of user and market needs by new services development expressed in percentages

i = 9 The time to development success The ratio of the number of working days spent on the development and the number of anticipated working days for the project expressed in percentages

NSD performance	KPI	Indicators	Description
Idea Generation	i = 1	Idea verification	Percentage of reviewed and approved new services proposals for further development by leaders and managers
	i = 2	Feedback time	The average time elapsed from the submission of ideas to the decision by the leaders and managers to initiate the development
Concept development	i = 3	Customer engagement	Yes/no metric that indicates whether users are involved in the development process, i.e. if there is feedback received from users regarding new ideas
	i = 4	Time-to-market	The time between the creation of the service specification to the determination of its validity
Concept implementation	i = 5	Staff training effectiveness	Rate of success of employees training to provide new services
	i = 6	New services testing success	Success rate of the new services final test
Full Prototype Verification	i = 7	Development costs allocation success	Ratio of costs allocated to the development of new services and the total income of the SME in the same period expressed in percentages
	i = 8	Process capability	Fulfilment of user and market needs by new services development expressed in percentages
	i = 9	The time to development success	The ratio of the number of working days spent on the development and the number of anticipated working days for the project expressed in percentages

Formally, presentation of SMEs is denoted by set ɛ ={ 1, …, e, … , E }. In each SME relevant decisions at the NSD process level are carried out by decision makers. These decision makers (general manager, quality manager, process owner, and development manager) are denoted by set K = (1, …, k, … , K), with k representing each decision maker, and K representing total number of decision makers. In general, from each observed SME management team, four or less decision makers were delegated. Number of managers on the level of each SME involved in decision making depends on the enterprise size. These managers’ opinions in terms of decision-making and NSD effectiveness process evaluation are not of the same significance. Having this in mind, managers are divided into groups, so that set K may be represent as union of K_g sets. K_g represents total number of managers in a set denoted as γ ={ 1, …, g, …, G }, with g representing the group of managers with the same significance in SMEs, while G represents the total number of decision maker groups. The importance value of each group W_g, g = 1, …, G, is determined by the expert team (leaders and top managers of the observed SMEs).

An NSD process is composed of activities focused on the implementation of creative ways for service creation. For NSD corresponding KPIs have been determined. KPIs are formally presented by the set of indices ι ={ 1, …, i, …, I }. Each KPI is denoted by i, while the total number of KPIs is I. The assumption is that all KPIs are immeasurable and of benefit type.

On these grounds, mathematical methods could be used to resolve three considered management tasks: (1) to determine fuzzy rank of KPIs at the level of all SMEs; (2) to determine optimal improvement values of selected KPIs by applying GA and fuzzy if then rules.

In this paper, it is assumed that KPIs do not have the same importance. In order to determine values of these KPIs importance, a certain number of steps is defined within the model. Firstly, the relative importance of each KPI is assessed by each decision maker. Decision makers form their opinions based on results of a good practice. They express their opinion with pre-defined linguistic expressions. Quantitative description of linguistic variables was based on the theory of fuzzy sets [6, 39] and presented bytriangular fuzzy numbers (TFNs).

Secondly, the relative importance of KPIs at the level of each management group have been obtained by using the fuzzy average value operator. Afterwards, aggregated relative importance values of each KPIs at the level of all considered SMEs have been obtained by using the Fuzzy Ordered Weighted Averaging (FOWA) operator, defined by Merigo and Casanovas [40].

The managers used seven predefined linguistic expressions which are modelled by TFNs for describing of the relative importance of KPIs: very low importance- ${\tilde{R}}_{1} = (0, 0, 0.15);$ low importance- ${\tilde{R}}_{2} = (0, 0.15, 0.3);$ fairly moderate importance- ${\tilde{R}}_{3} = (0.2, 0.35, 0.5);$ moderate im portance- ${\tilde{R}}_{4}$ = (0.35, 0.5, 0.65) ; fairly high importance- ${\tilde{R}}_{5} =$ (0.5, 0.65, 0.8) ; high importance- ${\tilde{R}}_{6} = (0.7,$ 0.85, 1) ; and very high importance- ${\tilde{R}}_{7} = (0.85, 1, 1) .$

According to benchmarking results, it can be concluded that KPI values should be determined at the level of each SME. Respecting selected KPIs of NSD process it is clear that their values are almost impossible to get by measuring. Hence, values of selected KPIs are assessed by decision makers. It is regarded that, decision making of management team at the level of SME is performed by consensus. Decision maker judgments are based on their knowledge and experience. The management teams used five predefined linguistic expressions that are modelled by TFN: very low value ${\tilde{v}}_{1} = (0, 0, 0.25)$ ; low value ${\tilde{v}}_{2} = (0.15, 0.3, 0.45)$ ; middle value ${\tilde{v}}_{3} = (0.35, 0.5, 0.65)$ ; high value ${\tilde{v}}_{4} = (0.55, 0.7, 0.85)$ ; and very high value ${\tilde{v}}_{5} = (0.75, 1, 1)$ .

Determination of KPIs fuzzy ranks coefficients values marked as r_i, i = 1, … I, with respect to all the analysed SMEs, GA has been used. Optimization functions used to determine optimal values of rank coefficients of KPIs are sum and variance of ranks coefficients values, for each level of TFNsmembership function.

Optimization variables existing in objective function are: the weights ${\tilde{w}}_{i}$ at the level of all SMEs, the KPI value at the level of each SME $d_{i}^{e}$ , and the mean value $\bar{d}$ of KPIs for all SMEs. These values are coded with decimal chromosomes, and optimal KPIs ranks coefficients values with respect to all SMEs have been obtained by MATLAB GA toolbox application. According to the obtained results rank of KPIs is determined, so that the first place in the rank is assigned to KPI that has the smallest associated KPI rank coefficient value, while the last place assigned to KPI that has the highest associated KPI rankcoefficient value.

Optimal objective function values are obtained in the form of Pareto front, so that the decision-maker may choose closest solution to his business concept. Having this in mind, it may be stated that multi-criteria optimization is among the methods applied before decision making, namely search prior to the decision [41]. Within MATLAB environment GA parameters for optimal ranking determination were defined.

Second task, which may be accomplished by using the GA, refers to determination of the optimal values of KPIs to improve the quality of the entire NSD process in any observed SME. However, for KPIs values enhancement there is a wide range of strategies, namely large search space in which it is difficult to find the solution that will satisfy all the constraints. This kind of problem, also, presents a good field for the application of GA optimization method. In this case, determined optimal KPIs values are derived from the condition for the minimum seeking of the proposed objective function f, which may lead to enhancement of process effectiveness, with respect to the defined constraints. Objective function f can be defined as difference between summarized ranks coefficients (S^i′) ₁ after and before the modification of selected KPIs for the randomly selected SME; constraint is defined as the modification sum of all selected KPIs, β. The total modification belongs to interval form 0% to 100%. Discretization step of the considered interval in which β is defined is determined by decision makers. They base their assessment on knowledge and experience, with respect to the region development in which SMEs are operating. This value is marked as Δβ. The optimal values of KPI improvements are defined according to the fuzzy if-then rules. Based on the obtained results, it can be concluded whether the applied measures of KPIs improvements would lead to effectiveness improvement of NSD process.

4 The proposed algorithm

Step 1. The fuzzy rating of the relative importance of KPIs at the level of each group g, g = 1, …, G is given by manager kg = 1, …, Kg marked as: ${\tilde{w}}_{i}^{kg}$ , i = 1, …, I .

Step 2. The aggregated relative importance of KPI i, i = 1, …, I at the group g, g = 1, …, G level is calculated by using fuzzy averaging method: ${\tilde{w}}_{i}^{g} = \frac{1}{K_{g}} \sum_{k = 1}^{K_{g}} {\tilde{w}}_{i}^{kg}$

Step 3. The overall relative importance of KPI i, i = 1, …, I with respect to all groups of decision makers is obtained by FOWA [40]: ${\tilde{w}}_{i} = \sum_{g = 1}^{G} ({\tilde{w}}_{i}^{g} \cdot w_{g})$

Step 4. KPI fuzzy values at the level of each SME e, e = 1, …, E are given by management team. These values are denoted as ${\tilde{v}}_{i}^{e}$ .

Step 5. The weighted KPI values at the level of each SME, ${\tilde{d}}_{i}^{e}$ i, i = 1, … I, e = 1, …, E, is given: ${\tilde{d}}_{i}^{e} = {\tilde{w}}_{i} \cdot {\tilde{v}}_{i}^{e}$

Step 6. Determine the representative scalar of TFN $d_{i}^{\sim^{e}}, d_{i}^{e}$ by using the maximum possibility method [39].

Step 7. Optimization functions are: $max (S_{total}^{i}) = \sum_{i = 1}^{I_{j}} r_{i} \cdot \sum_{e = 1}^{E} d_{i}^{e}$ $min ({Var}_{i}) = \frac{\sqrt{\sum_{i = 1}^{I} r_{i} \cdot \sum_{e = 1}^{E} {(d_{i}^{e} - \bar{d})}^{2}}}{E \cdot I}$

where:

$S_{total}^{i}$ is sum of KPIs ranks coefficient values,

Var_i is the ranks coefficient values variance.

Step 8. By using GA, KPIs ranks coefficient values, with respect of all SMEs, are determined.

Step 9. KPIs ranks coefficient values are sorted in decreasing order. The rank of KPIs corresponds to the KPIs ranks coefficient values.

Step 10. Selection of KPIs which should be improved is given with respect to obtained rank of KPIs. Total number of selected KPIs is denoted as I and determined by decision makes jugdments. Index of selected KPI is i, i = 1, …, I′.

Step 11. To determine the modified value of selected KPIs i ∈ I′ at the level of randomly selected SME e^*, by application of GA method and fuzzy algebra rules:

Objective function: $\min [(1 - δ_{i^{'}}) S^{i^{'}} - S^{i^{'}}]$

Subject to: $\sum_{i^{'}}^{I^{'}} δ_{i^{'}} \leq β$

and: ${\tilde{v}}_{i^{'}}^{e^{*}} = (1 + δ_{i^{'}}) {\tilde{v}}_{i^{'}}^{e^{*}} i = 1, \dots, I^{'}$

Step 12. By application of fuzzy if-then rules, modified values of selected KPIs for certain β may be determined:

IF the modified KPI value equals to ${\tilde{v}}_{i^{'}}^{e^{*}}$ , THAN KPI value is described by linguistic expression where: $max_{i = 1, . ., 5} μ_{{\tilde{v}}_{i}} (\underset{i^{'}}{{\tilde{v}}^{e^{*}}}) = μ_{{\tilde{v}}_{i^{*}}}$

Step 13. Based on the determined results it can be determined at which value of β it is necessary to take management initiatives with a goal of significant improvements of KPIs, which are further propagated to increase the effectiveness of the NSD process.

5 Application of the proposed model

In this paper, a certain number of service oriented SMEs with applied ISO 9001 : 2015 standard have been considered. By applying ISO 9001 : 2015 standard, SMEs are obligated to implement processes approach as one of the basic principles of the quality management, and to measure, analyse and continually improve process and business performance.

It is known that service industry has an important role in developing counties, such as Republic of Serbia. Proposed model is tested with real life data. Sample consists of 100 SMEs in Serbian service industry is used.

As it has already been said, managers of all SMEs are classified in four groups: general managers (g = 1), quality managers (g = 2), process owners (g = 3), and development managers (g = 4),respectively. On treated sample there are 100 general managers, 51 quality managers, 18 process owners, and 18 development managers. It this study, the team of experts, composed of leaders of the surveyed SMEs and stakeholders, determined the importance of manager groups, w₁ = 0.4 for general managers, w₂ = 0.3 for quality managers, w₃ = 0.2 for process owners, and w₄ = 0.1 for development managers.

Thereafter, overall relative importance for the KPIs have been obtained by using the proposed Algorithm (Step 1 to Step 3). The proposed procedure is illustrated on example of determination of the relative importance of KPI denoted as Verification of ideas (i = 1).

Aggregated relative importance of KPI (i = 1) at the level of each group of managers is calculated by using fuzzy averaging method: $\begin{matrix} \begin{matrix} {\tilde{w}}_{1}^{1} = 0.4 \cdot \frac{1}{100} \cdot (\begin{matrix} 11 \cdot {\tilde{R}}_{7} + 21 \cdot {\tilde{R}}_{6} + 28 \cdot {\tilde{R}}_{5} + {\tilde{25 \cdot R}}_{4} + \\ + 9 \cdot {\tilde{R}}_{3} + 6 \cdot {\tilde{R}}_{2} \end{matrix}) = \\ = (0.187, 0.254, 0.307) \end{matrix} \\ \begin{matrix} {\tilde{w}}_{1}^{2} = 0.3 \cdot \frac{1}{51} \cdot (13 \cdot {\tilde{R}}_{7} + 12 \cdot {\tilde{R}}_{6} + 15 \cdot {\tilde{R}}_{5} + 5 \cdot {\tilde{R}}_{4} + 5 \cdot {\tilde{R}}_{3} + {\tilde{R}}_{2}) = \\ = (0.174, 0.219, 0.253) \end{matrix} \\ \begin{matrix} {\tilde{w}}_{1}^{3} = 0.2 \cdot \frac{1}{18} \cdot (6 \cdot {\tilde{R}}_{7} + 5 \cdot {\tilde{R}}_{6} + 2 \cdot {\tilde{R}}_{5} + 3 \cdot {\tilde{R}}_{4} + {\tilde{R}}_{3}) = \\ = (0.120, 0.148, 0.167) \end{matrix} \\ \begin{matrix} {\tilde{w}}_{1}^{4} = 0.1 \cdot \frac{1}{18} \cdot (6 \cdot {\tilde{R}}_{7} + 5 \cdot {\tilde{R}}_{6} + 3 \cdot {\tilde{R}}_{5} + 1 \cdot {\tilde{R}}_{4} + 3 \cdot {\tilde{R}}_{3} + 2 \cdot {\tilde{R}}_{2}) = \\ = (0.061, 0.072, 0.089) \end{matrix} \end{matrix}$ The overall relative importance for the KPI (i = 1), is obtained by using FOWA: ${\tilde{w}}_{1} = 0.5 \cdot {\tilde{w}}_{1}^{1} + 0.3 \cdot {\tilde{w}}_{1}^{2} + 0.2 \cdot {\tilde{w}}_{1}^{3} + 0.1 \cdot {\tilde{w}}_{1}^{4} = (0.375, 0.696, 0.818)$

In a similar way, overall relative importance for the rest of KPIs have been calculated, as follows: $\begin{matrix} {\tilde{W}}_{2} = (0.571, 0.720, 0.831), {\tilde{W}}_{3} = (0.540, 0.722, 0.813), \\ {\tilde{W}}_{4} = (0.528, 0.678, 0.800), {\tilde{W}}_{5} = (0.490, 0.640, 0.768), \end{matrix}$

$\begin{matrix} {\tilde{W}}_{6} = (0.462, 0.580, 0.742), {\tilde{W}}_{7} = (0.444, 0.591, 0.728), \\ {\tilde{W}}_{8} = (0.483, 0.625, 0.758), and {\tilde{W}}_{9} = (0.498, 0.644, 0.777) . \end{matrix}$ The weight values of KPIs are calculated by using the proposed Algorithm (Step 4 to Step 6). The proposed procedure is illustrated on example of weighted value KPI denoted as Time-to-market (i = 4) determination for SME (e = 15): $\begin{matrix} {\tilde{d}}_{4}^{15} = (0.528, 0.678, 0.8) \cdot (0.15, 0.3, 0.45) = \\ (0.079, 0.203, 0.360) . \end{matrix}$

The representative scalar of TFN ${\tilde{d}}_{4}^{15}$ is 0.203.

In a similar manner, the weighted KPI values for all SMEs from sample have been calculated.

The software solution for KPI weights and values imputation and further model application is created within MATLAB environment. In the next steps of the proposed Algorithm (Step 7 to Step 9), ranking optimization functions were defined and optimization has been carried out within developed MATLAB environment. In MATLAB GA toolbox different variations of GA parameters (population size, crossover function, selection function, mutation function, stopping condition) were used (Table 2). Within these steps, MATLAB GA toolbox has been used for optimal KPI fuzzy ranking.

Table 2
Used GA parameters value variations in MATLAB environment

Variation Population size Selection Mutation Crossover Stopping condition

1 150 tournament adaptive feasible heuristic after 100 generations

2 100 uniform uniform at one point after 150 generations

3 75 roulette constraint dependent in two points after 200 generations

Variation	Population size	Selection	Mutation	Crossover	Stopping condition
1	150	tournament	adaptive feasible	heuristic	after 100 generations
2	100	uniform	uniform	at one point	after 150 generations
3	75	roulette	constraint dependent	in two points	after 200 generations

After the optimization, results of KPI rank coefficient values were obtained in a form of a Pareto front. Obtained Pareto front consists of asterisks which are presented in Fig. 1a-c. Asterisks present possible optimal solutions for both optimization functions [42]. On the abscissa solutions for the equation (4) and on the ordinate solutions for the equation (5) are presented.

Fig.1

Pareto solutions for optimal key performance indicators ranking a) variation 1, b) variation 2, and c) variation 3.

Decision makers may select any solution from Pareto front, due to their mutual non-dominance. The number of non-dominant solutions depends primarily on the number of optimization variables, but it may also depend on parameters of the GA algorithm. In this paper, authors have chosen solution with the lowest variance and the lowest ranking sum from 70 obtained solutions (Fig. 1c). This allowed the authors to rank KPIs with minor deviations, which means that KPIs with better ranks did not get very high scores, and, also, KPIs with poor ranks did not get very low scores. Based on the chosen solution, ranks were determined and presented on a Fig. 2. On the abscissa axis KPIs fuzzy ranks r_i.

Fig.2

The ranks for New service development key performance indicators.

The KPIs rank coefficient values and rank of KPIs are presented in Table 3.

Table 3

KPI ranks and corresponding ranks of fuzzy rank coefficients

KPI	KPIs rank coefficient values	Rank
i = 1	0.896	8
i = 2	1	9
i = 3	0.868	7
i = 4	0.539	6
i = 5	0.299	3
i = 6	0.339	4
i = 7	0.262	1
i = 8	0.377	5
i = 9	0.298	2

According to obtained results from Table 3 following KPI are selected by decision makers (Step 10 of the proposed algorithm): Development costs allocation success (i = 7), The time to consumption success (i = 9), and Staff training effectiveness (i = 5). The reason for this may be because the allocation of expenses is rarely taken into account, rarely taken care of the training for the employees and the number of days spent on the development of new services in the Serbian SMEs is neglected. The selected KPIs were the subject of optimization, for randomly selectedSME e^*.

Desired level of NSD process enhancement is expressed through constraint factor β. In SME that are operating in developed countries improvement of business processes may be based on the implementation of Kaizen approach that doesn’t involve high costs. Unfortunately, in developing countries, such as Republic Serbia incremental improvements of business processes performances don’t lead to the improvement of their effectiveness. In this paper, assumption is that the overall modification of business process performances belongs to the interval (0% –100%). It is proposed that the discretization step should be 25%.

According to different values of constraint factor β, which are: 25%, 50%, 75% and 100%, selected KPI modified values are determined (Step 11 of the proposed Algorithm) and presented in Table 4.

Table 4

KPI ranks and corresponding ranks of fuzzy rank coefficients

Constraint factor	Observed indicators – improvement values in percentages
β	i = 5	i = 7	i = 9
0.25	0.103	0.001	0.145
0.5	0.187	0.106	0.207
0.75	0.267	0.189	0.294
1	0.340	0.300	0.359

The modified values of selected KPIs at each discretization step Δβ are determined by the proposed Algorithm (Step 11) and presented at the Fig. 3a–c.

Fig.3

KPI improvement presented through linguistic expression a) i = 5 b) i = 7, and c) i = 9.

By applying fuzzy if then rules it can be determined if the modified KPI values should be described with the same or some other pre-defined linguistic expressions (Step 12 of the proposed Algorithm). In other words, application of this rule provides an opportunity for decision makers to determine whether the KPI modified values for each β value led to the improvement of process performance or not.

Based on the results of the fuzzy if-then rules application, determination of optimal improvements through linguistic expression is obtained (Figs. 4–6). According to these results decision maker may define improvement strategies (Step 13 of the proposed Algorithm).

Fig.4

Improvement of KPI (i = 5) for different β values.

Fig.5

Improvement of KPI (i = 7) for different β values.

Fig.6

Improvement of KPI (i = 9) for different β values.

Business process improvement is associated with investments, meaning that the improvement can be achieved only with a consumption of resources. However, decision makers’ aim is to achieve the highest possible improvement of business processes with the least possible consumption of resources. By applying the proposed model this objective can be achieved.

Based on the results that are shown in Figs. 4–6, it can clearly be concluded that only at high overall modifications may lead to the performance improvement, which could further be propagated on the considered process efficiency. If the overall modification is 50% (the least resource consumption) it will lead to an improvement of Time to development success (i = 9) (Fig. 6). At the same level of the overall modification other two KPIs values are not improved, they wound not have an impact on the increase of this business process effectiveness. One of the possible strategies of improvement may be defined as taking of appropriate management measures, such as: (1) introduction of new technologies that are based on information and communication technologies; if this management activity is applied, following benefits may be achieved: a) improvement of communication between members of the management team, b) a better understanding of the stakeholders demands, c) reduced duration of the NSD process, d) improvement of the connection between the NSD and other processes in the organization; also (2) the introduction of project management in accordance with ISO 10006 in order to provide: development of team members, services, and employees development; especially when it is known that enhancement of employees competence level in the NSD process provides an opportunity to achieve higher long-term sustainability and competitiveness of theorganization.

Improvement of Staff training effectiveness (i = 5) may be realized only if the overall modification of all considered KPIs is 75% (Fig. 4). The most important management measures that can lead to improvements in (i = 5) is to design and implement staff trainings. By applying these measures, it is possible to increase the awareness of employees about the importance of knowledge enhancement, and to increase the knowledge and skills of employees that are needed to realize the development strategy of the organization. Improvement of (i = 5) should lead to enhancement of organizational knowledge level, which is the most important resource for competitive advantages achievement in a long term period.

The greatest improvement in the effectiveness of business process can be achieved if the value of the overall modification of KPIs is 100% (the highest resource consumption). At this level of the overall modification, also the value of Development costs allocation success (i = 7) would be improved (Fig. 5). Improvement of (i = 7) could be made by using the process costing method at the level of each activity of the development project in the NSD process.

6 Conclusion

The development of new technologies leads to changes in customers’ preferences, which are further propagated on effectiveness of business processes, especially on process of NSD. Solution of treated problem has the critical impact of the organizational competitiveness. It may be concluded that improvement of new service development effectiveness presents an interesting topic in research and practice domain in last decades.

Estimation of NSD process effectiveness may be achieved by using novel two-steps model. Uncertainties in the relative importance and values of identified KPIs are described by predefined linguistic terms, which are modelled by TFNs. The first novelty, determination of the relative importance of KPIs is stated as fuzzy group decision making problem. The aggregated fuzzy relative importance of KPIs are calculated by using different operators. The second novelty, the KPIs ranking problem is stated as multi-criteria decision making problem. The weighted KPI values present input data for the proposed model application within MATLAB GA toolbox. By using the proposed GA multi-objective optimization method, KPIs rank coefficient values are obtained. Selection of KPIs that should be improved is performed according to obtained KPIs rank coefficient values. Modified KPI values at the level of each constraint factor β and at the level of each SME are obtained by applying GA single-objective optimization method and fuzzy algebrarules.

The third novelty presents the use of fuzzy if-then rules to determine level of KPIs improvement and management initiatives which lead to improvement of the NSD process effectiveness. The model was tested on a sample of 100 SMEs from Serbian service industry.

The advantages of the proposed model in comparison with the models which can be found in the literature may be presented as: 1) taking into account different uncertainties and imprecise for calculation of KPIs relative importance and their values; 2) aggregated KPIs relative importance are obtained with respect to all decision makers and their weights; 3) the rank of KPIs is determined on a large sample of SMEs by applying the multi-objective GA method; 4) to improve level of each selected KPI fuzzy if-then rules are used; 5) by applying developed model, the improvement strategy that demands lower resources consumption may be defined.

By applying the proposed model it is possible to identify weak points in organizations, to provide corrective actions, and to enhance the effectiveness of the observed process. Consequently, the model can represent a suitable solution for reengineering and obtain improvement of the process performance, in terms of cost, quality, service and time consumption.

Beside above mentioned various advantages of the proposed model, this research work can be easily extended to the analysis of other business processes in different research areas.

The constraints of the presented research are linked to the observed SMEs (SMEs operating within Serbian service industry). Other constraints are connected to the defined parameter values within GA method. If the other parameter values were used, then other optimal solutions would be possible.

Consequently, directions for further research are arising from the constraints of research; they will focus on expanding research on the other countries SMEs; and on covering the implementation of the model in a greater number of service organizations (from different sectors, sizes and markets) in order to confirm the generality of the model.

References

Pulakos

, Performance management, Wiley-Blackwell, Chichester, West Susssex, UK, 2009.

Pan

, Kuo

and Bretholt

, Developing a new key performance index for measuring service quality, Industrial Management & Data Systems110(6) (2010), 823–840.

Agarwal

, Krishna Erramilli

and Dev

, Market orientation and performance in service firms: Role of innovation, Journal of Services Marketing17(1) (2003), 68–82.

Kaplan

and Norton

, The balanced scorecard, Harvard Business Review Press, Boston, Mass; 2010.

Nudurupati

, Bititci

, Kumar

and Chan

, State of the art literature review on performance measurement, Computers & Industrial Engineering60(2) (2011), 279–290.

Zimmermann

, Fuzzy set theory-and its applications, Kluwer Academic Publishers, Boston, Mass; 2001.

Bala

and Ratnoo

, A genetic algorithm approach for discovering tuned fuzzy classification rules with intra- and inter-class exceptions, Journal of Intelligent Systems25(2) (2016), 263–282.

Guo

, Li

, Zuo

and Chen

, Influence of organizational elements on manufacturing firms’ service-enhancement: An empirical study based on Chinese ICT industry, Technology in Society43 (2015), 183–190.

ISO/IEC 9001:2015, ISO 9001:2008 Quality management systems – Requirements, ISO/IEC, Geneva, Switzerland; 2015.

10.

Urbański

and Wąsowski

, Fuzzy measurement theory, Measurement41(4) (2008), 391–402.

11.

Sadi-Nezhad

and Khalili Damghani

, Application of a fuzzy TOPSIS method base on modified preference ratio and fuzzy distance measurement in assessment of traffic police centers performance, Applied Soft Computing10(4) (2010), 1028–1039.

12.

Ishibuchi

, Nozaki

, Yamamoto

and Tanaka

, Selecting fuzzy if-then rules for classification problems using genetic algorithms, IEEE Transactions on Fuzzy Systems3(3) (1995), 260–270.

13.

Yang

, Lee

and Cheng

, Continuous improvement competence, employee creativity, and new service development performance: A frontline employee perspective, International Journal of Production Economics171 (2016), 275–288.

14.

Chen

, Wang

, Huang

and Shen

, Service innovation and new product performance: The influence of market-linking capabilities and market turbulence, International Journal of Production Economics172 (2016), 54–64.

15.

Homburg

and Kuehnl

, Is the more always better? A comparative study of internal and external integration practices in new product and new service development, Journal of Business Research67(7) (2014), 1360–1367.

16.

Gil Saura

, Servera Francés

, Berenguer Contrí

and Fuentes Blasco

, Logistics service quality: A new way to loyalty, Industrial Management & Data Systems108(5) (2008), 650–668.

17.

Carbonell

, Rodríguez-Escudero

and Pujari

, Customer involvement in new service development: An examination of antecedents and outcomes, Journal of Product Innovation Management26(5) (2009), 536–550.

18.

Rosyidi

, Murtisari

and Jauhari

, A concurrent optimization model for suppliers selection, tolerance and component allocation with fuzzy quality loss, Cogent Engineering3(1) (2016), 1222043.

19.

Rouzafzoon

and Helo

, Developing service supply chains by using agent based simulation, Industrial Management & Data Systems116(2) (2016), 255–270.

20.

Kouki

, Joint impact of ERP systems and non-financial performance indicators on corporate financial performance: Evidence from French listed companies, Management Science Letters5(1) (2015), 1–10.

21.

El-Haik

and Shaout

, Software design for six sigma, John Wiley & Sons, Hoboken, N.J.; 2010.

22.

Aguinis

, Performance management, Pearson Prentice Hall, Upper Saddle River, N.J.; 2009.

23.

Parmenter

, Key performance indicators, John Wiley & Sons, Hoboken, N.J.; 2010.

24.

Nestic

, Djordjevic

, Puskaric

, Djordjevic

, Tadic

and Stefanovic

, The evaluation and improvement of process quality by using the fuzzy sets theory and genetic algorithm approach, Journal of Intelligent & Fuzzy Systems29(5) (2015), 2017–2028.

25.

Dorigatti

, Guarnaschelli

, Chiotti

and Salomone

, A service-oriented framework for agent-based simulations of collaborative supply chains, Computers in Industry83 (2016), 92–107.

26.

Menor

, Tatikonda

and Sampson

, New service development: Areas for exploitation and exploration, Journal of Operations Management20(2) (2002), 135–157.

27.

Cocca

and Alberti

, A framework to assess performance measurement systems in SMEs, International Journal of Productivity and Performance Management59(2) (2010), 186–200.

28.

, Tzeng

and Chen

, A fuzzy MCDM approach for evaluating banking performance based on Balanced Scorecard, Expert Systems with Applications36(6) (2009), 10135–10147.

29.

Lee

, Chen

and Chang

, A fuzzy AHP and BSC approach for evaluating performance of IT department in the manufacturing industry in Taiwan, Expert Systems with Applications34(1) (2008), 96–107.

30.

Yüksel

İ.

and Dağdeviren

, Using the fuzzy analytic network process (ANP) for Balanced Scorecard (BSC): A case study for a manufacturing firm, Expert Systems with Applications37(2) (2010), 1270–1278.

31.

Feili

, Farahani

and Vesaghi

, Integration of fuzzy analytic hierarchy process (FAHP) with balance score card (BSC) in order to evaluate the performance of information technology in industry, The Journal of Mathematics and Computer Science2(2) (2011), 271–283.

32.

Talebnia

, The major perspectives weighted model for balanced scorecard system in the case of auto industries, Indian Journal of Science and Technology5(10) (2012), 3412–3420.

33.

Seçme

, Bayrakdaroğlu

and Kahraman

, Fuzzy performance evaluation in Turkish Banking Sector using Analytic Hierarchy Process and TOPSIS, Expert Systems with Applications36(9) (2009), 11699–11709.

34.

Tadic

, Aleksic

, Mimovic

, Puskaric

and Misita

, A model for evaluation of customer satisfaction with banking service quality in an uncertain environment, Total Quality Management & Business Excellence (2016), 1–20.

35.

Costea

, Applying Fuzzy Logic and Machine Learning Techniques in Financial Performance Predictions, Procedia Economics and Finance10 (2014), 4–9.

36.

Chaudhuri

, Dayal

and Narasayya

, An overview of business intelligence technology, Communications of the ACM54(8) (2011), 88.

37.

Zyngier

, An uncertainty management framework for industrial applications, Optimization and Engineering18(1) (2016), 179–202.

38.

Lundkvist

and Yakhlef

, Customer involvement in new service development: A conversational approach, Managing Service Quality: An International Journal14(23) (2004), 249–257.

39.

Dubois

and Prade

, Fuzzy sets and systems, Academic Press, New York; 1980.

40.

Merigó

and Casanovas

, The fuzzy generalized OWA operator and its application in strategic decision making, Cybernetics and Systems41(5) (2010), 359–370.

41.

Onat

, Kucukvar

, Tatari

and Zheng

, Combined application of multi-criteria optimization and life-cycle sustainability assessment for optimal distribution of alternative passenger cars in U.S., Journal of Cleaner Production112 (2016), 291–307.

42.

Caramia

and Dell’Olmo

, Multi-objective management in freight logistics, Springer, London; 2008.