7.2.2    The enterprise model development phase

Section describes some lessons learned pertaining to the enterprise model development phase.

Objectives of enterprise modeling

The empirical studies have provided some lessons related to both objectives espoused and objectives in-use. Assertions about the objectives of shared understanding and holistic thinking are also made, as these were the two most outstanding in the empirical studies.
Objectives espoused versus objectives in-use
The observed differences in intended use and actual use of enterprise modeling indicate that modelers ought to keep this distinction mind:
 
Assertion 3 

When the objectives of an enterprise modeling effort are left unattended for a prolonged period of time, there is a risk for differences in objectives espoused and objectives in-use. The consequence may be that the objectives espoused are not met.

By leaving objectives unattended is meant that the objectives are not recalled and discussed explicitly in the project group. The phrase prolonged period of time is used to indicate that objectives do not have to be discussed constantly, only often enough to ensure conformance between objectives and the activities carried out.

Looking for empirical evidence, both the Gazz and TEK-S projects offer observations that support assertion 3. In Gazz, the initial goal of enterprise modeling was to make people think holistically relying on qualitative relationships, while the final goal became quantitatively accurate models. In TEK-S (recalling table 6.2), several of the intended applications of enterprise modeling were not met (and they were not abandoned explicitly).

From a theoretical point of view, a plausible reason for the deviations concerning objectives may be that as the project group gain more understanding of the problem situation, they see that the original objective was not appropriate or the one most influential. This makes sense when the work process is dominated by design risk.

Another reason for deviations is resource consumption: If the amount of resources allocated to the project is not sufficient, something has to be sacrificed. In TEK-S, the focus on enterprise modeling was reduced as compared to the initial intentions, as the strategy document itself required more resources than originally foreseen.

Shared understanding
Shared understanding was the single most frequently stated objective of enterprise modeling in the empirical studies.
 
Assertion 4 

Enterprise modeling as a means to develop shared understanding in an organization may require extensive communication between the modelers and the enterprise audience. Hence, both the development phase and the dissemination phase must receive sufficient attention. 

No attempt is made to suggest what sufficient attention might be, but some measure has to be used to decide if actors agree upon the modeled reality. Assertion 4 may seem obvious, but observations from the empirical studies indicate that the dissemination phase was considered subordinate to the development phase. In VPT, the dissemination phase mainly consisted of distribution of the enterprise models. Incorporation of feedback from the enterprise audience was not done to any significant degree. In TEK-S, all the enterprise models were removed from the final strategy document and did not play any significant role in the dissemination phase.

From a constructivist point of view, shared understanding implies closure. Attaining closure requires a dissemination phase that allows the project group to both externalize their views on the enterprise as well as internalization of the audience's feedback on the models. The rest of the enterprise audience must also have the opportunity to internalize the project group's models and externalize their views.

Holistic thinking
Holistic thinking was an objective in all the four empirical studies.
 
Assertion 5 

When the purpose of enterprise modeling is holistic thinking, simplicity and completeness are more preferable qualities of an enterprise model than accuracy and consistency.

By model simplicity is meant that an enterprise model includes few types and instances of language concepts, and that the model elements are expressed in a language that the enterprise audience have internalized. Model completeness refers to how comprehensive the model is (the number of aspects that are included in the model). Model accuracy refers to the precision of a model (e.g., is the model oversimplified?), and model consistency denotes the existence of incompatible statements in the model (e.g., matter being output from a process without matter being input).

Assertion 5 is first and foremost based on the observation of how enterprise models were used as structuring devices: In discussions and presentations, the models were traversed element by element, discussing each element by itself and in combination with other model elements.

Looking at the various models, they are mostly simple. The VPT model relies on six language elements, with only three of them being used extensively (process, product and flow). The Gazz models are simple in the sense that they include few types of elements, and the types are all within the professional language of the enterprise audience. The models in TEK-S are all simple (see section 6.6.4).

The importance of model completeness was observed in the TEK-S incidence reported on page 109, where a WG was fortified with an actor with different professional background. Significantly more attention was directed towards his field of expertise.

Looking at the models, they are not necessarily accurate or consistent. The VPT model was neither, as discussed in section 2.2.5. In TEK-S, most of the models involved gross simplifications, making them inaccurate.

Searching for theoretical support for assertion 5, Weick (1995:57) may be read as supporting such a claim in his discussion of plausibility as a more important trait than accuracy in sense-making:

"The strength of sensemaking as a perspective derives from the fact that it does not rely on accuracy and its model is not object perception. Instead, sensemaking is about plausibility, pragmatics, coherence, reasonableness, creation, invention, and instrumentality."
Hence, when holistic thinking is sought supported by enterprise modeling and enterprise modeling is dominated by sense-making, accuracy is subordinate to factors like plausibility.

Activities in the development phase

The activities observed to dominate the development phase are focussed on improving the understanding that the project group have of the domain that is modeled (corresponding to the "inner loop" of figure 7.1).
Creation and use of intermediate artifacts
When enterprise modeling is dominantly sense-making and modeling takes place in a face-to-face meeting, the following observation has been made:
 
Assertion 6 

An effective approach to enterprise model development is creation of a number of intermediate artifacts through an iterated process of construction, presentation, interpretation and feedback.

The empirical evidence for assertion 6 is taken from TEK-S and WS2 (page 112) where the creation of intermediate artifacts was particularly visible. Actors constructed a proposal for the issue under consideration, presented it to the rest of the project group along with an explanation (allowing the group to interpret the artifact) and received feedback on the artifact (initiating a discussion).

From a theoretical point of view, the four activities bear close resemblance to the activities discussed in Figure 3.1, representing two iterations of the process: Construction is related to an actor's local reality. Presentation implies externalizing the local reality, making it available to fellow actors for interpretation (internalization). The actors are then in position to provide feedback (externalization).

Use of creative techniques
The project groups were observed using creative techniques (both deliberately and more casually) in order to create some of the intermediate artifacts.
 
Assertion 7 

Existing, well-known creative and analytical techniques may be effective means to create intermediate artifacts in the enterprise model development phase.

Empirical evidence in favor of assertion 7 can be found in TEK-S: Brainstorming was observed at the project kick-off meeting (page 99) and most notably on WS2 (page 113), concept clustering was employed at WS2, a QFD variant was also observed (as a part of the M2 techniques). In addition, questioning was a widely employed activity (A9), and posing questions was also included as an activity in the M2 model. Posing questions may be seen as a diverging activity, opening up to creativity (A8).

From a theoretical perspective, the use of creative techniques makes sense in an enterprise modeling effort that is dominated by sense-making: Creative techniques are devices for sense-making. They are devices for provision of requisite variety that is called for by Weick (1995:56), being a prerequisite for complexification of one's understanding. Seeing enterprise modeling as complexification is consistent with seeing enterprise model development as perspective making.

Summarizing and enterprise modeling
The use of summaries was observed to be effective in the sense-making process.
 
Assertion 8

Summarizing can be used in enterprise modeling as a means to make sense of the domain that is modeled.

 
By summarizing is meant the construction of a brief account of a domain, leaving out details that are perceived to be of little value to the audience. Empirical evidence for assertion 8 is found in TEK-S (A13), e.g., as a frequently observed activity in WG Gas.

From a theoretical point of view, a summary can be considered as an abstraction that for a given purpose removes the dead ends, insignificant events, etc. from the domain, and thereby makes the account easier to understand (by introducing order and reducing complexity).

Forums for enterprise modeling
The activities that have been observed as being a part of enterprise modeling impose constraints on the forums that are most appropriate.
 
Assertion 9

Enterprise model development is a communication-intensive activity that requires a forum with multi-directional, real-time communication facilities.

 
By a communication-intensive activity is meant an activity that requires much communication in little time. Multi-directional communication means that actors communicate with each other, as opposed to one-way presentation. By real-time communication is meant that the listening actor hears what the presenting actor has to say without delay in time (for practical purposes).

In practice, assertion 9 states that enterprise modeling is most effectively performed in face-to-face meetings with a restricted number of actors. Empirical evidence in favor of the assertion can be found in all four studies, as they all relied on the meeting as the main way of working. In VPT, they created models sitting around a table, communicating over the model. In PA30, the use of meetings to interview plant workers supports assertion 9. In TEK-S, work was mostly done in meetings in the creative stages (I and II).

Still, most of the models in TEK-S were not developed in meetings, but rather adapted from previous projects. How can this be consistent with assertion 9? The models in question were mostly dominated by replication risk and not controversial. The meaning of the M2 model, on the other hand, was constructed in meetings.

Assertion 9 is also supported by Weick (1995:185) in his discussion of implications of taking sense-making seriously in organizations:

"Perhaps the most perverse-sounding implication of the preceding chapters is the suggestion that people need to meet more often. That implication arises from the reluctance with which people acknowledge that they face problems of ambiguity and equivocality, rather than problems of uncertainty."
What Weick refers to as problems of ambiguity and equivocality require posing new questions and new arguments, and need a certain "messiness" that can only be found in a meeting. Problems of uncertainty (according to the terminology of Weick discussed in section 3.2.5) are approached by information scanning and retrieval.

Artifacts in the development phase

The artifacts created in the development phase are mostly intermediate in the sense that they are not exposed to the enterprise audience.
The unfinished state of intermediate artifacts
Intermediate artifacts have been observed to play a role in the sense-making process.
 
Assertion 10

The intermediate artifacts developed as a part of enterprise modeling are often partial and incompatible, in the sense that they describe some aspects of the problem situation, but not all. The main role of an intermediate artifact is to exemplify and concretize the issue being discussed.

 
Hence, the intermediate artifacts were not just improvements of each other. They were alternatives that resembled each other, but with distinct properties. The artifact production was not a rational process of stepwise approaching the best or most correct model through refinements, but rather a creative process that produced a plethora of alternatives.

The empirical evidence for the characteristics and role played by intermediate artifacts is mainly taken from TEK-S, as described on page (A17).

From a theoretical point of view, intermediate artifacts play a role in perspective making, being means to complexify the perspective. Again, Weick (1995:56) can be read supporting such a claim when calling for requisite variety as the remedy to make sense of complex situations:

"Complicate yourself if you want to understand complicated environments."
The idea of construction of intermediate artifacts very well illustrates how accuracy has to yield to variety: The importance of accuracy is subordinate to the conceptual outline that the intermediate artifact conveys.
Development of a terminology
Enterprise modeling involves development of a common vocabulary as a part of attributing meaning to the model. The challenges of this task were observed in TEK-S.
 
Assertion 11

Construction and use of a shared and precise terminology based on definitions requires agreement on wordily formulation, remembering the formulation and finally consistent use of the terms. Such a process is time-consuming and is not decisive to the success of the enterprise model development process.

 
Assertion 11 concerns both the activities involved in construction of a terminology and the perceived importance related to successful enterprise modeling. The activities, or rather difficulties of conducting them, were observed in TEK-S (page 136). The relatively limited importance of successful definition of a terminology is indicated by the assessment that the project groups attained most of their objectives although they did not succeed in the construction and use of a shared and precise terminology. Neither VPT nor Gazz attempted to construct a terminology, but still succeeded with their projects.

From a theoretical point of view, the meaning of terms can be considered to depend upon the actual use of language in a community of knowing, where the meaning of a term evolves and varies according to context and actor. Boland and Tenkasi (1995) discuss this as a language game within the community of knowing and a required part of developing a perspective.

Medium for development of enterprise models
The roles played by intermediate artifacts have consequences for their preferred medium.
 
Assertion 12

The most appropriate medium for representation of intermediate artifacts is one that allows rapid construction, presentation and modification (due to the limited life-time of an intermediate artifact) and shared access (due to the requirement to forums).

 
Empirical evidence in favor of assertion 11 can be found in all the four projects. A whiteboard and non-permanent markers, or plastic foils and an overhead projector were observed in VPT, PA30, Gazz and TEK-S. In VPT, they explicitly claimed that introduction of more advanced technology, e.g., a computer with sophisticated software, in the creative development phase was assumed to be counterproductive. In TEK-S, the use of a flip-over with marker and Post-It notes on the wall were also observed (e.g., at WS2).

None of the projects used technology like computers in the creative modeling phase. Computers were used to create exposed artifacts or when working alone in between meetings.

Construction of a modeling language
Although the focus in chapter 7 is on activities and artifacts pertaining to sense-making, representation is also a required part of enterprise modeling. Representation requires a modeling language.
 
Assertion 13

The actors in an enterprise modeling project group are capable of developing a simple, informal modeling language as an integrated part of the modeling effort. A benefit of developing their own language is that they can tailor it to a specific domain and their professional language.

 
By a simple modeling language is meant a language that has few and easily comprehensible concepts (relative to a community). By informal is meant a language that lacks both precisely defined syntax and precisely defined semantics.

Empirical evidence for assertion 13 can be found in all four projects. The languages in VPT and PA30 each consists of only six concepts with no definitions of syntax or semantics. The language in PA30 was tailored to the particular application through the concepts problem and opportunity. In Gazz, the language elements were predefined by the project group, as discussed in section 2.4.5. In TEK-S, all the models M1 - M7 are created using informal languages developed as a part of the project (or another project), and tailored to the specific needs and domain. An example is the M7 model, consisting of the overall diagram (with two concepts activity and market) and the assessment matrices for each activity, describing strengths, weaknesses, opportunities, threats, etc.

From a theoretical point of view, creation of a language tailored to a specific use in a particular project can be seen as an opportunity to base enterprise modeling on the professional language of the modelers, and not on the professional language developed by another community of knowing with a totally different background. In this way, the language do not become more powerful than the community's understanding of their perspective.

Still, remember that assertion 13 is stated in the context of an enterprise modeling development process dominated by human sense-making. In other situations, e.g., if computers were intended to interpret the models, the assertion might be invalid.

Enterprise modeling project group

Two assertions about the actors doing enterprise modeling and their relationships to the domain that is modeled conclude the development phase.
Composition of project groups
The knowledge of the actors that participate in the project group clearly influences the modeling process.
 
Assertion 14

When the actors in the enterprise modeling project group have general knowledge about the domain that is modeled, the modeling process is more effective and coverage of aspects of the domain is ensured.

 
By general knowledge is meant that they are familiar with the domain, e.g., understand the professional language within the area. By an effective process is meant one that results in models that are agreed to represent the domain. The term coverage denotes to what degree the model is perceived to be complete.

Empirical evidence for the assertion can be found in all four projects. In VPT, one of the actors had general knowledge of the variety of petroleum products and contributed this to the model. In PA30, interviewing all the plant workers were perceived to result in more correct models. In Gazz, the elements of the modeling language were well known to several of the actors and represented no problem. In TEK-S, the work groups were composed to match the production chain (like in WG Gas, see page 108).

Concerning coverage, an illustrative example was provided in TEK-S (page 109). When one of the work groups were fortified with an actor with knowledge of information technology, more attention was allocated to this aspect of the domain.

From a theoretical point of view, seeing the enterprise modeling process as a communication-intensive construction process implies that the actors need knowledge in order to contribute in the construction. Without at least a general knowledge of the domain, the perspective making process will not result in an effective perspective (i.e., a perspective that can be used to do work within the domain, as discussed in section 3.2.4). The degree of coverage of the domain can be considered as depending on the number of distinct perspectives that are involved in the modeling process.

Knowledge versus competence
The difference between knowledge and competence influences the accomplishment of the development process.
 
Assertion 15

Enterprise model development requires competence in the basic activities of the development process. Without competence, the activities may consume excessive amounts of resources.

 
Recall the difference between knowledge and competence from page 146. Empirical evidence for assertion 15 can be found in both PA30 and TEK-S. In PA30, the project group stated explicitly that a pilot project would have prepared them for the problems they encountered. Running a pilot project is essentially to gain competence through training. In TEK-S, the project group were advised to use a few techniques by the two consultants. However, even if they understood the techniques, they had no training in using them efficiently. The project group experienced a lot of trouble in trying to use the techniques and finally rejected them. Another example from TEK-S is the creative session at WS2 when the competent facilitator organized the brainstorming and concept clustering (page 113). This was considered highly successful and effective.

From a theoretical point of view, the difference between knowledge and competence is that between knowing about practice and being a practitioner, as discussed on page 146. Claims in favor of the need for basic skills in team-based modeling are also stated by Richardson and Andersen (1995) in their discussions of group modeling scripts. The scripts are described as (ibid.:130)

"planned and rehearsed routines for accomplishing subgoals in the course of a group model-building workshop."
Scripts are seen as basic skills that a member of a modeling team must have.