Hello,
as still a newbee to SD and the modelling process I am right now crossing the edge of transforming a CLD to a working simulation model. The model will have 5 stocks (should this be downsized still?) and the CLD connections seem pretty straigth forward -as is the result from deep discussions with a friend.
Now occurs the problem: how to minimize the spaghetti connections between stock / flows in a way that the model would be still understandable for others (with less SD background, let's say managers)!
Is there a methodology that could be used to build the visual model (before the functions are put in) in a reasonable manner?
Advice is greatly appreciated:)
I guess other newcomers to the field would be also interested in this. I am quite positive there is no single lane solution, but ideas would be a good trigger to deepen understanding of the practical use of SD in the business world.
Best regards
Ralf
[Edited on 12-12-2007 by Ralf]
Develop model from CLD
Develop model from CLD
Ralf Lippold
-
bob@vensim.com
- Senior Member
- Posts: 1107
- Joined: Wed Mar 12, 2003 2:46 pm
The best advice is actually to rearrange or rebuild the model after you thoroughly understand what is important and what you need to emphasize in your presentation to others. If you actually have the time even better advice is to build a completely new and simpler model that gets to the heart of your important conclusions.
During the model construction phase it can be fatal to include everything that is perceived as important and try to make it work. It is much better to start with one or two loops, and build a model of those. After that is working and understand additional structure can be added. Putting in everything at once often leads to something that takes so long to get to run sensibly there is no time for understanding. Remember, 5 is the number of levels in Jay's World Dynamics model.
During the model construction phase it can be fatal to include everything that is perceived as important and try to make it work. It is much better to start with one or two loops, and build a model of those. After that is working and understand additional structure can be added. Putting in everything at once often leads to something that takes so long to get to run sensibly there is no time for understanding. Remember, 5 is the number of levels in Jay's World Dynamics model.
CLD
Hi Ralf
I do not think that ther is a real methodology for what you explain.
A CLD can have two functions.
One is to prepare for further quantitative modelling and another one improving the understanding of a dynamic problem.
It is not evident to make both at the same time.
There are no formal rules to represent a problem with a diagram as long as it works.
To learn to build diagrams with different level of details, one must practice and think a lot. It comes with experience.
A good book for diagram simplification is Geoff Coyle (system dynamics modelling, a practical approach). The cover on the book shows the notion of the cone of diagrams with different level of details adapted to different kilnd of people.
Regards.
JJ
I do not think that ther is a real methodology for what you explain.
A CLD can have two functions.
One is to prepare for further quantitative modelling and another one improving the understanding of a dynamic problem.
It is not evident to make both at the same time.
There are no formal rules to represent a problem with a diagram as long as it works.
To learn to build diagrams with different level of details, one must practice and think a lot. It comes with experience.
A good book for diagram simplification is Geoff Coyle (system dynamics modelling, a practical approach). The cover on the book shows the notion of the cone of diagrams with different level of details adapted to different kilnd of people.
Regards.
JJ
Move from CLD to modelling
Hello Jean-Jacques,
thanks a lot for your comment that highlights the problem very well.
We use the CLD for our own understanding of the processes and this was even quite a journey so downsize the rather large -monitor filliing- CLD in the beginning. As the connections are quite clear but still the dynamics over time can't be fully grasped by CLD.
So we decided to expand our thoughts into a modell in order to make it more workable (as a simulation).
In the meanwhile -from designing the first CLD to the present day- we both (he is doing a PhD in Auckland, New Zealand) put our outcomes in the real world asking colleagues, professors, friends in order to verify our findings. This is giving us both great insights in the perception of others and the mental models that we all share -but mostly don't know (not before we ask the right questions).
The book by Geoff Coyle you mention seems to be a good choice and I will check that out for further enquiries on how to model for different audiences.
What exactly is Coyle covering in his book? What's the difference to John Sterman's Business Dynamics (he builds also CLD and models from there on)?
Happy to gain further drive in my learning through discussions like that:-)
Regards
Ralf
thanks a lot for your comment that highlights the problem very well.
We use the CLD for our own understanding of the processes and this was even quite a journey so downsize the rather large -monitor filliing- CLD in the beginning. As the connections are quite clear but still the dynamics over time can't be fully grasped by CLD.
So we decided to expand our thoughts into a modell in order to make it more workable (as a simulation).
In the meanwhile -from designing the first CLD to the present day- we both (he is doing a PhD in Auckland, New Zealand) put our outcomes in the real world asking colleagues, professors, friends in order to verify our findings. This is giving us both great insights in the perception of others and the mental models that we all share -but mostly don't know (not before we ask the right questions).
The book by Geoff Coyle you mention seems to be a good choice and I will check that out for further enquiries on how to model for different audiences.
What exactly is Coyle covering in his book? What's the difference to John Sterman's Business Dynamics (he builds also CLD and models from there on)?
Happy to gain further drive in my learning through discussions like that:-)
Regards
Ralf
Ralf Lippold
CLD
Hi Ralf
I have been working with Coyle’s book about 4 months (20 hours a week), and I have only studied a fifth of the book.
When I started studying it, I quickly realized that it was written by somebody that had a very good understanding
of practical SD.
It looked as if I was starting studying something completely new.
In fact his book, makes SD take a step down from its mythical and magical ivory tower to become a nearly
common place field for the great ease and profit of the reader.
It delivers a method that goes from the nebulous expression of a problem to a more structures written definition, to
a diagram that represents exactly the written definition, to a simulation quantitative model with policy analysis, to
a robust policy from the correct utilization of optimization.
It does not only give one method, but shows different options. For instance there are three ways to construct diagrams.
It has given me the missing link between the expression of a problem (the reality) and the formal mathematical
formulation.
With the method you can see the continuous transformation of a badly formalized problem to a completely formalized one, not hiding the difficulties of the process.
If you want to study the book seriously, you will have to study from the beginning to the end, nearly 20 different cases, avoiding to look at the explanations the book gives, before you try to work out the solutions by yourself (this is critical condition of learning and this explains the time I take to study it).
Thus the book is built on understanding by doing.
The book has a back bone, that leads you chapter by chapter in a continuous learning path (not the case in Sterman’s book, that is constructed in a very heteroclite way).
I have worked nearly 6 years ago, with Sterman’s book, making all the challenges. I found it extraordinarily interesting, but I have realized through the years, that having a good understanding of academic and theoretical SD, does not help at all, building useful models. And Sterman’s book is academical. There is no clear method to build models, and of course no training to use it.
Coyle’s book addresses equally well, open and closed systems.
It shows too very well, what you can already get from the study of a qualitative diagram (much more than what I thought at first) and that this study will help you when you take the step to quantitative modelling.
It shows too that it is not always useful to tak that step.
Unfortunately, I believe that Coyle is too abrupt in his thinking.
The way he demystifies most of the SD myths, was certainly very badly perceived by the SD scholar community.
Scholars do not like that what they teach look too simple.
There is surprisingly no reference of his book, in the huge references of Sterman’s book.
Nor is there any reference in Warren’s book either.
I will add that Jay Forrester who does not like compromise and is too a bit abrupt, prefaced Coyle’s book, and not
Sterman’s book.
At the time when the book was produced, Coyle was still considered a reference.
I think he should not have written an article with David Exelby in the SD review about the difference of validity between a consultant’s work and an academic work.
He is completely right in his article, but there are some things that must be said with extreme precaution
to avoid hurting susceptibilities.
I am too sometimes abrupt.
I verified it in my recent posting to the Sd mailing list, where one very known academic sent me
an insulting personal e-mail. He considered being insulted by my public e-mail relating personal experiences
in contradiction with actual research and practice.
It is my fault, I should have been more careful about what I wrote.
Coyle’s article is extremely interesting and is joined.
With his article he must have transformed all the scholar and research SD community into personal enemies.
I will explain in a later post, how I concretely build a simplified diagram, from a more complex one, giving an example.
Regards.
[Edited on 14-12-2007 by LAUJJL]
I have been working with Coyle’s book about 4 months (20 hours a week), and I have only studied a fifth of the book.
When I started studying it, I quickly realized that it was written by somebody that had a very good understanding
of practical SD.
It looked as if I was starting studying something completely new.
In fact his book, makes SD take a step down from its mythical and magical ivory tower to become a nearly
common place field for the great ease and profit of the reader.
It delivers a method that goes from the nebulous expression of a problem to a more structures written definition, to
a diagram that represents exactly the written definition, to a simulation quantitative model with policy analysis, to
a robust policy from the correct utilization of optimization.
It does not only give one method, but shows different options. For instance there are three ways to construct diagrams.
It has given me the missing link between the expression of a problem (the reality) and the formal mathematical
formulation.
With the method you can see the continuous transformation of a badly formalized problem to a completely formalized one, not hiding the difficulties of the process.
If you want to study the book seriously, you will have to study from the beginning to the end, nearly 20 different cases, avoiding to look at the explanations the book gives, before you try to work out the solutions by yourself (this is critical condition of learning and this explains the time I take to study it).
Thus the book is built on understanding by doing.
The book has a back bone, that leads you chapter by chapter in a continuous learning path (not the case in Sterman’s book, that is constructed in a very heteroclite way).
I have worked nearly 6 years ago, with Sterman’s book, making all the challenges. I found it extraordinarily interesting, but I have realized through the years, that having a good understanding of academic and theoretical SD, does not help at all, building useful models. And Sterman’s book is academical. There is no clear method to build models, and of course no training to use it.
Coyle’s book addresses equally well, open and closed systems.
It shows too very well, what you can already get from the study of a qualitative diagram (much more than what I thought at first) and that this study will help you when you take the step to quantitative modelling.
It shows too that it is not always useful to tak that step.
Unfortunately, I believe that Coyle is too abrupt in his thinking.
The way he demystifies most of the SD myths, was certainly very badly perceived by the SD scholar community.
Scholars do not like that what they teach look too simple.
There is surprisingly no reference of his book, in the huge references of Sterman’s book.
Nor is there any reference in Warren’s book either.
I will add that Jay Forrester who does not like compromise and is too a bit abrupt, prefaced Coyle’s book, and not
Sterman’s book.
At the time when the book was produced, Coyle was still considered a reference.
I think he should not have written an article with David Exelby in the SD review about the difference of validity between a consultant’s work and an academic work.
He is completely right in his article, but there are some things that must be said with extreme precaution
to avoid hurting susceptibilities.
I am too sometimes abrupt.
I verified it in my recent posting to the Sd mailing list, where one very known academic sent me
an insulting personal e-mail. He considered being insulted by my public e-mail relating personal experiences
in contradiction with actual research and practice.
It is my fault, I should have been more careful about what I wrote.
Coyle’s article is extremely interesting and is joined.
With his article he must have transformed all the scholar and research SD community into personal enemies.
I will explain in a later post, how I concretely build a simplified diagram, from a more complex one, giving an example.
Regards.
[Edited on 14-12-2007 by LAUJJL]
CLD
Hi
When I have built a diagram, example given in French from Coyle’s book and I want to build a simpler diagram I must first study each of the loops of the diagram, to decide if I want to represent it in the simpler diagram.
Coyle recommends to build up to 4 different diagrams at different level of details.
I first work on a copy of the diagram, because I will have to erase successively different arrows.
In my diagram where the stocks are represented in Bold and the physical input and output flows in full lines and the information arrows in dotted lines, I choose successively all the stocks.
I start with a first stock, preferably one that moves at the beginning of the simulation.
I click on the stock and at the loop button in the analysis tools, and copy the result in a word processor, for me Word.
I erase the arrows that represent the inflow and outflow of the first stock, to avoid duplication of loops.
I go to the second stock, click on it and on the loop button, copy the loops to the same file and so on until I have copied all the loops of the model.
After that I go to the text document, and start to study each loop in detail.
I go back to the original diagram not modified, and click on each variable on the loop to
visualize the loop. I try to understand how it is working and if is coherent with the problem definition. It is good way to verify if the diagram is correct.
It is rare not to find an error when doing this.
I note under the loop, the remarks about the loops, its polarity, its strength, its effect, when it starts, if it starts after the beginning of the simulation, the delay. I note too its eventual importance and if it its action is straightforward or complex.
The process is rather laborious, but after a while, you begin to feel that the diagram becomes more and more understandable and looks less complex.
In the process I note the loops whose action is to zero the action of another loop, generally after a delay.
In the example joined, there are 43 loops.
I have noticed that it is often possible to eliminate studying loops going through an outflow, because they are generally zeroing a precedent loop.
To do that I erase all the outflows from all the stocks.
If I do again the preceding study there are no more than 16 loops to study.
When I have finished I decide what loops I want to appear at the chosen level, here number 1, the simpler level.
At that level the importance is to stress the action of the chosen loops.
You do not need either to represent a stock with its inflow and outflow, one variable is enough for the three.
I think that you must work with many different models to become experienced.
I think that my actual method will vary with the time and I gain in experience.
The loops that one decides to represent must be simple enough, easy to understand and be of some importance.
In the diagram the policy variable is in bold and italic.
Joined the initial diagram, the livel 1 diagram and the word file with the 16 loops.
you will note too, that the dimension of all the variables are defined at this qualitative step, as it helps verify the understanding and the quality of the diagram.
Anybody having other ideas are welcome.
Regards.
JJ
[Edited on 14-12-2007 by LAUJJL]
When I have built a diagram, example given in French from Coyle’s book and I want to build a simpler diagram I must first study each of the loops of the diagram, to decide if I want to represent it in the simpler diagram.
Coyle recommends to build up to 4 different diagrams at different level of details.
I first work on a copy of the diagram, because I will have to erase successively different arrows.
In my diagram where the stocks are represented in Bold and the physical input and output flows in full lines and the information arrows in dotted lines, I choose successively all the stocks.
I start with a first stock, preferably one that moves at the beginning of the simulation.
I click on the stock and at the loop button in the analysis tools, and copy the result in a word processor, for me Word.
I erase the arrows that represent the inflow and outflow of the first stock, to avoid duplication of loops.
I go to the second stock, click on it and on the loop button, copy the loops to the same file and so on until I have copied all the loops of the model.
After that I go to the text document, and start to study each loop in detail.
I go back to the original diagram not modified, and click on each variable on the loop to
visualize the loop. I try to understand how it is working and if is coherent with the problem definition. It is good way to verify if the diagram is correct.
It is rare not to find an error when doing this.
I note under the loop, the remarks about the loops, its polarity, its strength, its effect, when it starts, if it starts after the beginning of the simulation, the delay. I note too its eventual importance and if it its action is straightforward or complex.
The process is rather laborious, but after a while, you begin to feel that the diagram becomes more and more understandable and looks less complex.
In the process I note the loops whose action is to zero the action of another loop, generally after a delay.
In the example joined, there are 43 loops.
I have noticed that it is often possible to eliminate studying loops going through an outflow, because they are generally zeroing a precedent loop.
To do that I erase all the outflows from all the stocks.
If I do again the preceding study there are no more than 16 loops to study.
When I have finished I decide what loops I want to appear at the chosen level, here number 1, the simpler level.
At that level the importance is to stress the action of the chosen loops.
You do not need either to represent a stock with its inflow and outflow, one variable is enough for the three.
I think that you must work with many different models to become experienced.
I think that my actual method will vary with the time and I gain in experience.
The loops that one decides to represent must be simple enough, easy to understand and be of some importance.
In the diagram the policy variable is in bold and italic.
Joined the initial diagram, the livel 1 diagram and the word file with the 16 loops.
you will note too, that the dimension of all the variables are defined at this qualitative step, as it helps verify the understanding and the quality of the diagram.
Anybody having other ideas are welcome.
Regards.
JJ
[Edited on 14-12-2007 by LAUJJL]
- Attachments
-
- cld.zip
- (9.4 KiB) Downloaded 295 times
Validation of problem for further modelling
Hello Jean-Jacques,
thanks a lot for your thoughts and the model you have attached (even though it is in French;-) -a good way to learn some French).
In quite the same way our modelling around the problem has taken place so far: starting small, adding lots of points and arrows, downsizing again, switching to the main points of the problem.....
Right now there is no client (management) really involved which makes the task even more difficult. The faced problem is the lack of sustainability of certain change programs over time (even though they are quite positive in the beginning, due to siloing the overall effects are not seen by the actors). This reminds me of the mentioning of George Richardson' comment on chapter 9 in Ed Roberts "Managerial Applications of System Dynamics" which gives good hints on what to achieve with modelling and what not (this makes some things clear).
Which article by Geoff Coyle have have you had in mind in your earlier posting? I checked the old SDR and one article dated 1998 (The validation of commercial system dynamics models) seemed like a fit, but I am not sure that's the one you are talking about.
The new ideas have to settle and we will sure come back with deeper understanding of what we are trying to model (or should I say explain and fix through the right policy;-)).
Best
Ralf
thanks a lot for your thoughts and the model you have attached (even though it is in French;-) -a good way to learn some French).
In quite the same way our modelling around the problem has taken place so far: starting small, adding lots of points and arrows, downsizing again, switching to the main points of the problem.....
Right now there is no client (management) really involved which makes the task even more difficult. The faced problem is the lack of sustainability of certain change programs over time (even though they are quite positive in the beginning, due to siloing the overall effects are not seen by the actors). This reminds me of the mentioning of George Richardson' comment on chapter 9 in Ed Roberts "Managerial Applications of System Dynamics" which gives good hints on what to achieve with modelling and what not (this makes some things clear).
Which article by Geoff Coyle have have you had in mind in your earlier posting? I checked the old SDR and one article dated 1998 (The validation of commercial system dynamics models) seemed like a fit, but I am not sure that's the one you are talking about.
The new ideas have to settle and we will sure come back with deeper understanding of what we are trying to model (or should I say explain and fix through the right policy;-)).
Best
Ralf
Ralf Lippold
Hello Bob,
thanks for your short but nevertheless deep thought comment (we are already thinking about reducing the 5 stocks into a smaller and aggregated number)
The phenomenon of building up a model to largeness that can't be handled is occurring everywhere (like the "Golden Pony") and one has to find the braking of the vicious cycle.
Would it be another way to use the different views in Vensim to focus on specific dynamics of the complete model, or is this not meant to be?
To be honest we all are pretty much in SD because the problems we face and try to understand and solve are complex. So it's no wonder the models get a bit messy;-)
Best regards
Ralf
thanks for your short but nevertheless deep thought comment (we are already thinking about reducing the 5 stocks into a smaller and aggregated number)
The phenomenon of building up a model to largeness that can't be handled is occurring everywhere (like the "Golden Pony") and one has to find the braking of the vicious cycle.
Would it be another way to use the different views in Vensim to focus on specific dynamics of the complete model, or is this not meant to be?
To be honest we all are pretty much in SD because the problems we face and try to understand and solve are complex. So it's no wonder the models get a bit messy;-)
Best regards
Ralf
Ralf Lippold
CLD
Hi Ralf
The article from Coyle that you mention is the right one.
My problems are messy too and I started to solve them
some years ago with not enough experience having already solved simpler problems.
To learn SD one has to learn first to manage simple problems, and then progressively less simpler one.
If you start with complex problems too early you will get drowned.
One solution is to take your complex problem and to simplify the definition.
But the process of simplification is part of the difficulty to manage complex problems.
How to simplify the problem and still respect it?
What path to take from the simpler model to the definitive one?
And suppose that you succeed to do that and start with a simpler two loops model.
If you are not prepared to manage models of a certain level of difficulty you will get drowned
when you have exceeded the level of problem and model complexity you can manage, when increasing the size of the model.
I have tried that method but it works well and is often mandatory if you have already sufficient experience managing problems of a certain level of complexity.
I think that one must accept to take the time to first work with easier problems, be sure that they are correctly built and analyzed, that the policies chosen are robust, and then repeat the process with less simple problems. The time does not respect what has been built without him.
Coyle in his book writes:
The key to successful modelling is to keep one’s understanding of the model and what it says about the problem ahead of its size.
Is it the case with your model?
If you have a complete written definition of your problem, I can have a look at it, and tell you what I think about its level of difficulty.
May be Bob or an administrator will give an advice more valuable than mine too.
One important thing too is to be relatively sure that the problem has sufficient dynamical problems to justify the use of SD, and evaluate the expected benefit of solving the problem.
This question is important.
If both questions are not clearly stated, there will not be enough motivation to solve the problem. A model needs a lot of efforts, and it is preferable to be sure of their justification.
Regards.
JJ
[Edited on 15-12-2007 by LAUJJL]
The article from Coyle that you mention is the right one.
My problems are messy too and I started to solve them
some years ago with not enough experience having already solved simpler problems.
To learn SD one has to learn first to manage simple problems, and then progressively less simpler one.
If you start with complex problems too early you will get drowned.
One solution is to take your complex problem and to simplify the definition.
But the process of simplification is part of the difficulty to manage complex problems.
How to simplify the problem and still respect it?
What path to take from the simpler model to the definitive one?
And suppose that you succeed to do that and start with a simpler two loops model.
If you are not prepared to manage models of a certain level of difficulty you will get drowned
when you have exceeded the level of problem and model complexity you can manage, when increasing the size of the model.
I have tried that method but it works well and is often mandatory if you have already sufficient experience managing problems of a certain level of complexity.
I think that one must accept to take the time to first work with easier problems, be sure that they are correctly built and analyzed, that the policies chosen are robust, and then repeat the process with less simple problems. The time does not respect what has been built without him.
Coyle in his book writes:
The key to successful modelling is to keep one’s understanding of the model and what it says about the problem ahead of its size.
Is it the case with your model?
If you have a complete written definition of your problem, I can have a look at it, and tell you what I think about its level of difficulty.
May be Bob or an administrator will give an advice more valuable than mine too.
One important thing too is to be relatively sure that the problem has sufficient dynamical problems to justify the use of SD, and evaluate the expected benefit of solving the problem.
This question is important.
If both questions are not clearly stated, there will not be enough motivation to solve the problem. A model needs a lot of efforts, and it is preferable to be sure of their justification.
Regards.
JJ
[Edited on 15-12-2007 by LAUJJL]
-
bob@vensim.com
- Senior Member
- Posts: 1107
- Joined: Wed Mar 12, 2003 2:46 pm
Hi Ralf,
Different views are a great way to break up large models but they aren't always the best way to emphasize feedback loops - important loops can involve variables in several views.
I recognize that we face complex problems. My advice was not to build simple minded models, but to build up to more complete models. If you have identified a totally obvious loop, just building a model of that one loop can be very informative. It might be the case the model does not do what you expect. When it does do what you expect you can look at the behavior of all the variables and understand something about their ranges and perhaps limits or countervailing forces that must come into play. If these are already in your diagram then proceed as planned, otherwise you might need to revise thinking.
Also, the full simulation model is not the only thing to show to people. Simplified loop diagrams as a way of explaining both the model and reality can be very useful.
Different views are a great way to break up large models but they aren't always the best way to emphasize feedback loops - important loops can involve variables in several views.
I recognize that we face complex problems. My advice was not to build simple minded models, but to build up to more complete models. If you have identified a totally obvious loop, just building a model of that one loop can be very informative. It might be the case the model does not do what you expect. When it does do what you expect you can look at the behavior of all the variables and understand something about their ranges and perhaps limits or countervailing forces that must come into play. If these are already in your diagram then proceed as planned, otherwise you might need to revise thinking.
Also, the full simulation model is not the only thing to show to people. Simplified loop diagrams as a way of explaining both the model and reality can be very useful.
Hi Bob,
it's been a while since posting any questions - work is still under way and the time that is passing offers new views, ideas and directions.
Just the other day I had a game of chess with friends and it reminded me strongly of the model I want to build. Never look at one aspect of the game (system) without getting an idea on other parts of it due to your specific action. It was a real thrill to get connected so deeply with current situation.
In a way we got the initial idea into one single view model (even though it is not yet producing "correct" output) to still have the overall view. Aggregation was the concept behind it as some variables can be put together into one broader aggregated variable (either stock or flow, depending on the situation).
As we have come to a turning point having the model discussed from our different mental model positions (that are actually quite different in some parts) we are searching for the "key" to open new doors of the model and thinking.
I sometimes step into friends' imaginations when I connect a present situation (such as, why are some quite successful companies reduce their workforce despite the future necessity of people's knowledge, networks and ideas that will end in further prosperity -if used and seen wisely (in a long-run perspective!). You are right that the "small" models are the ones to start with and to get a sense what is understood by non-system dynamicists around the business world (there are actually quite a few) so you have a chance to rethink the models you have built on the ground of your own mental models.
Thanks a lot for all your comments so far.
Cheers
Ralf
it's been a while since posting any questions - work is still under way and the time that is passing offers new views, ideas and directions.
Just the other day I had a game of chess with friends and it reminded me strongly of the model I want to build. Never look at one aspect of the game (system) without getting an idea on other parts of it due to your specific action. It was a real thrill to get connected so deeply with current situation.
In a way we got the initial idea into one single view model (even though it is not yet producing "correct" output) to still have the overall view. Aggregation was the concept behind it as some variables can be put together into one broader aggregated variable (either stock or flow, depending on the situation).
As we have come to a turning point having the model discussed from our different mental model positions (that are actually quite different in some parts) we are searching for the "key" to open new doors of the model and thinking.
I sometimes step into friends' imaginations when I connect a present situation (such as, why are some quite successful companies reduce their workforce despite the future necessity of people's knowledge, networks and ideas that will end in further prosperity -if used and seen wisely (in a long-run perspective!). You are right that the "small" models are the ones to start with and to get a sense what is understood by non-system dynamicists around the business world (there are actually quite a few) so you have a chance to rethink the models you have built on the ground of your own mental models.
Thanks a lot for all your comments so far.
Cheers
Ralf
Ralf Lippold