Collaborative engineering decision-making for building information channels and improving Web visibility of product manufacturers

Advanced Engineering Informatics 38 (2018) 264–276

Contents lists available at ScienceDirect

Advanced Engineering Informatics journal homepage: www.elsevier.com/locate/aei

Collaborative engineering decision-making for building information channels and improving Web visibility of product manufacturers

T

⁎

Sylvain Sagota, , Alain-Jérôme Fougèresb, Egon Ostrosic a

ELLIADD, ESTA, School of Business and Engineering, Belfort 90000, France ECAM Rennes, Louis de Broglie, Campus de Ker Lann, Bruz, 35091 Rennes, France c ELLIADD, Université de Bourgogne Franche-Comté, UTBM, Belfort 90010, France b

A R T I C LE I N FO

A B S T R A C T

Keywords: Information channel Web visibility Search Engine Optimization (SEO) Multi-agent systems Decision support systems

Product manufacturers have spent the last years improving productivity and process efficiency in order to face increasingly competitive markets. Today, the visibility of technological innovations has become essential to achieve the targeted market. It is now very difficult for a product manufacturer to reach customers without owning a website that is visible on search engine results pages. The goal of this paper is to build information channels between a company and its customers through improving both a company’s content of information on the Web and its website rank on the Internet through search engine results pages. Company information and knowledge are distributed through multiple stakeholders. The problem of building information channels between a company and customers is solved through a collaborative and distributed approach, on the one hand, and is supported by decision-making tools, on the other hand. The paper proposes an engineering model for building information channels and improving the visibility of the company on the Web. Agents are used for the implementation of the approach. The proposed model and its implementation handle the requirements, constraints, functions and solutions for improving Web visibility. The prototype tool, called CAWIS (Computer Aided Web Information Sharing), examines Web visibility in real time and evaluates the performance of the proposed content of information. CAWIS allows an exploratory and open way for building information channels and improving the visibility of product manufacturers on the Web.

1. Introduction The Internet affects the way people buy, enquire and express their opinions about various products. This transition from traditional business to e-business has forced product manufacturers to change their visibility strategy. Customers are now online, they use search engines to find products that meet their requirements and express their opinions publicly. To take advantage of this, product manufacturers have to have an online presence through a website presenting their expertise and products’ characteristics. However this is not sufficient; to be visible on the Internet the website also has to be present on search engine results pages and, ideally, in the top positions. Indeed, information overload has made consumers highly selective [1]; in this way, Internet users are only between 5% and 10% to consult the second page of search engines [2]. Potential customers prefer to reformulate their query rather than looking at the other results pages. To improve the position of their websites, product manufacturers can apply search engine optimization (SEO). By using some techniques

⁎

touching the source code and the website environment, SEO permits a ranking improvement. The objective is to rank a website on a list of keywords in order to attract qualified traffic through these new information channels. The final goal is to generate new contacts by promoting product expertise and technological innovation in order to acquire the voice of customers. Indeed, the design and manufacturing of the products in a company are results of collaborative and distributed processes. Distributed stakeholders work and interact actively on the design and manufacturing of the product according to the aspects and constraints of their domains [3]. In this way, the improvement of product visibility on the Web permits the gathering of online customer reviews and comments, which are useful for stakeholders in identifying their products’ potential enhancement [4], and can motivate potential customer decisions. A webmaster in charge of the SEO strategy faces a number of problems: search engines frequently update their search algorithms in order to always propose relevant results to the users, ranking models are not made public and the SEO process is unstructured. Moreover,

Corresponding author. E-mail address: [email protected] (S. Sagot).

https://doi.org/10.1016/j.aei.2018.07.003 Received 4 November 2017; Received in revised form 12 June 2018; Accepted 13 July 2018 Available online 27 July 2018 1474-0346/ © 2018 Elsevier Ltd. All rights reserved.

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

are effective, will have to meet the stakeholders’ needs. The optimization of on-site criteria and off-site criteria is a fundamental step of increasing visibility [10]. Since the existence of search engines, researchers [11,12] have applied several studies on SEO criteria. They identify some SEO criteria touching the website content, structure, and source code called “on-site criteria”. Improving on-site criteria is necessary to obtain a good ranking based on a specific keywords characterized by a high content of information. Keywords have to be relevant and in agreement with the page’s theme. The targeted keywords also have to be present in the text content but also in the HTML source code of the webpage. The 〈TITLE〉 tag, for example, has to be unique for each webpage and must be completed using targeted keywords. It should also not be longer than 70 characters. The 〈H1〉 tag is also important, and contrary to the 〈TITLE〉 tag, it is visible on the webpage. The presence of targeted keywords at the beginning of these tags is preferable for good results. Off-site criteria are criteria related to the website environment and popularity. The criterion “website popularity” which depends on the number of inbound links, also called backlinks, can be highlighted. The more a website acquires links from other websites, the more it will improve its ranking [13,14]. This criterion was first used in the Google ranking algorithm [15]. But this is not sufficient, as the quality of backlinks is also considered. Indeed, the theme of the website from which the backlinks are coming from has to be the same as the theme of the website that receives the backlink. A backlink between websites with different thematic content is not considered a good backlink by search engines, consequently there is no improvement in ranking. In fact, if the quality of backlinks is not good, the ranking of the website could decrease [16,17]. The creation date of the website is also important. Search engines give advantage to websites that have a long history. Moreover, after the optimization implementation, the ranking of an older website improves faster than a new one. The SEO process is the process that permits optimization of the ranking of a website on search engines while responding to the stakeholders’ requirements; it is a process that can be repeated. A SEO project is different from the SEO process because it is unique and temporary; it has a starting and an ending point [18]. In a manufacturing company, the insourced SEO process includes at least two types of human actors: the webmaster and the stakeholders working on product design and manufacturing. Other forms of organization can be found in the SEO process. In the case when a company outsources its SEO process, the involved human actors are the SEO practitioner and the client [9]. The SEO process is not clearly formalized and controlled [19]. In this way, there is no structured engineering method to conduct this activity. Moreover, webmasters adapt their work by using their own experience. Several authors proposed tools facilitating the SEO process. A system to predict ranking [20]; a content editing system working on relevant keywords collecting [21]; and a monitoring system to support SEO strategy [22] are proposed. However, these tools are only dedicated to SEO experts and neglect the stakeholder’s role and requirements. Therefore, there is a necessity to design the SEO process such as to support the decision making while integrating the stakeholders’ needs. Consequently, the development of a collaborative decisionmaking tool becomes also necessary [23,24]. In the insourced SEO process, the stakeholders are the people who express the need. According to future product development, the stakeholders give the webmaster a list of keywords to optimize. Additionally the role of stakeholders is to validate the actions to be implemented and to confirm the success of the SEO project. The stakeholders are not greatly involved in the SEO process because the SEO process is complex and requires a SEO expert assessment. Moreover, the knowledge difference between these human actors makes the exchanges and collaboration during the SEO process difficult [3]. Indeed, the webmaster has to satisfy the stakeholders’ needs and respect the search engine rules, which are sometimes in conflict. Observations made on the webmaster SEO work shows that non-experienced

collaboration between the webmaster and stakeholders working on product design and manufacturing is difficult due to knowledge differences. Finally, SEO results take time and are uncertain while product life cycle quickens and product manufacturers cut development time [5]. This leads to a situation where new innovative products are unknown because they cannot be found in time by using the right keyword on search engines. For all these reasons, the SEO process is considered complex [6]. The goal of this paper is to build information channels between a company and customers through improving both a company’s content of information on the Web and its website rank on Internet search engine results pages. Company information and knowledge are distributed through multiple stakeholders. The problem of building information channels between a company and customers is solved through a collaborative and distributed approach, on the one hand, and is supported by decision-making tools, on the other hand. The paper proposes an engineering model for building information channels and improving the visibility of the company on the Web. The formalization of different steps of the SEO process through different bridged models creates consistency towards a collaborative SEO process assisted by computer Web information sharing tools. Agents are used for the implementation of the approach. The proposed model and its implementation handle the requirements, constraints, functions and solutions for improving Web visibility in real time. The prototype tool called CAWIS (Computer Aided Web Information Sharing), examines Web visibility and evaluates the performance of the proposed content of information. CAWIS allows an exploratory and open way for building information channels and improving the visibility of product manufacturers on the Web. In addition, CAWIS allows a better control of the SEO process. The proposed model and the implementation through agent technology permitted the decomposition of the SEO process in several entities. Interactions that took place in this process allow the building of a collaborative and intelligent approach. In the proposed approach, the solution is the result of collaboration between the webmaster and stakeholders. It satisfies stakeholders’ requirements as well as Web constraints that include: competition, technological evolution, and algorithm changes. The prototype tool CAWIS implementing the approach allows building bidirectional information channels between a company and customers as well as improving the visibility of the company through the ranking of a product manufacturer website. The paper is structured as follows. Section 2 presents the literature review. Section 3 describes the proposed approach, including the use of an engineering meta-model to define and describe the SEO process, and the modeling of a multi-agent system to assist the SEO process. Section 4 presents a case study to validate the approach. The discussion and conclusion present and analyze the findings. 2. Literature review In this section, an overview of different techniques used to improve the visibility of product manufacturers on the Web is provided and the engineering research gap is highlighted. Different techniques have been developed to improve website visibility on search engines, such as SEO (search engine optimization). SEO [7] consists of optimizing some technical criteria (off-site and on-site) according to the search engines’ algorithm evolution to be visible in the organic results area. A product manufacturer can use SEO to maximize the visibility of its website. The SEO implementation requires knowledge of ranking criteria [8]. Product manufacturers that do not have time or resources to apply SEO can resort to a SEO firm in order to improve their visibility in the organic results. On the contrary, product manufacturers that want a better control of costs and confidentiality as well as better reactivity in exchanges and the preservation of knowledge can insource their SEO activity to their webmaster [9]. However, this implies that the webmaster takes on extra work and requires knowledge she or he may not possess. Moreover, the technical optimizations proposed, even if they 265

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

selected keywords is generated. This first report will serve as a starting point for the SEO project. The ranking evolution (increase, decrease, or stagnant) identified in subsequent reports will permit the measure of success for the SEO project relative to the starting point. During this first step, human actors have little collaboration. Typically, the only exchange takes place if the webmaster needs to convince the stakeholders to modify the keywords list to achieve better search results. In few cases, if the stakeholders do not accept the webmaster’s advice, a conflict may occur. After the identification of the stakeholders’ requirements, the analysis of the website to identify the website technology (e.g., HTML5, Flash, etc.) or the used Web platform (e.g., Wordpress, Magento, Joomla!, etc.) is the next step. Sometimes the website does not permit the SEO implementation. For example, if the website is developed in Flash, which is incompatible with SEO [27], or if the Web platform does not allow modification of HTML tags, or if the source code access is protected. The best solution in such cases is to build a new website by keeping the same domain name in order to avoid losing history. Adaption to the website technology is an important SEO constraint. During this second step, if the website is compatible with the SEO techniques implementation, the webmaster directly moves on to the next step; there is no exchange with the stakeholders. If the website does not permit the SEO techniques’ implementation, the webmaster has to convince the stakeholders to build a new website. If the stakeholders do not accept building a new website, a conflict may occur. When the website is identified as being compatible with SEO implementation, the best actions to optimize the ranking should be defined. This work does not concern the stakeholders because it requires an SEO expert assessment. The actions should consider the stakeholders’ needs, the search engine rules, and the stakeholders’ modification acceptance degree. During this third step, the stakeholders can accept or reject actions to implement. If the stakeholders accept, there is no exchange between human actors. If the stakeholders do not accept, the results can be deceptive; therefore, the webmaster has to convince the stakeholders of the importance and effectiveness of his future actions. If the proposed actions have been accepted or a consensus has been reached, implementation of these actions on the website is the next step. This work consists of implementing on-site criteria (e.g., insert keywords in HTML tags and text content, create optimized text content, create new webpages, change the website structure, etc.) or off-site criteria (e.g., create quality backlinks from relevant websites, create a blog, create social media pages, etc.). When all the actions have been implemented, a final check (operational acceptance testing) to verify that there are no problems or mistakes preventing ranking improvement is done. During this fourth step, there is no exchange with the stakeholders. After the implementation of actions into the website, the next step is monitoring. The evolution of ranking has to be reported to the stakeholders in order to measure the effectiveness of the work. For this step, a keywords ranking report at a frequency of one per month, which is an adequate frequency to perceive the quality for ranking modification, is produced. If the results obtained are far from those envisaged at the outset, new actions should be proposed. Modification of the initial keywords by introducing new keywords with a greater potential is an option. During this step, if results are good, exchanges between human actors stop until the next SEO project. If actions do not provide expected results, the SEO process begins a new cycle. In summary, Fig. 1 shows human actors involved in an empirical SEO process and its five steps: (1) needs identification, (2) study of the project feasibility, (3) actions proposal, (4) actions implementation, and (5) impact measurement. It can be noticed that collaboration between human actors only takes place during steps 1 and 5. Indeed, steps 2, 3 and 4 require SEO expertise that only the webmaster has, resulting in the absence of collaboration between human actors, which could impact the SEO project

stakeholders could negatively impact the success of a SEO project. Indeed, they almost always chose keywords that are too generic (with a high competition) and with a low content of information (e.g., a ball bearing manufacturer chooses the keyword “new ball bearing”, a tires manufacturer the keyword “innovative tire”). In this way the results could be deceptive (i.e., not as good as expected). In addition, nonexperienced stakeholders could refuse to implement the SEO webmaster's recommendations (e.g., by not adding text and keywords in webpages) by thinking that these kinds of modifications will negatively affect a website visitors’ product perceptions. In such cases, the website will be poorly optimized. In few cases, when the stakeholders are not experienced and when human actors do not reach a mutual arrangement, conflicts may even occur. Finally, the stakeholders’ experience has a high impact on the consensus with the webmaster, which will have consequences on the SEO project success. The literature review shows that building bidirectional information channels between a company and customers and improving the visibility of a company can be done through the SEO process. However, today, SEO can be summarized in a set of empirical techniques. Overall, the solutions are empirically found based on the experience of the webmaster by involving a lot of trial and error. The SEO process lacks underlying scientific and engineering principles. Therefore, this paper proposes that building bidirectional information channels between a company and customers can only be done through the design of a collaborative SEO process. The design of a collaborative SEO process, as an interplay between what should be achieved and how it could be achieved [25,26] should allow the development of good Web solutions based on collaborative decisions between the webmaster and stakeholders. Therefore, the design of collaborative SEO process should explicitly consider the customers’ and stakeholders’ requirements, the functions of the SEO process, and the dynamic constraints of the Web (webmaster constraints). It should allow good Web solutions for building bidirectional information channels and increase the visibility between a company and the customers; collaborative engineering and decision tools should support the collaborative SEO process. This paper tackles these issues. 3. Methodology To answer the research question, the study was divided in three parts. Firstly, observations were made on the working methods used by webmasters on their stakeholders’ projects. Secondly, an engineering meta-model permitted to formalize the SEO process. Thirdly, the use of software agents allowed to model the SEO process. The connection between observations and engineering theories can offer a consistent method to develop and test the models. 3.1. Observations During three years in a company, observations were made on the working methods used by webmasters on their stakeholders’ projects [8]. The stakeholders can be represented by several human actors (e.g., manager, product designer, product engineer, innovation manager, technician, etc.) and the webmaster often works in a team composed of programmers, community managers and copywriters. The observation permits to identify five important steps in the SEO process and highlights the level of collaboration between human actors. A usual SEO process begins with the identification of the stakeholders’ requirements. A list of keywords that need to be optimized is defined. The keyword selection is the key of a good SEO strategy. In fact, result time could be considerably reduced if the keyword selection is relevant. If the chosen keywords are too generic (i.e., relative to competitors) or not in agreement with the keyword theme of the website, the webmaster could propose to the stakeholders to modify the list. Once the keywords have been chosen, the first report of keyword rankings to observe the current position of the website based on 266

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

Fig. 1. Empirical SEO process.

success. This is accentuated by differences in knowledge between human actors, particularly when the stakeholders are inexperienced. Indeed, the success of a SEO project requires taking into account the stakeholders’ experience in order to avoid conflicts and to provide good results. To solve this problem, collaboration between human actors has to be efficient. Thus, the development of an approach permitting human actors to efficiently collaborate during the SEO process is necessary [28].

formula; e.g., “a webpage with a keyword density between 3% and 8% could win two positions in one month on this keyword.” ● Computational model: it is the representation of a mathematical model from a computational point of view; it can be a monitoring tool permitting to study the ranking evolution of keywords. ● Experimental model: its results are observations or tendencies; it can be an experiment made on several webpages to compare the influence of a SEO criterion on ranking.

3.2. An engineering meta-model building

The four domains of the engineering meta-model are the following:

As presented previously, there is a lack of collaboration between human actors during the SEO process. Developing computer aided Web information sharing tools allows improving the collaboration between actors as well as decision-making. The creation of an innovative approach [29] that improves collaboration between human actors and support decision making in an upgradable environment is typically a design problem. Therefore, the formalization of the different steps of the SEO process through different bridged models create a consistent ground towards a collaborative SEO process assisted by Web information sharing tools. An engineering meta-model derived from product design [30,31] is used in this study because it is complete and allows the flexibility of bridging. It covers the design domains as well as the engineering models. This meta-model presented in Table 1 is structured in four models and four domains. The four models of the engineering meta-model are the following:

● Customer domain: it expresses a stakeholders’ need; e.g., “We want our website to reach the top five results based on that keyword.” ● Functional domain: processes are designed using functions; e.g., “to increase rank of webpages, to increase the website visibility on search engines, etc.” ● Physical domain: solutions that respond to the functional domain; e.g., “insert relevant keywords, improve the page load time, etc.” ● Process domain: constraint elements linked to the physical domain; e.g., “to be adapted to criteria impact evolution, to be adapted to the website changing.” To formalize the SEO process, some parts of the presented engineering meta-model were selected in accordance with observations done in the company [8]. The SEO process identification is illustrated (Fig. 2), where each numbered step is described below: Step (1), Needs identification: The stakeholders start to formulate their requirements (e.g., “We want our website to reach the 3rd position on this keyword”). This step is purely conceptual and it only belongs to the customer domain. Step (2), Study of the project feasibility: This step is done following several iterations with the stakeholders in order to understand the needs precisely. After this, the webmaster identifies constraints (e.g., “be compatible with the technology used by the website”) in order to study

● Conceptual model: it represents a concept or rules, in the case of the SEO process it can be the expression of a SEO technical concept; e.g., “a webpage which contains important keywords in the full text and the 〈TITLE〉 tag has more of a chance to improve its ranking than a webpage which has none.” ● Mathematical model: it expresses itself from a mathematical point of view, in the case of the SEO process it can be a mathematical Table 1 Engineering meta-model proposed to model the SEO process.

Conceptual model Mathematical model Computational model Experimental model

Customer domain

Functional domain

Physical domain

Process domain

Conceptual model of customer domain Mathematical model of customer domain Computational model of customer domain Experimental model of customer domain

Conceptual model of functional domain Mathematical model of functional domain Computational model of functional domain Experimental model of functional domain

Conceptual model of physical domain Mathematical model of physical domain Computational model of physical domain Experimental model of physical domain

Conceptual model of process domain Mathematical model of process domain Computational model of process domain Experimental model of process domain

267

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

Fig. 2. Identification of the SEO process steps by using an engineering meta-model.

scale of agents has been defined: reflex-based agents, rule-based agents, and knowledge-based agents. The proposed SEO agents to model the collaborative and intelligent approach of SEO process are a part of the second level. These agents belong to one of the four communities: Cr, the Requirement agents’ community; Cf, the Function agents’ community; Cc, the Constraint agents’ community; and Cs, the Solution agents’ community. A system of SEO agents MSEO is defined by the following 4tuples (1):

the feasibility of the SEO project. The stakeholders do not participate in the identification of constraints. Finally, the stakeholders’ requirements are transformed into functions. During this conceptual step, only the functional domain specific to the functions and the process domain specific to the constraints are concerned. Step (3), Actions proposal: When all functions are identified, the webmaster finds technical solutions. They can belong to the four models discussed (i.e., Conceptual, e.g., the webmaster identifies a list of concepts and technical rules to be implemented, Mathematical, e.g., the webmaster creates an equation combining the needs of the stakeholders and the results to obtain, Computational: e.g., the webmaster adapts the HTML code of the website, and Experimental, e.g., the webmaster exposes experimental results to the stakeholders that can be chosen to apply to the website). During this step, only the physical domain specific to the solutions is concerned. The solutions may belong to all models. Step (4), Actions implementation: Once the stakeholders have validated the solutions proposed by the webmaster, they are then implemented in the website. This implementation step always concerns the physical domain specific to the solutions, but the implementation of the actions only affects the experimental model. Step (5), Impact measurement: The impact measurement allows human actors to validate the implemented actions. This step remains in the physical domain specific to the solutions, but the measurement of the results concerns only the mathematical model. Finally, at the end of the SEO process, if the work does not give the expected outcomes, the webmaster should reevaluate all the steps. During these five steps, the webmaster’s goal is to propose effective solutions permitting a quick improvement of the website ranking while limiting the number of exchanges. In this way, the SEO process can be considered a dynamic process.

MSEO = 〈A, I , P , O〉

(1)

where A, I , P , and O are respectively the set of SEO Agents, the set of possible Interactions between SEO agents, the set of Roles played by SEO agents, and the Organization of SEO agents formed by the four communities Cr, Cf, Cc, and Cs. A SEO agent αi ∈ A is defined by (2):

αi = 〈ΦΠ(αi) , ΦΔ(αi) , ΦΓ(αi) , K αi 〉

(2)

where ΦΠ(αi) , ΦΔ(αi) , and ΦΓ(αi) are respectively the functions of observation, decision, and action, of the SEO agent αi . The set of knowledge K αi of the SEO agent αi includes the values of the domain (e.g., current values of criteria), network of affinities with other SEO agents (e.g., relationship with the agents of his community or those of the other three agent communities), dynamic knowledge (e.g., observed events in a webpage, move of internal states, etc.), and fuzzy decision rules that are described later in the Eqs. (5) and (6). The proposed model in the Eq. (2) permits to define an agent continuous behaviour as illustrated in Fig. 3: SEO agents can interact with human actors or other agents. An

3.3. Agent-based modeling The engineering meta-model presented in the previous section is composed of four domains: customer domain, functional domain, physical domain, and process domain. The five steps of the SEO process identified are included into these four domains. This distribution and cooperation between these different entities permits us to model the SEO process by using a multi-agent system, which is a computerized system composed of multiple interacting intelligent agents within an environment [32–35]. Indeed, thanks to the agents’ properties (distribution, adaptability, autonomy, cooperation, and flexibility), agent-based approaches allow the solving of distributed problems [36,37] and facilitates exchange of information. Moreover, researchers [38–40] have already suggested a multi-agent model that is adapted here in order to design the proposed collaborative and intelligent computing approach for the SEO process. To model complex systems by using software agents, a three-level

Fig. 3. Agent decision making process. 268

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

Fig. 4. Collaborative and intelligent agent-based SEO process.

interaction ιi ∈ I between a source and one or more destination SEO agents is defined by the Eq. (3):

ιi = γc (αs, A′ , τj, (π1, …, πm))

(3)

(2)

where γc ∈ Γ is a cooperative act (nine cooperative acts are defined in the model for collaborative and intelligent SEO process: Γ = {inform, reply, diffuse, propose, declare, ask, accept, refuse, acknowledge}); αs ∈ A and A′ = {α d1, …, α dn} ⊆ A , are respectively the source SEO agent and the n destination SEO agents; τj is a type of action that destination SEO agents are able to perform (for example: ADD, RANK, TIMELIMIT, APPEARANCE, …), and π1, …, πm are the m parameters of the action referred by τj . The model for the collaborative and intelligent based agent SEO process is presented (Fig. 4). As has been specified before it is composed of the four agents’ communities: 〈Requirement〉, 〈Function〉, 〈Constraint〉 and 〈 Solution〉 where interactions take place. These interactions could occur between human actors, between software agents from the same community (intra-communities interactions), or between software agents from other communities (inter-community interactions). The five steps of the SEO process previously presented in Section 3.2 are implemented in the model: (1) Needs identification, (2) Study of the SEO project feasibility, (3) Actions proposal, (4) Actions implementation, and (5) Impact measurement. The intelligent SEO process involves the exchanges of messages between human actors and software agents. To illustrate this process, a basic scenario is presented below. Table 2 describes the different messages exchanged during this scenario.

(3)

(4) (1) Needs identification: the stakeholders ask the webmaster to improve website ranking on the keyword KWi. The stakeholders want the website WS to reach the first results page of the search engine within six months without altering the visual appearance more than 30%. The webmaster suggests to work on a specific page Pj because he considered it as the most suitable page to improve the ranking of

(5)

the keyword KWi on the website WS. Indeed, the theme of the page Pj is similar to one of the keyword KWi. The respect of the theme is an important element to get good results on search engines. Study of the SEO project feasibility: the webmaster knows which type of optimization could be implemented on the stakeholders’ website WS because this Web platform is often used: the SEO project is considered realizable. The webmaster declares the message “The optimization must be compatible with the Web platform” to the constraint agents. Then, the constraint agents inform the solution agents. The webmaster also notices that the keyword KWi does not repeat enough in the text content and in HTML tags. So he declares the message “Add the keyword KWi in the page Pj” to the requirement agents. Moreover, the stakeholders want to have results in six months at a maximum. Thus, the webmaster declares the message “Respect the maximum project time duration of six months” to the requirement agents. Finally, the stakeholders want the design of the page Pj to not be altered over 30% after the implementation of SEO actions. This means, the webmaster declares the message “The visual appearance change of the page Pj shall not exceed 30%” to the requirement agents. After this, the requirement agents inform the function agents, then the function agents inform the solution agents. Actions proposal: the solution agents inform the webmaster about the solutions that were found (e.g., “repeat the keyword KWi three times in the 〈TITLE〉 tag of the page Pj”). Then, the webmaster proposes this solution to the stakeholders. The stakeholders can accept or refuse to implement this solution. Actions implementation: if the solutions are accepted by the stakeholders, the webmaster implements them within the website WS. Impact measurement: at the end of the SEO project period (e.g., “six months”) the webmaster and the stakeholders monitor the results. If the results are not expected or desired, the process begins again. In this scenario, it can be noticed that messages are only exchanged

269

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

Table 2 Typical exchanged messages between human actors and software agents during the SEO process. Source Stakeholders (human actor) Webmaster (human actor) Webmaster (human actor)

Requirement agent (rs)

Function agent (fs)

Constraint agent (cs) Solution agent (ss) Webmaster (human actor) Stakeholders (human actor)

Message

Destination

The stakeholders ask the webmaster to rank the website WS on a new keyword KWi declare(Webmaster, cd, PLATFORM, (WS, KWi)) e.g., the optimization must be compatible with the Web platform declare(Webmaster, rd, ADD, (KWi, Pj))e.g., add the keyword KWi on the page Pj declare(Webmaster, rd, TIMELIMIT, (Td = TdMax)) e.g., the maximum project time duration (Td) is six months (TdMax) declare(Webmaster, rd, APPEARANCE, (Pj, Vac⩽VacMax)) e. g., the visual appearance change (Vac) of the page Pj shall not exceed 30% (VacMax) inform(rs, fd, ADD, (KWi, Pj,) e.g. inform the function agent to create the keyword KWiinform(rs, fd, TIMELIMIT, (Td = TdMax)) e.g. inform the function agent about the maximum project time durationinform(rs, fd, APPEARANCE, (Pj, Vac⩽VacMax)) e.g. inform the function agent about the limit in the visual appearance change inform(fs, sd, ADD, (KWi, Pj,))e.g., inform the solution agent to add KWi on the page Pj inform(fs, sd, TIMELIMIT, (Td = TdMax)) e.g., inform the solution agent about the maximum project time durationinform(fs, sd, APPEARANCE, (Pj, Vac⩽VacMax)) e.g., inform the solution agent about the visual appearance change limit inform(Webmaster, sd, PLATFORM, (WS, KWi)) e.g., inform that the optimization must be compatible with the Web platform reply(ss, Webmaster, ACTION, (Sk, Pj)) e.g., reply that the Sk solution is possible on the page Pj The webmaster proposes the Sk solution to the stakeholders The stakeholders accepts the Sk solution proposed by the webmaster

from steps 1 (needs identification) to 3 (actions proposal). Indeed, steps 4 and 5 contain actions and monitoring. In the proposed model, the Page and Criterion are considered solution agents because the majority of solutions could be implemented by improving the SEO criteria on webpages. A Page could interact with Onpage Criteria (e.g., number of keyword repetitions in the full text, 〈TITLE〉 tag length, etc.) and Off-page Criteria (e.g., number of inbound links, number of Facebook likes, etc.). To improve the Page ranking, the webmaster improves these pages criteria. A solution agent (Page or Criterion) shares the knowledge of the other solution agents (Page or Criterion). A solution agent also knows its selfcharacteristics. After studying the literature [13,20,41], 15 criteria that can influence behaviors of the page towards the best ranking were distinguished. These 15 criteria are made into agents. The set of these Criterion agents is shown in Table 3: According to the literature, each Criterion agent has an influence on the ranking process. Three kinds of influences are distinguished: excellent, medium, or bad. In order to adapt the approach to the constant search engine algorithm evolution, the fuzzy logic [42] is chosen to manage these influences. Indeed, SEO criteria values (excellent, medium and bad) cannot be static (as Boolean logic) but constantly evolve according to the search engine algorithm updates. The fuzzy logic representation allows the consideration of evolving values [43] by adjusting the influence of criteria to overcome uncertainty of the ranking process. Fig. 5 shows the three membership functions corresponding to fuzzy sets: “bad” {0;12 months}, “medium” {4;34 months}, and “excellent” {24;40 months} of the impact of the criterion c1 (domain’s age) on the ranking. After having identified the membership functions of each Criterion to the three fuzzy sets (bad, medium, and excellent), fuzzy decision rules were built. Indeed, solution agents (Page and Criterion) take decisions according to fuzzy decision rules contained in their knowledge [44]. Theses fuzzy decision rules are built based on observation on SEO criteria influence [45] and the study of the literature [13,20,41]. They are formalized as following (5):

IF {x1 IS i x1} …AND {xn IS i xn} THEN {y1 IS j y1 } …AND {ym IS jym }

Webmaster (human actor) Constraint agent (cd) Requirement agent (rd)

Function agent (fd)

Solution agent (sd)

Solution agent (sd) Webmaster (human actor) Stakeholders (human actor) Webmaster (human actor)

Table 3 SEO criteria list. Criteria

Name of criteria

c1 c2 c3 c4 c5 c6 c7 c8 c9 c10 c11 c12 c13 c14 c15

Page load score mobile Domain’s age Keyword first in 〈TITLE〉 tag Keyword first in 〈H1〉 tag Referring domains number 〈TITLE〉 tag length Number of Facebook likes Keyword repetitions in 〈H1〉 tag Meta description tag length Session duration Keyword repetitions in the full text Backlinks number Number of Google + 1 Bounce rate Keywords repetitions in 〈TITLE〉 tag

(6) shows an example of a fuzzy decision rule concerning the criteria c11 (number of keyword repetitions in the full text) and c15 (number of keyword repetitions in the 〈TITLE〉 tag).

IF {c11 IS excellent } AND {c15 IS excellent } THEN {Ranking IS excellent }

(6)

In this case, the ranking of the webpage is excellent (i.e., on the first results page) if the values of the criteria c11 and c15 are excellent. Examples of other fuzzy decision rules relative to criteria c11 and c15 are presented in Fig. 6: According to fuzzy decision rules, the solution agents can propose the best actions to the webmaster in order to improve the page ranking. Fig. 7 shows the decision surface graph derived from the membership functions and decision rules of criteria c11 and c15. The “hot zone” represents a good ranking. All the fuzzy decision rules were then incorporated in a prototype tool called CAWIS (Computer Aided Web Information Sharing) developed in Java and based on the proposed approach. The architecture of CAWIS is composed of:

(5) - Server: it catches every day criteria data on the website by using an HTML parser and several APIs (e.g., Ahrefs, MyPoseo, Whois, Google Analytics, PageSpeed Insights, etc.), then the captured data

where i xk , jyk ∈ I , I = {bad, medium, excellent} is the influence degree on ranking; xk, k = 1,n is an input element (e.g., criteria c1, c2, c3, etc.) and yk, k = 1,m is an output element (e.g., Ranking). The following Eq. 270

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

Fig. 5. Membership functions of the impact of the criterion c2 (domain’s age) on ranking.

Fig. 6. Examples of fuzzy decision rules (criteria: c11 and c15).

stakeholders), it displays the SEO engine processed dataFig. 8 illustrates the architecture of the prototype tool CAWIS. 4. CAWIS for decision-making In this part, a case study to analyze the implementation of a SEO project with the help of CAWIS (Computer Aided Web Information Sharing) is presented. The objective was to improve the ranking of a smartphone product manufacturer website in the Google.com search engine. At the beginning of the SEO project, the webmaster and the stakeholders agreed on a keywords list noted KWList, which contains three keywords defined as follows (7):

KWLIST = {KW1, KW2, KW3}

(7)

where KW1=“ultra slim smartphone”, KW2=”slimmest phone”, KW3=“thinnest mobile phone”. The keywords choice was approved by CAWIS that checks the keywords detected in the website, their competition level, and their search volume level. The URL of the website also permitted CAWIS to check if the website was compatible with SEO optimizations regarding the technology used by the website (e.g., Web platform, developed from scratch, etc.). All these elements permitted the webmaster to approve the start of the SEO project. After the website’s compatibility confirmation, CAWIS, used as a decision support system, selected for each keyword the best-ranked page on the website and it instantaneously provided recommendations to implement in order to improve the ranking. Indeed, CAWIS identified SEO criteria that have a bad, medium, and excellent impact on the

Fig. 7. Surface graph of ranking (c11 and c15).

are saved in a “criteria” database - SEO engine: it processes data from the “criteria” database according to fuzzy decision rules from “decision rules” database. - Client⁎: it includes two interfaces (i.e., webmaster and

⁎ The client–server model is a distributed application structure that partitions tasks or workloads between the providers of a resource or service, called servers, and service requesters, called clients.

271

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

Fig. 8. Architecture of the prototype tool CAWIS.

Fig. 9. SEO audit for each criterion, CAWIS expert interface, keyword KW3.

red color can be optimized (CBO), the criterion c1 (Page load score mobile) in light green color is almost optimized (IAO), and finally, the criterion c2 (Domain’s age) in dark green color is already optimized (IO). According to CAWIS, to improve the ranking of the webpage (position 7 on the 11th of July) the webmaster should optimize the criteria c4 and c7 in priority, so he needs to add the keyword KW3 = “thinnest mobile phone” at the beginning of the 〈TITLE〉 tag and he has to improve the number of Facebook likes. After that, he has to improve all the criteria with the values “Must be optimized” (MBO) and “Can be optimized” (CBO) which are highlighted in red color in CAWIS.

ranking by using color (Fig. 9). Dark red color indicates criteria that must be optimized (MBO: MustBeOptimized), light red color indicates criteria that can be optimized (CBO: CanBeOptimized), light green color indicates criteria that are almost optimized (IAO: IsAlmostOptimized), and finally dark green indicates criteria that are already optimized (IO: IsOptimized). The linguistic values (MBO, CBO, IAO, and IO) of each criterion are given in Table 4. According to the results of Fig. 9 and Table 4, on the 11th of July, the criterion c4 (Keyword first in 〈H1〉 tag) in dark red color must be optimized (MBO), the criterion c7 (Number of Facebook likes) in light

272

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

Table 4 Criteria values and their influences on the ranking, keyword KW3. Dates

Ranking

c1

c2

c3

c4

c5

c6

c7

c8

c9

c10

c11

c12

c13

c14

c15

07/11/2017 07/10/2017 07/09/2017 07/08/2017 07/07/2017

7 6 6 7 6

IAO IAO IAO IAO IAO

IO IO IO IO IO

CBO CBO CBO CBO CBO

MBO MBO MBO MBO MBO

MBO MBO MBO MBO MBO

IAO IAO IAO IAO IAO

CBO CBO CBO CBO CBO

IO IO IO IO IO

CBO CBO CBO CBO CBO

IO IO IO IO IO

IO IO IO IO IO

MBO MBO MBO MBO MBO

CBO CBO CBO CBO CBO

IO IO IO IO IO

CBO CBO CBO CBO CBO

months after the actions implementation, the website ranking was improved on the keywords KW1, KW2, and KW3, indeed, all the keywords reached the first page of the search engine. Finally, the information channel created by the stakeholders and the webmaster permitted the customers to find the product manufacturer website on keywords that expressed the right content of information. The detailed agents’ actions performed during the case study are described in Table 6. During the needs identification, the requirement agents deliver the keywords list (e.g., KW1, KW2 and KW3) to the function agents, then function agents deliver it to the solution agents. After that, during the study of the SEO project feasibility, the constraint agents receive the constraints identified by CAWIS (e.g., Web platform: Wordpress) and they deliver it to the solution agents. During the actions proposal the solution agents propose to CAWIS to optimize some criteria according to their decision rules (e.g., c2: IO, c3: CBO, c4: MBO, etc.). Then, for the actions implementation, the webmaster implements the actions on the website and the solution agents are automatically informed. Finally, during the impact measurement step the webmaster and the stakeholders consult the monitoring view, agents are automatically updated.

Once the human actors take note of actions proposed by CAWIS, the webmaster can then begin to implement them into the website. The webmaster can be guided by CAWIS, which allowed him to avoid mistakes. For example, if the webmaster adds the keyword KW3 at the end of the 〈TITLE〉 tag, the criterion c4 will stay in red color. Indeed, to optimize the criterion c4 the keyword should be placed in first position in the 〈TITLE〉 tag. If the keyword KW3 has a correct location in the 〈TITLE〉 tag, CAWIS will change the color of the criteria c4 from red to green. So CAWIS permitted him to add the right volume of keywords on the relevant page and place. The webmaster instantaneously checked the actions implementation and the time where they were implemented on the website. Finally, the webmaster used CAWIS to make the operational acceptance testing in order to check if all the actions were correctly implemented on the website. After this correct implementation of SEO actions, the webmaster and the stakeholders consulted the “monitoring view” (Fig. 10) that illustrates the ranking of the three keywords (Table 5). The evolution of the keywords’ ranking can be followed. For example, on the 11th of May, the position of the keyword KW1 = “ultra slim smartphone” was 100, then on the 11th of July the position was 4. In the case study, two

Fig. 10. Ranking monitoring, CAWIS stakeholders’ interface. 273

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

Table 5 Ranking values evolution of keywords KW1, KW2 and KW3. Dates

KW1: “ultra slim smartphone”

KW2: “slimmest phone”

KW3: “thinnest mobile phone”

07/11/2017 07/10/2017 07/09/2017 07/08/2017 07/07/2017 … 05/14/2017 05/13/2017 05/12/2017 05/11/2017

4 4 4 4 4 … 3 3 8 100

2 2 2 2 2 … 100 100 100 100

7 6 6 7 6 … 100 100 100 100

Table 6 Agents’ actions performed during the case study. Steps

Agents action

Examples with formalization

1. Needs identification

The requirement agents deliver the keywords list KWLIST to the function agents. Then, function agents deliver it to the solution agents Constraint agents receive the constraints identified by CAWIS. They deliver it to the solution agents

KWLIST = {KW1, KW2, KW3}

2. Study of the SEO project feasibility 3. Actions proposal 4. Actions implementation

Solution agents propose to CAWIS to optimize some criteria according to their decision rules The webmaster implements the actions on the website

5. Impact measurement

The webmaster and the stakeholders consult the monitoring view

Web platform:Wordpress Competition level: high Search volume level: high Influence on the ranking: c1:IAO, c2:IO, c3:CBO, c4:MBO, etc The actions implementation is performed by the webmaster who inform the solution agents The affected agents values are updated

Fig. 11. New collaborative and intelligent SEO process assisted by CAWIS.

5. Discussion The development of CAWIS, which is based on the proposed approach allows a better collaboration between human actors, leading to an improvement in the success of information channels building between customers and stakeholders. Each step’s gain in efficiency of the SEO process is highlighted below:

•

• Needs identification: the stakeholders are now totally involved in this

•

first step of the SEO process. After discussing the core information of product manufacturers, the stakeholders and the webmaster converge towards a list of keywords, which express the content of information. Moreover, CAWIS gives them some advice in order to improve the SEO project success. The webmaster can also rely on 274

CAWIS to support his decision about the keywords selection. CAWIS allows needs identification and encourages stakeholders to collaborate with the webmaster. In this way, the SEO project has more chance for success. Study of the SEO project feasibility: CAWIS is able to make the difference between a website that permits the implementation of SEO recommendations and a website that does not. This function enables quick improvement by providing the stakeholders and the webmaster information on the feasibility of the future implementation. Actions proposal: from now on, the choice of actions to implement does not only depend on the webmaster's experience. Indeed, CAWIS identifies and prioritizes the actions to implement in order to improve the page ranking. Thus, it could assist the webmaster in critical decision-making situations. CAWIS also permits the

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

The proposed approach is based on the development of an engineering model for SEO process. The proposed engineering SEO model process explicitly considers the relationships between requirements, functions, solutions and constraints. Overall, the proposed approach improves the collaboration between human actors involved in the SEO process. In this way, the collaboration between the webmaster and the stakeholders become essential for the success of a SEO project in product design and manufacturing context. Until now, the stakeholders were not heavily involved in the SEO process. The proposed approach permits a reconsideration of the stakeholders’ role. Due to its strong involvement, the stakeholders are now an integral part of the SEO process. The modeling and the implementation of the proposed engineering model of SEO process use the knowledge from different human actors efficiently. Updates on fuzzy rules and criteria can easily be performed into an agent-based platform called CAWIS (Computer Aided Web Information Sharing). This permits an efficient support for decision making despite the constant changing of search engine algorithms. The proposition, validation, and implementation of the actions are more efficient. CAWIS, which is developed by using the proposed model, permits the webmaster and the stakeholders to be supported in the decision making and to not only use the webmaster experience. Moreover, it allows the webmaster to have a better control of different steps in the SEO process. This enables the webmaster to better structure tasks and accomplishes work more effectively. The proposed method considers the webmaster’s knowledge for the definition of fuzzy rules. The inference and learning of SEO fuzzy rules is a new direction for research that can be further explored in future work. An advanced SEO process should address the issue concerning culture-difference, in different geographical locations, for customer requirements acquisition. It constitutes a new research topic for the SEO process.

stakeholders to have a better perception of the actions to implement.

• Actions implementation: CAWIS allows one to check the correct im•

plementation of SEO actions. The webmaster has to implement manually the actions proposed by CAWIS (e.g., add text content, add keywords, add webpages, etc.). Impact measurement: the monitoring view provided by CAWIS permits to follow the evolution of ranking in real time. It permits the webmaster and the stakeholders to check the success of the SEO project.

Overall, the collaborative and intelligent computing approach implemented in CAWIS allowed better collaboration between human actors during the SEO process (Fig. 11). It also supported the human actors’ decisions through all the steps of the process, from the needs identification to the impact measurement step. The use of the fuzzy logic permitted us to efficiently adjust the decision rules and contributed to the evolution of CAWIS according to the search engine algorithm’s updates. To obtain efficient recommendations and results, it is important to frequently update the fuzzy decision rules and criteria in order to follow the evolution of search engine algorithms. CAWIS essence is to assist the various employees of a company in their SEO approach to improve the visibility of their products. But other product manufacturers may in the same time try to improve their ranking on the same keyword by using their own SEO approach. In that case a competition between different SEO approaches occurs. So the issue is: How can CAWIS help a company to deal intelligently with this competition? Currently, CAWIS permits a competitors analysis by studying input variables (criteria values) and output variable (ranking) of their websites. CAWIS cannot identify the SEO approaches of competitors but CAWIS can warn the webmaster dynamically if it considers that its own evaluation is not coherent with the ranking. In that case CAWIS could propose to extend the criteria panel or to update the decision rules. Actually, CAWIS is able to evaluate dynamically the impact of its own prediction. Finally, the developed approach allows building bidirectional information channels between stakeholders, webmasters and customers; the website becomes a real interaction node. Henceforth customers can directly interact to share their opinion on a product, thus product manufacturers can better work on the content of information to put forward. The resulting collaborative environment also permits the stakeholders to be involved in the SEO approach; it no longer solely depends on webmasters. The company can take advantages of the developed tool and does not necessarily depend on an external SEO firm. The SEO approach can be completely integrated into the product design process to improve the product reputation. All these benefits permit product manufacturers to improve their visibility on the Web and to intensify their proximity to their customers to become more efficient. Though the proposed approach integrates the customers into the product design development, the clear elicitation of multicultural factors amongst the customer needs should be integrated in the SEO process. It will help to determine the success level of specific new product development organization, its competitiveness and its benchmarking [46]. Therefore, an advanced SEO process should address the issue concerning culture-difference, in different geographical locations, for customer requirements acquisition. It constitutes a new research topic for the SEO process.

Acknowledgments This work was supported by a CIFRE convention of the ANRT (n°2013/0967) funded by the French Ministry of Higher Education and Research. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.aei.2018.07.003. References [1] P.A. Herbig, H. Kramer, The effect of information overload on the innovation choice, J. Consum. Mark. 11 (2) (1994) 45. [2] A. Chuklin, P. Serdyukov, M. de Rijke, Modeling clicks beyond the first result page, 2013, pp. 1217–1220. [3] E. Ostrosi, L. Haxhiaj, S. Fukuda, Fuzzy modelling of consensus during design conflict resolution, Res. Eng. Des. 23 (1) (Jan. 2012) 53–70. [4] J. Qi, Z. Zhang, S. Jeon, Y. Zhou, Mining customer requirements from online reviews: a product improvement perspective, Inf. Manage. 53 (8) (2016) 951–963. [5] C.-H. Chen, Y.T. Chong, W. Chang, W. Yan, A quality–time–cost-oriented strategy for product conceptualization, Adv. Eng. Inform. 26 (1) (2012) 16–25. [6] S. Sagot, A.J. Fougères, E. Ostrosi, Search engine optimization process: a concurrent intelligent computing approach, in: Proceedings of 22nd ISPE International Conference on Concurrent Engineering (CE), 2015, pp. 603–614. [7] N. Yalçın, U. Köse, What is search engine optimization: SEO? Procedia - Soc. Behav. Sci. 9 (2010) 487–493. [8] S. Sagot, Contribution à la conception et à la maîtrise du processus de référencement Web (Ph.D. Dissertation), Belfort-Montbéliard, 2016. [9] S. Sagot, A.J. Fougères, E. Ostrosi, Business constraints integration in a search engine optimization fuzzy decision support system, in: 2017 IEEE 19th Conference on Business Informatics (CBI'2017), vol. 01, 2017, pp. 302–311. [10] S. O’Neill, K. Curran, The core aspects of search engine optimisation necessary to move up the ranking, Int. J. Ambient Comput. Intell. IJACI 3 (4) (2011) 62–70. [11] J. Zhang, A. Dimitroff, The impact of webpage content characteristics on webpage visibility in search engine results (Part I), Inf. Process. Manage. 41 (3) (2005) 665–690.

6. Conclusion Building bidirectional information channels between a company and customers and improving Web visibility can be done only through designing a collaborative SEO process. In this paper, a collaborative engineering decision-making approach for building information channels and improving Web visibility of product manufacturers is proposed. The contributions of the proposed approach are the following: 275

Advanced Engineering Informatics 38 (2018) 264–276

S. Sagot et al.

Springer, London, 2013, pp. 545–556. [31] E. Ostrosi, F. Pfaender, D. Choulier, A.-J. Fougères, M. Tzen, et al., Describing the engineering modeling knowledge for complexity management in the design of complex city, in: DS 75-6: Proceedings of the 19th International Conference on Engineering Design (ICED13), Design for Harmonies, Vol. 6: Design Information and Knowledge, Seoul, Korea, 19-22.08.2013, 2013. [32] R.D. Beer, A dynamical systems perspective on agent-environment interaction, Comput. Theor. Interact. Agency 72 (1996) 173–215. [33] M. Wooldridge, N.R. Jennings, et al., Intelligent agents: theory and practice, Knowl. Eng. Rev. 10 (2) (1995) 115–152. [34] M. Wooldridge, Agent-based software engineering, Softw. Eng. IEEE Proc. 144 (1997) 26–37. [35] N.R. Jennings, On agent-based software engineering, Artif. Intell. 117 (2) (2000) 277–296. [36] E.H. Durfee, J.S. Rosenschein, Distributed problem solving and multi-agent systems: Comparisons and examples, Ann Arbor 1001 (48109) (1994) 29. [37] K. Slimani, D. Pallez, L. Médini, P. Ghodous, Multi-agent system architecture for concurrent design and manufacturing, in: 11th ISPE International Conference on Concurrent Engineering, 2004, pp. 209–215. [38] A.-J. Fougères, Modelling and simulation of complex systems: an approach based on multi-level agents, Int. J. Comput. Sci. Issues 8 (6) (2011) 8–17. [39] E. Ostrosi, A.-J. Fougères, M. Ferney, D. Klein, A fuzzy configuration multi-agent approach for product family modelling in conceptual design, J. Intell. Manuf. 23 (6) (2012) 2565–2586. [40] E. Ostrosi, A.-J. Fougères, M. Ferney, Fuzzy agents for product configuration in collaborative and distributed design process, Appl. Soft Comput. 12 (8) (2012) 2091–2105. [41] T. Mavridis, A.L. Symeonidis, Identifying valid search engine ranking factors in a Web 2.0 and Web 3.0 context for building efficient SEO mechanisms, Eng. Appl. Artif. Intell. 41 (2015) 75–91. [42] L.A. Zadeh, Outline of a new approach to the analysis of complex systems and decision processes, IEEE Trans. Syst. Man Cybern. 1 (1973) 28–44. [43] G.P. Farias, G.P. Dimuro, A.C. da R. Costa, BDI Agents with Fuzzy Perception for Simulating Decision Making in Environments with Imperfect Information, in: Proceedings of The Multi-Agent Logics, Languages, and Organisations Federated Workshops (MALLOW 2010), Lyon, France, August 30 - September 2, 2010. [44] A.-J. Fougères, A modelling approach based on fuzzy agents, Int. J. Comput. Sci. Issues 9 (6) (2013) 19–28. [45] S. Sagot, A.-J. Fougères, E. Ostrosi, A multi-agent approach for building a fuzzy decision support system to assist the SEO process, in: IEEE International Conference on Systems, Man, and Cybernetics (SMC'16), October 9–12, Budapest, Hungary, 2016, pp. 4001–4006. [46] C.-H. Chen, L.P. Khoo, W. Yan, Evaluation of multicultural factors from elicited customer requirements for new product development, Res. Eng. Des. 14 (3) (Nov. 2003) 119–130.

[12] J. Zhang, A. Dimitroff, The impact of metadata implementation on webpage visibility in search engine results (Part II), Inf. Process. Manage. 41 (3) (2005) 691–715. [13] M.P. Evans, Analysing Google rankings through search engine optimization data, Internet Res. 17 (1) (2007) 21–37. [14] S. Zhang, N. Cabage, Search engine optimization: comparison of link building and social sharing, J. Comput. Inf. Syst. 57 (2) (Apr. 2017) 148–159. [15] S. Brin, L. Page, The anatomy of a large-scale hypertextual Web search engine, Comput. Netw. ISDN Syst. 30 (1) (1998) 107–117. [16] R.A. Malaga, Worst practices in search engine optimization, Commun. ACM 51 (12) (2008) 147–150. [17] K. Dye, Website abuse for search engine optimisation, Netw. Secur. 2008 (3) (2008) 4–6. [18] D. Lidow, Startup Leadership: How Savvy Entrepreneurs Turn Their Ideas into Successful Enterprises, John Wiley & Sons, 2014. [19] S. Sagot, A.-J. Fougères, E. Ostrosi, P. Lacom, Search engine optimization: From analysis based on an engineering meta-model towards integrative approaches, in: Proceedings of International Conference on Information Society (i-Society), 2014, pp. 274–281. [20] A.-J. Su, Y.C. Hu, A. Kuzmanovic, C.-K. Koh, How to improve your Google ranking: Myths and reality, in: 2010 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT), vol. 1, 2010, pp. 50–57. [21] X. Zhu, Z. Tan, SEO keyword analysis and its application in website editing system, in: 2012 8th International Conference on Wireless Communications, Networking and Mobile Computing (WiCOM), 2012, pp. 1–4. [22] W. Di, L. Tian, B. Yan, W. Liyuan, L. Yanhui, Study on SEO monitoring system based on keywords & links, in: 2010 3rd IEEE International Conference on Computer Science and Information Technology (ICCSIT), vol. 5, 2010, pp. 450–453. [23] D.J. Power, Decision support systems: a historical overview, in: Handbook on Decision Support Systems, vol. 1, Springer, 2008, pp. 121–140. [24] J.T. Yao, An introduction to web-based support systems, J. Intell. Syst. 17 (1–3) (2008) 267–282. [25] N.P. Suh, Axiomatic design: Advances and applications (the oxford series on advanced manufacturing), 2001. [26] N.P. Suh, Complexity: Theory and Applications, Oxford University Press on Demand, 2005. [27] K. Curran, Tips for achieving high positioning in the results pages of the major search engines, Inf. Technol. J. 3 (2) (2004) 202–205. [28] M. Bernal, J.R. Haymaker, C. Eastman, On the role of computational support for designers in action, Des. Stud. 41 (2015) 163–182. [29] M.C. Schut, On model design for simulation of collective intelligence, Inf. Sci. 180 (1) (2010) 132–155. [30] E. Ostrosi, D. Choulier, M. Kurth, A meta-model for engineering analysis in product design, in: J. Stjepandić, G. Rock, C. Bil (Eds.), Concurrent Engineering Approaches for Sustainable Product Development in a Multi-Disciplinary Environment,

276