Once a product is examined for preferred capabilities by a consumer population using the Consumer Preference Meta-Model, the collected information needs to be transformed further to requirements models, with the purpose to configure a software product line (SPL).
As shown in the previous section, consumer drivers in CPMM (Fig. 2) are derived using value-based frameworks for individuals. These drivers are related to a product of interest (Value Object) by emphasizing its desired properties, although in different ways. In early phases of requirements engineering, business models describing constellations of actors, provisioning and use of a product, or a group of related products, are closely related to goal frameworks, which can be used to elicit high-level system requirements for the product(s) [32, 33]. In the SPL discipline, goal-orientation is also recognized as an affirmative way to elicit variable and common system requirements by analyzing goals of stakeholders [1, 11–13].
Based on this argumentation, in this section we present a goal-based method for linking the preferences of consumers with requirements for software product lines. At first, we propose the mappings from CPMM to an established RE goal-oriented framework, namely i* ; second, we derive product-specific feature models of SPL from i* goal models. We illustrate the method through our empirical study by mapping the values, as elicited by a student population, to an i* model, and further, to feature models for SPL for online education.
Mapping Consumer Preference Meta-Model to i* framework
CPMM is a consumer-centric extension of a value-based business model, showing how a product intended for exchange between a provider and consumers is desired from consumer’s perspective. On the other side, the i* framework is meant for capturing intentions of a group of dependent actors, such as stakeholders in a requirements engineering process. i* provides a rich modeling notation in this context .
In this section, we propose the use of i* based on CPMM through mappings shown in Table 6 and through a set of accompanying guidelines. Relevant to this study, an i* SDM (Strategic Dependency Model) diagram is used to model all actors (student segments, the online education system, and the university) as well as their interdependencies (Fig. 5), while i* SRM (Strategic Rationale Models) diagrams are used to model internal actor’s interests and intentions (Figs. 5, 6). Table 5 presents intentional elements of i* used in the mappings of Table 6. Other elements of i* used are explained in line with the proposed guidelines.
Based on , where the mappings between e
3 business value model and i* goal framework are defined, we propose the mappings of our consumer-centric value meta-model CPMM, to i* as shown in Table 6.
The aforementioned mappings have been applied within the scope of our empirical study to build i* SDM diagram in Fig. 4 and i* SRM diagrams presented in Figs. 5 and 6.
For the SRM diagram, we present two parts of the complete diagram where the top basic values of each segment are elaborated. Top values are identified through the numbers used to annotate beliefs of agents: Universalism for non-master’s students and Self-determination for master’s students. These numbers are used to carry the priority of weight of the segments’ basic values as this is derived from CPMM and the results of our empirical study on these segments. Therefore, the SRM diagrams are focused on the corresponding consumer values of these top two basic values, based on the mappings of Table 6.
The corresponding consumer value for Universalism is Ethics, and thus one SRM diagram is focused on the soft-goal dependency “Ethics be Satisfied” (Fig. 5). The corresponding consumer value for Self-determination is Play, and thus the other SRM diagram is focused on the soft-goal dependency “Play be Satisfied” (presented in Fig. 6). Also, the SRM diagrams contain intentions for the core (domain) functionality of the online education system derived from the real-life setting of our example. These are exemplified with the goals “Online Examination be Supported,” “Course Material be Available,” and “Communication between Participants be Supported” (Figs. 5, 6). To distinguish which intentional elements are expressed by which segment we are following a consistent coloring scheme in the SRM diagrams. Darkly shaded elements express Non-Master students, lightly shaded elements express Master students, and non-shaded elements are those commonly expressed by both segments.
Within each agent, the belief Universalism is associated with the soft-goal “Ethics be Satisfied” (Fig. 4) and the belief Self-determination is associated with the Soft-Goal “Play be Satisfied” (Fig. 5). Within the online education system, actor each soft-goal coming from the soft-goal dependency expresses the students’ consumer value needed to be satisfied by the online education system. Therefore, the qualitative measures of CPMM are used to elaborate how these consumer soft-goals should be satisfied through other soft-goals and goals. Tasks and resources are omitted, as they are too specific to capture the intentionality needed to satisfied the consumer value soft-goals.
Guideline 1: Use answers to leading questions for each of the consumer values to identify intentions and preferences. Leading questions for each consumer value make use of examples for value archetype (e.g., fun for Play: How would you find fun in using an online education system?). Answers to such leading questions provide intentions and preferences, which can be expressed through goals and/or soft-goals affecting consumer value, and can be further elaborated.
The outcome of this guideline is a set of goals and/or soft-goals associated with the consumer value soft-goal directly. Thus, intentional elements directly associated with the consumer value soft-goal are derived from answers on leading questions. For example, Ethics is directly associated with soft-goals “Cheating be Prevented”, “Proper Rules of Conduct be Established”, “System be trusted” (Fig. 5), and Play is directly associated with soft-goals “System be Fun to Use”, “System be Interactive” (Fig. 6).
Guideline 2: Elaborate on the intentions affecting the elements derived from Guideline 1. Based on the qualitative measures of CPMM used to record refinements of the generic sets of values as expressed by consumers, intentions of consumers can be identified and expressed either as a soft-goal or as a goal.
The outcome of this guideline is a complete set of goals and/or soft-goals expressing consumers’ intentions of how the system can satisfied their consumer values.
Guideline 3: Group elements in respect to the goals/soft-goals derived from Guideline 1. Goals and soft-goals derived from Guideline 2 are grouped with respect to the elements identified from Guideline 1, since the latter are derived directly from the leading questions about consumer values. This grouping will enhance the identification of appropriate contribution links between goals and sub-goals. Issues to consider include:
If there are goals/sub-goals not relevant to those coming from Guideline 1, then one needs to identify/interpret how such goals/soft-goals can be related to the consumer value soft-goal, which will allow for an appropriate type contribution link (in Guideline 4).
If there are goals/soft-goals that appear to be more general than goals/soft-goals coming from the leading questions, then one should re-assess the grouping.
The outcome of this guideline is a complete set of grouped goals/soft-goals forming a hierarchical structure from the most general one (consumer value soft-goal) to concrete ones (ending leaf goals).
Guideline 4: Identify contribution links between intentional elements derived during the previous steps. In accordance with the i* guide, soft-goals should be ending leaves of a goal model , and therefore, soft-goals are decomposed through contribution links either further to other soft-goals or to goals. The ending leaves of this decomposition are goals as they can only be further decomposed into i* tasks, which are specific. The types of contribution links identified in our example (Figs. 5, 6) are motivated as follows:
Soft-goals directly contributing to the consumer value soft-goal are associated through an And Contribution Link because they altogether express the complete set of intentions based on the answers of the leading questions. An And Contribution Link implies that the parent is satisfied if all offspring are satisfied .
If there is only one goal/soft-goal contributing to another (from the same segment), then the association used is Make Contribution Link because it indicates a positive contribution enough to satisfied the soft-goal  (e.g., from goal “Layout be Customizable” to soft-goal “Good Layout be Designed” in Fig. 6).
If there is more than one goal/soft-goal contributing to another the association used is Some + Contribution Link because it indicates some positive contribution to satisfied a soft-goal but whose strength is not explicit . Therefore, using the Some + indicates that the sub-elements positively contribute to the soft-goal be satisfied but in an unknown strength (Figs. 5, 6).
After all contribution types have been identified, every group has been assess against the goal/soft-goals contributing to the consumer value soft-goal, since each group must have at least one goal/soft-goal contributing to the consumer value soft-goal.
The outcome of this guideline is a set complete set of goals and soft-goals associated with contribution links having only goals as ending leaves as shown in the SR diagrams presented in Figs. 5 and 6.
Figures 5 and 6 are parts of a complete i* SR model for the online education system and constitute input to the process for identifying domain and application features for the SPL, which is presented in the following section.
Identifying features and requirements for SPL
Once an i* SR model is obtained from a CPMM, the “system” is the actor representing future SPL. To link the i* SR model with a configuration for SPL, the theory of feature modeling is applied, where features are used as the basis for analyzing and representing commonality and variability of systems in a solution domain [35, 36].
For the elaboration of the requirements for SPL from an i* SRM diagram using features, the proposal concerning feature configuration based on stakeholder goals  is adopted because it is based on the intentional variability of goals coming the variation of different stakeholders’ goals. The reason for choosing this proposal stems from the fact that in the prior steps (Sect. 5.1) intentionality of consumer preference has been expressed in i* in the form of a hierarchical structure of goals and soft-goals under a generic consumer value soft-goal based on the mappings of CPMM to i*. Motivation for this choice has been based on the fact that consumer values are qualitative in nature and thus cannot be explicitly satisfied; rather, specifics of their satisfaction cannot be necessarily derived solely from values. This makes understandable why proposals such the G2SPL  is not suitable to derive feature models—the basis for such derivation under G2SPL are i* Tasks and Resources, in line with PRiM (Process Reengineering i* Method) , and not goals. Liaskos et al.  have proposed a goal-based framework about preference variability on requirements which is focused on stakeholders’ preference goals in terms of quality desires and temporal preferences. While this work also acknowledges the influence of context when setting/defining priorities, it does not relate to feature models for SPL, and in terms of priorities the approach makes use of a given relative importance among preference goals. Priorities coming from CPMM could be input to this approach, however, not within the context of SPL.
With respect to this proposal, goals are mapped to system features and soft-goals are used to generate qualitative constraints for the feature model. Within the scope of our work, features derived from consumer values are considered as optional with respect to the core (domain) functionality of the system. However, once a consumer value is selected for consideration in a system, all derived features from this consumer value become mandatory, as they are required to satisfied the consumer value soft-goals.
In our example scenario, the mapping of goals into features and the derivation of feature constraints from soft-goals is summarized in Figs. 7 and 8, where features for each of the two segments are presented. Features derived from goals unique to Non-Masters Students are darkly shaded (Fig. 7), features derived from goals unique to Masters Students are lightly shaded (Fig. 8), and features derived from goals common to both segments are non-shaded (Figs. 7, 8).
The features identified for Non-Master students in Fig. 7 will be further used as the source for deriving low-level requirements for developing the online education product configuration for this consumer segment, while the features identified in Fig. 8 for Master students will be part of the other product configuration. To repeat, the choice of the products in the product line is done in Step 1 of the method by segmenting the consumers’ population using demographics and context of use in a desired way—according to the organization’s goals, to foster the grouping of dominant differences regarding the preferences, or in some other way. In Step 2, the selected products are configured with user specific features, using goal modeling.
Transforming the obtained feature models to the system requirements artifacts modeled with Use Cases involves the following activities:
For common features across the feature models, Use Cases are elicited from a common feature by creating a corresponding Use Case Diagram and further documenting the interactions for each Use Case. The obtained requirements artifacts are valid for the entire product line and labeled as common, i.e., they complement the core functional modules of the line. The stakeholders involved in the elicitation of the use cases could be domain experts, and/or the representatives from the consumer segments. For instance, “Submitted Material Verification” (see the feature in Figs. 7, 8) will be used to derive the Use Cases to, for instance, choose a verification method, perform the verification, and present the outcome.
For those features specific to one or more product, configurations are transformed to Use Cases similarly to the previous alternative; however, apart from the domain experts, the consumer stakeholders are chosen from the segments requiring the features. The obtained requirements artifacts represent variability in the line and must be labeled accordingly, i.e., to complement the core and common functionalities for the products containing those features. An example of a feature specific to a product is “Subtitled Videos” in Fig. 7.
In addition to the above guidelines, it is needed to decide how the alternative and conflicting preferences will be handled in the process, when such are elicited during the goal modeling. Both behaviors may be found in a single segment (product) and can either be resolved based on the organizational preferences in the goal models itself or upon feature modeling, i.e., when documenting Use Cases for development.
Apart from the qualitative measures of Consumer Values used in Step 2 to elicit the intentions and the features for different product configurations in the line, the quantitative measures of Basic Values enable setting the rankings on the qualitative ones to use them as prioritizations in the development of products. To repeat, Schwartz’s theory defines values as desirable, trans-situational goals, varying in importance, that serve as guiding principles in peoples’ lives. The ranking of these values via the PVQ (see Sect. 3.2) is an important aspect of how they can be utilized within this method for the prioritization of system features discovered during requirements elicitation. Below a description of one possible method for accomplishing this is shown.
The Kano model  was developed to categorize the attributes of a product or service, based on how well they are able to satisfy customers’ needs. Eliminating problems and failures can be linked to expected (basic) requirements. Figure 9 illustrates the Kano model.
“Basic” (i.e., must-be or expected) attributes are those requirements which often go unnoticed by most customers, since customers expect these requirements to be met in the product or service , but their absence is very dissatisfying. “Performance Needs” (i.e., “One-dimensional”) attributes are also termed “more is better” but could also be “faster is better” or “easier is better”. “Delighter” (i.e., “Attractive”) attributes are beyond customers’ expectations. Their absence does not dissatisfy customers but rather their presence excites them. Each in turn has a measure of success: Threshold, Performance, and Excitement Attributes, respectively. These categories are discovered through qualitative analysis of customer requirements, where an evaluation table is created containing data for each of the attributes described above. Customers complete the table to rank those requirements, which populate the model as shown in Fig. 9.
Numerous quantitative extensions to Kano have been proposed within the field of satisfaction and customer requirements (S-CR) among them [46–48]. For example, Matzler et al.  propose the development of a questionnaire that relates the specific needs that the product and features under consideration must address to engage a customer. While this does not directly address values as understood by Schwartz, it could be easily appropriated for the purposes of the present method.
Recalling the description of the PVQ in Sect. 3.2, Schwartz’s values include a division between positive and negative values, where positive associations shift to more negative associations. This can be considered as the inflection point at Performance Needs, whereupon either Delighters or Basic Needs can be addressed. Lower weights, which indicate a high degree of connectedness to a value, indicate in Schwartz a higher affinity to that value, and when applied to the current process, a higher priority. Using this inflection point, values of 1–3 or higher are granted a higher priority than those greater 4–6. A second refinement is possible within the sub-areas of the model in which the same values provided by Schwartz are utilized for additional prioritization (see Sect. 6, Preference Capture). Finally, CPMM can be directly applied to Kano  for additional refinement. For example, the Basic Requirements, described as implied, self-evident, and taken for granted  are identical to those of Maslow’s Physiological Needs as shown in Sect. 3.1.