Hello,
how do one formulate where the goal (compared to currrent state) is rising.
The goal in question is to improve a process constantly, so there is constant change towards the "ulitmate" vision!
After every time step the actual process is compared to the past time step:
1. has it decreased >> the goal for next period is the past one
2. has it increased >> the goal is x% on the present one
3. equal >> the goal is x% on the present/past one
For me it looks either like a DELAY- or MIN/MAX-function. Am I right? Which one would be appropriate?
How do one get this into a workable VENSIM formulation?
Thanks a lot and best regards
Ralf
Formulating a rising goal
Formulating a rising goal
Ralf Lippold
Perhaps to make the "story" behind a bit more clearer (actually this was the story getting to point what I want to achieve -in a general way- during discussions with a friend):
Imagine a field & track sportsman who is doing preparation for the Olympics.
During his training sessions he always has the final vision "getting the gold medal" in mind (even though I won't reach it, it will be the "ultimate" goal).
So his training goal is to become steadily better, meaning faster on the e.g. 100 Meters. His sub-goal is to be at least faster than in the past. If that doesn't work, getting worse time on some training runs the former better run times will be the goal for next training (if not put on paper than at least mentally in his head).
I guess most of use know the feeling reaching new hights in whatever field and the tension for getting better by it.
This was just to make the point of my first posting a bit clearer.
The generally positive re-inforcing process could switch to a vicious cycle when the gap between the present output and the "achievable" goal seems to be too large and therefor reduces the efforts of improvement undermining the initial positive tension impacts.
Best
Ralf
Imagine a field & track sportsman who is doing preparation for the Olympics.
During his training sessions he always has the final vision "getting the gold medal" in mind (even though I won't reach it, it will be the "ultimate" goal).
So his training goal is to become steadily better, meaning faster on the e.g. 100 Meters. His sub-goal is to be at least faster than in the past. If that doesn't work, getting worse time on some training runs the former better run times will be the goal for next training (if not put on paper than at least mentally in his head).
I guess most of use know the feeling reaching new hights in whatever field and the tension for getting better by it.
This was just to make the point of my first posting a bit clearer.
The generally positive re-inforcing process could switch to a vicious cycle when the gap between the present output and the "achievable" goal seems to be too large and therefor reduces the efforts of improvement undermining the initial positive tension impacts.
Best
Ralf
Ralf Lippold
formulating a rising goal
Hi Ralf
I reformulate your problem into what I think to have understood.
New text and unformal definition while still clear and precise.
< A trainer is considering different policies to manage a sportsman towards making him reach a goal
that to his opinion he considers reachable taking into account his physical possibilities.
The problem is how to split the longer goal into workable short term goals, so as to keep the motivation of the sportsman high while coping with the time left until the competition starts. >
This is a very short definition, that has nothing to do with any methodology like SD, and that is centred on the problem of the client (the trainer).
Before wanting to solve a problem it is better to concentrate on the client’s problem (which is the more difficult part of modelling), so as to be sure we solve the correct problem!
Is it what you want to solve or does it need some modifications or clarifications?
The problem is certainly dynamic, and SD might help choose a policy of setting goals that will maximize the chance of
reaching the goal in the time left.
The next step will be to build the model.
Regards.
JJ
I reformulate your problem into what I think to have understood.
New text and unformal definition while still clear and precise.
< A trainer is considering different policies to manage a sportsman towards making him reach a goal
that to his opinion he considers reachable taking into account his physical possibilities.
The problem is how to split the longer goal into workable short term goals, so as to keep the motivation of the sportsman high while coping with the time left until the competition starts. >
This is a very short definition, that has nothing to do with any methodology like SD, and that is centred on the problem of the client (the trainer).
Before wanting to solve a problem it is better to concentrate on the client’s problem (which is the more difficult part of modelling), so as to be sure we solve the correct problem!
Is it what you want to solve or does it need some modifications or clarifications?
The problem is certainly dynamic, and SD might help choose a policy of setting goals that will maximize the chance of
reaching the goal in the time left.
The next step will be to build the model.
Regards.
JJ
-
bob@vensim.com
- Senior Member
- Posts: 1107
- Joined: Wed Mar 12, 2003 2:46 pm
There is a fair literature on goals, goal gap formulations and floating goals - most tend to work somewhat the opposite of the way you propose which is a reflection of reality - perhaps why it sometimes seems so dismal.
If you want a goal that can only increase the you use something like
floating goal = SMOOTH(indicated floating goal,time to adjust goal)
indicated floating goal = floating goal +
IF THEN ELSE(performance < floating goal,0,
(perfection - floating goal) * performance impact)
performance impact = performance impact f(ZIDZ(performance-floating goal),(perfection - floating goal))
performance impact f((0,0),(.25,.1),(.5,.5),(.75,.9),(1,1))
Note that this is continuous in the change in performance relative to the floating goal.
Note also that TIME STEP is an artificial construct that should not be used as if it had concrete meaning. You should always try to conceptualize formulations in continuous time.
If you want a goal that can only increase the you use something like
floating goal = SMOOTH(indicated floating goal,time to adjust goal)
indicated floating goal = floating goal +
IF THEN ELSE(performance < floating goal,0,
(perfection - floating goal) * performance impact)
performance impact = performance impact f(ZIDZ(performance-floating goal),(perfection - floating goal))
performance impact f((0,0),(.25,.1),(.5,.5),(.75,.9),(1,1))
Note that this is continuous in the change in performance relative to the floating goal.
Note also that TIME STEP is an artificial construct that should not be used as if it had concrete meaning. You should always try to conceptualize formulations in continuous time.
formulating a rising goal
Hi Bob and Ralf.
It seems to me that setting a fixed goal to be reached or working with changing goals with no
clear knowledge of the final result, is something very different, from a practical point of view.
For instance, knowing what the final goal is, may be important under special considerations:
It is much more motivating to look for a final predefined goal, than to have a general objective of just becoming better, which everybody tries to achieve anyway.
It is too better to settle a set of more or less interdependent goals, than to let them float.
Setting a final goal, eases too the overall organisation of the process and obliges to think before one starts. With changing goals you can start without thinking and adjust accordingly as new experiences are made.
So working with the hope for the best possible goal or working for a final defined goal is very different.
I choose to develop a model of a predefined goal, considering that the chosen goal is one of the more important policy if not the only one.
Any idea about that reflection?
If not I will try to build a model step by step from that consideration.
If Ralf is not interested by a definite final goal and prefers a model whose objective is to reach the maximum reachable objective, I will change the purpose of the problem accordingly.
Regards.
JJ
It seems to me that setting a fixed goal to be reached or working with changing goals with no
clear knowledge of the final result, is something very different, from a practical point of view.
For instance, knowing what the final goal is, may be important under special considerations:
It is much more motivating to look for a final predefined goal, than to have a general objective of just becoming better, which everybody tries to achieve anyway.
It is too better to settle a set of more or less interdependent goals, than to let them float.
Setting a final goal, eases too the overall organisation of the process and obliges to think before one starts. With changing goals you can start without thinking and adjust accordingly as new experiences are made.
So working with the hope for the best possible goal or working for a final defined goal is very different.
I choose to develop a model of a predefined goal, considering that the chosen goal is one of the more important policy if not the only one.
Any idea about that reflection?
If not I will try to build a model step by step from that consideration.
If Ralf is not interested by a definite final goal and prefers a model whose objective is to reach the maximum reachable objective, I will change the purpose of the problem accordingly.
Regards.
JJ
Hello Jean-Jacques,
Now occurs the difficult question (actually that seems to be the most difficult part) in what units is the goal measured?
It could be profit (in $$ or €€, or else): then one could reduce head counts and "improve" the process output in the short run?
It could be a productivity number: producing more products in the same quality with the same input?
It could be a quality number: producing the same amount of products with a higher quality (question in case is will the customer notice that and therefor would it have an impact on future sales?)?
Perhaps my comments enlighten a few "darker" corners of my initial question (hopefully;-)).
Best regards
Ralf
;)
the "final" goal is to eliminate every process in the complete process chain that adds no value to the product/service. Of course this is more a "moving" target and should be like a vision to up to (like looking up to your father as a small boy, never really getting exactly to him!).Originally posted by LAUJJL
Hi Bob and Ralf.
.....
If not I will try to build a model step by step from that consideration.
If Ralf is not interested by a definite final goal and prefers a model whose objective is to reach the maximum reachable objective, I will change the purpose of the problem accordingly.
Regards.
JJ
Now occurs the difficult question (actually that seems to be the most difficult part) in what units is the goal measured?
It could be profit (in $$ or €€, or else): then one could reduce head counts and "improve" the process output in the short run?
It could be a productivity number: producing more products in the same quality with the same input?
It could be a quality number: producing the same amount of products with a higher quality (question in case is will the customer notice that and therefor would it have an impact on future sales?)?
Perhaps my comments enlighten a few "darker" corners of my initial question (hopefully;-)).
Best regards
Ralf
;)Ralf Lippold
rising goal
Hi Ralf
Your problem looks more like a moving target, than a fixed one. Look at Sterman page 533, floating goals. it gives some ideas although being very scholar.
About units, the difficulty to choose among them is that you are looking at increasing the added values of the processes, while decresasing the costs.
The difficulty to choose a unit comes often from the fact that the objective is multiple and that each objective has a different unit. One of the solution is to choose a payoff that is a mix of the different objectives, and to choose a common unit, deciding how much each unit is worth in the common unit. You can choose for instance dollar as a common unit and you will have to convert every different objective into dollars. It is the same thing that you have to do if you choose multiple objectives for the payoff when optimizing in Vensim, you have to choose a weight for every objective.
I wonder if you are not triying to misuse SD, thinking that it will give you detailed operationnal policies.
Geoff Coyle thinks that a good model is a simple one, and
the more is becomes complex and the more chance it has to become unuseful.
You should first try to formulate precisely your purpose in very agregated terms and work first with simple models, so as to gain global understanding before eventually deciding to add details. It is alway easy to add details later on and painful to cut away details that proved to be marginal.
Regards.
JJ
Your problem looks more like a moving target, than a fixed one. Look at Sterman page 533, floating goals. it gives some ideas although being very scholar.
About units, the difficulty to choose among them is that you are looking at increasing the added values of the processes, while decresasing the costs.
The difficulty to choose a unit comes often from the fact that the objective is multiple and that each objective has a different unit. One of the solution is to choose a payoff that is a mix of the different objectives, and to choose a common unit, deciding how much each unit is worth in the common unit. You can choose for instance dollar as a common unit and you will have to convert every different objective into dollars. It is the same thing that you have to do if you choose multiple objectives for the payoff when optimizing in Vensim, you have to choose a weight for every objective.
I wonder if you are not triying to misuse SD, thinking that it will give you detailed operationnal policies.
Geoff Coyle thinks that a good model is a simple one, and
the more is becomes complex and the more chance it has to become unuseful.
You should first try to formulate precisely your purpose in very agregated terms and work first with simple models, so as to gain global understanding before eventually deciding to add details. It is alway easy to add details later on and painful to cut away details that proved to be marginal.
Regards.
JJ
Stuck in my own mental model
Dear fellow dynamicists,
a while ago I have asked about how to build a "moving target" into a VENSIM model.
After I have reflected for quite some time trying to cope with the challenge there still is glimpse of puzzling as the model doesn't behave as reality shows.
The model will be attached to the post and it is about sustainable change of a "lean implementation". This will be affected by appropriate rewarding policies and the actual workload that is coming.
I am very grateful to any helping insight in this tough question (at least it looks like that to me!).
Thanks a lot and cheers
Ralf
a while ago I have asked about how to build a "moving target" into a VENSIM model.
After I have reflected for quite some time trying to cope with the challenge there still is glimpse of puzzling as the model doesn't behave as reality shows.
The model will be attached to the post and it is about sustainable change of a "lean implementation". This will be affected by appropriate rewarding policies and the actual workload that is coming.
I am very grateful to any helping insight in this tough question (at least it looks like that to me!).
Thanks a lot and cheers
Ralf
- Attachments
-
- MOD_080606_Lean_Culture.mdl
- (3.77 KiB) Downloaded 312 times
Ralf Lippold
Changing goal
Hi Ralf
I cannot read your model.
Maybe is it because it is written for the Macintosh.
Logically you have just to click on the browse button beneath your written text and choose the model to send, the one with a mdl extension. Did you follow that procedure?
When I build a model, I write first down in plain common language, exactly the statement of the problem.
If this is done correctly, the modelling is not so complicate.
Regards.
JJ
I cannot read your model.
Maybe is it because it is written for the Macintosh.
Logically you have just to click on the browse button beneath your written text and choose the model to send, the one with a mdl extension. Did you follow that procedure?
When I build a model, I write first down in plain common language, exactly the statement of the problem.
If this is done correctly, the modelling is not so complicate.
Regards.
JJ
Rising goal
Hi Ralf
I have well received your model by mail.
Il looks very strange.
First it has unit errors which probably is a result of the strange equations in it.
I have never seen levels that have as input the ratio of the input by the output. What does it mean? It should be
integ (input - output and not integ (input / output
Why is the variable VA/VNA ratio a difference and not a ratio? This is misleading.
What is exactly the level standards. Is it the applied
lean thinking standards?
What is exactly the walk the talk ?
I do not understand why the workload has a negative influence on the learning and a positive on the unlearning.
I do not understand the va / nva ratio. What is it?
What variables do you find not behaving properly?
Can you formulate your model in plain english, explaining clearly what each variable is representing, and why is one variable influencing another one?
Regards.
JJ
[Edited on 6-6-2008 by LAUJJL]
I have well received your model by mail.
Il looks very strange.
First it has unit errors which probably is a result of the strange equations in it.
I have never seen levels that have as input the ratio of the input by the output. What does it mean? It should be
integ (input - output and not integ (input / output
Why is the variable VA/VNA ratio a difference and not a ratio? This is misleading.
What is exactly the level standards. Is it the applied
lean thinking standards?
What is exactly the walk the talk ?
I do not understand why the workload has a negative influence on the learning and a positive on the unlearning.
I do not understand the va / nva ratio. What is it?
What variables do you find not behaving properly?
Can you formulate your model in plain english, explaining clearly what each variable is representing, and why is one variable influencing another one?
Regards.
JJ
[Edited on 6-6-2008 by LAUJJL]
Modified model, error - expecting an operator
Hello Jean-Jacques,
thanks for your comments:-) I am especially working on the plain English explanation what is going on in the model (or at least my mental model;-)).
In the meanwhile I have done some modifications and have tried to implement the DELAY FIXED function (there occurs an error message (see attachment).
Regards
Ralf (really getting the learning grip at the moment - feedback feels great:-))
thanks for your comments:-) I am especially working on the plain English explanation what is going on in the model (or at least my mental model;-)).
In the meanwhile I have done some modifications and have tried to implement the DELAY FIXED function (there occurs an error message (see attachment).
Regards
Ralf (really getting the learning grip at the moment - feedback feels great:-))
- Attachments
-
- MEDIA_080606_DELAY_FIXED_Error.png (8.03 KiB) Viewed 11266 times
Ralf Lippold
Modified model version
Just to keep up with the model versions - here is the next one (with slight modifications).
Explanations of the variables (stocks, flows) are mainly in the comment fields of the variables.
Complete story is going to follow.
Regards
Ralf
Explanations of the variables (stocks, flows) are mainly in the comment fields of the variables.
Complete story is going to follow.
Regards
Ralf
- Attachments
-
- MOD_080606_Lean_Culture_1.mdl
- (4.28 KiB) Downloaded 325 times
Ralf Lippold
Model picture (as modified)
Here is the model as viewed in VENSIM grasped as a picture.
Especially the non-change of WALK-THE-TALK, especially when the REWARDING POLICIES are changed appropriate to foster the Lean Culture?
Regards
Ralf
PS.: VA = value adding work; NVA = non value adding work; all other variables will be talked about in the forthcoming story;)
[Edited on 6-6-2008 by Ralf]
Especially the non-change of WALK-THE-TALK, especially when the REWARDING POLICIES are changed appropriate to foster the Lean Culture?
Regards
Ralf
PS.: VA = value adding work; NVA = non value adding work; all other variables will be talked about in the forthcoming story;)
[Edited on 6-6-2008 by Ralf]
- Attachments
-
- MEDIA_MOD_080606_Lean_Culture_1.png (17.19 KiB) Viewed 11268 times
Ralf Lippold
Explaination of the situation - in plain English
Hello everybody,
here -finally- comes the story around which the model is woven. It is rather long - so please be patient;-)
Initial state:
- lean rewarding policies not in action (initial value = 1; better would be 0)
- there is no walk-the-talk (initial value =0)
The right hand side of the model is not in affect at this initial state.
The company gets customer demand that establishes the number of incoming new tasks on the work force. The new tasks flow into the stock of workload (that has an initial value = 10). As the workforce is working on fulfilling the customer demands work is getting done and this is represented by the outflow of finished tasks.
Over time the board realizes that the business is getting tougher and as they have learned from management courses and some consultants they see the "Lean Management" wave to be a useful way to get more competitive in the market.
So they install a special taskforce that is meant to forster the speed of processes and better use of resources, that ends in faster finished tasks (as before in the steady state) [even though there is not yet an appropriate variable on that goal in the model, this just have come up while thinking about how to write it in plain text].
What is all too often seen in reality is that this changes are going to start off really good only to completely die after a while getting the organization back to the "old" stage putting it in an even tougher shape against the competition.
As you can image the workforce is strongly watching what their bosses are doing due to the new way of working. As they -the bosses- walk the talk (meaning live to the new lean culture themselves) the impact on learning and adopting to the new ways amongst the workforce will increase. This will be the in flow into the stock of standards (lean standards being actively lived by everybody). Surely there will be also an unlearning, as people leave the organization taking knowledge of lean methods with them. Addtionally there will be an decrease of standards if there is not appropriate walk the talk or "slack" (meaning extra time being cut of the normal workday by improving the current processes) to learn and expirement with the new methods related to lean thinking.
This learning and unlearning will result in a VA/NVA (VA= value added work, NVA = non value added work) ratio that will have an effect (probably with a time delay) on the finishing of tasks. As this ratio is rising and the processes become more efficient the speed of finishing tasks will increase (there will be additional effects such as new piles of work to fill the "freed up" time by the improvements which will eventually cause further pressure on the workforce (to be added to the pressure coming out of following the new lean rules). [This has not yet been added to this model but analysed in another seperate model].
If the stock of workload is not increasing there will be also more time for the managers to do the "walk the talk" and this will increase the learning amongst the workforce and will lead to further leveraging of the lean standards in the overall organization. If the workload is rising, then there is less time for "walk the talk" and the managers can't fulfill the appropriate amount of "walk the talk" resulting in lowering the standards and through lowering the VA/NVA ratio as well will have negative impact on the workload (as finished tasks are slowing down).
Now comes the lean rewarding policies into play, as this impacts the willingness of managers to do the "walk the talk". These policies could be like rewarding the knowledge sharing between people in different department in order to get to better processes, crossing departmental boundaries or silo barriers, team work initiatives are rewarded (where not a single person or boss is getting the merrits).
The main question is: how can the system by influenced by policies given (lean rewarding policies, walk the talk, customer demand adjusting due to flawless workload increase) in order to get to a sustainable lean culture in the organization?
Of course this only shows a small part of the whole system and still this is difficult enough to model - especially if one is a newcomer to the method;-( The model probably will be expanded over time and build more to suit reality.
Concerning the reference modes of the variables over time I am not quite sure whether I could bring it over to you in the words and will try to add some graphs of the variables in case in a separate posting.
Thanks so much for taking the time to read my thoughts that have evolved over quite some time together with deep discussions with Markus Stamm and due to constant discussion and dialogue in the "Lean Thinking" special interest group at XING. It seems to be a huge topic and yet not very many -if not any- people are realizing what is really going on in this dynamics. Mostly it is said after a failed lean program, "The people didn't do it right. Let's change them and the right outcome will come!".
Yet this will lead to perpetual failures in changing to what Toyota is already doing for several decades (respect for people and constant rethinking of processes to become better!).
Cheers and awaiting tough and challenging comments, questions and -of course- hints on how to get closer to a possible solution or understanding of the situation at first.
Ralf
PS.: The rising goal is that VA/NVA ratio should constantly raise over time (not in big steps but in small increments - kaizen like;) in order to keep the lean culture sustain and so save resources which has on the other hand big impacti on other goals that we are after on a larger scale.
[Edited on 6-7-2008 by Ralf]
here -finally- comes the story around which the model is woven. It is rather long - so please be patient;-)
Initial state:
- lean rewarding policies not in action (initial value = 1; better would be 0)
- there is no walk-the-talk (initial value =0)
The right hand side of the model is not in affect at this initial state.
The company gets customer demand that establishes the number of incoming new tasks on the work force. The new tasks flow into the stock of workload (that has an initial value = 10). As the workforce is working on fulfilling the customer demands work is getting done and this is represented by the outflow of finished tasks.
Over time the board realizes that the business is getting tougher and as they have learned from management courses and some consultants they see the "Lean Management" wave to be a useful way to get more competitive in the market.
So they install a special taskforce that is meant to forster the speed of processes and better use of resources, that ends in faster finished tasks (as before in the steady state) [even though there is not yet an appropriate variable on that goal in the model, this just have come up while thinking about how to write it in plain text].
What is all too often seen in reality is that this changes are going to start off really good only to completely die after a while getting the organization back to the "old" stage putting it in an even tougher shape against the competition.
As you can image the workforce is strongly watching what their bosses are doing due to the new way of working. As they -the bosses- walk the talk (meaning live to the new lean culture themselves) the impact on learning and adopting to the new ways amongst the workforce will increase. This will be the in flow into the stock of standards (lean standards being actively lived by everybody). Surely there will be also an unlearning, as people leave the organization taking knowledge of lean methods with them. Addtionally there will be an decrease of standards if there is not appropriate walk the talk or "slack" (meaning extra time being cut of the normal workday by improving the current processes) to learn and expirement with the new methods related to lean thinking.
This learning and unlearning will result in a VA/NVA (VA= value added work, NVA = non value added work) ratio that will have an effect (probably with a time delay) on the finishing of tasks. As this ratio is rising and the processes become more efficient the speed of finishing tasks will increase (there will be additional effects such as new piles of work to fill the "freed up" time by the improvements which will eventually cause further pressure on the workforce (to be added to the pressure coming out of following the new lean rules). [This has not yet been added to this model but analysed in another seperate model].
If the stock of workload is not increasing there will be also more time for the managers to do the "walk the talk" and this will increase the learning amongst the workforce and will lead to further leveraging of the lean standards in the overall organization. If the workload is rising, then there is less time for "walk the talk" and the managers can't fulfill the appropriate amount of "walk the talk" resulting in lowering the standards and through lowering the VA/NVA ratio as well will have negative impact on the workload (as finished tasks are slowing down).
Now comes the lean rewarding policies into play, as this impacts the willingness of managers to do the "walk the talk". These policies could be like rewarding the knowledge sharing between people in different department in order to get to better processes, crossing departmental boundaries or silo barriers, team work initiatives are rewarded (where not a single person or boss is getting the merrits).
The main question is: how can the system by influenced by policies given (lean rewarding policies, walk the talk, customer demand adjusting due to flawless workload increase) in order to get to a sustainable lean culture in the organization?
Of course this only shows a small part of the whole system and still this is difficult enough to model - especially if one is a newcomer to the method;-( The model probably will be expanded over time and build more to suit reality.
Concerning the reference modes of the variables over time I am not quite sure whether I could bring it over to you in the words and will try to add some graphs of the variables in case in a separate posting.
Thanks so much for taking the time to read my thoughts that have evolved over quite some time together with deep discussions with Markus Stamm and due to constant discussion and dialogue in the "Lean Thinking" special interest group at XING. It seems to be a huge topic and yet not very many -if not any- people are realizing what is really going on in this dynamics. Mostly it is said after a failed lean program, "The people didn't do it right. Let's change them and the right outcome will come!".
Yet this will lead to perpetual failures in changing to what Toyota is already doing for several decades (respect for people and constant rethinking of processes to become better!).
Cheers and awaiting tough and challenging comments, questions and -of course- hints on how to get closer to a possible solution or understanding of the situation at first.
Ralf
PS.: The rising goal is that VA/NVA ratio should constantly raise over time (not in big steps but in small increments - kaizen like;) in order to keep the lean culture sustain and so save resources which has on the other hand big impacti on other goals that we are after on a larger scale.
[Edited on 6-7-2008 by Ralf]
Ralf Lippold
-
bob@vensim.com
- Senior Member
- Posts: 1107
- Joined: Wed Mar 12, 2003 2:46 pm
Hi Ralf,
I think step one is to study deeply the work that has been done in this area. Nelson Repenning has published a lot in this area and got the Forrester award some years back. I believe the model used in this work is also available. One thing that you will notice is that the model is quite a bit more complicated than your model. While small models are wonderful, they aren't always enough. Conceptual connections that make perfectly good sense in a causal loop diagram do not translate directly to a full simulation model. Details need to be fleshed out and physical rather than conceptual relationships specified.
I think step one is to study deeply the work that has been done in this area. Nelson Repenning has published a lot in this area and got the Forrester award some years back. I believe the model used in this work is also available. One thing that you will notice is that the model is quite a bit more complicated than your model. While small models are wonderful, they aren't always enough. Conceptual connections that make perfectly good sense in a causal loop diagram do not translate directly to a full simulation model. Details need to be fleshed out and physical rather than conceptual relationships specified.
Larger models - difficult for an outsider to understand
Hi Bob,
thanks for you comment on my recent postings.
You are right, there has been intensive work done by Nelson Repenning, John Sterman (some of his students).
The difficulty I sense while in constant discussion with members of the "Lean Thinking" group (where managers are also be part) is that they are looking for the quick solution and the larger models just scare them off and they don't even start talking about it.
So the intention is to learn from scratch in a very familiar situation (actually I am in this field for almost a decade now) and it hasn't changed much for years. Managers are driven by stakeholders (they take mistakenly only for shareholders;-() and there has to be quick results and positive numbers.
To understand the drivers in the model and the drivers that scare managers away from looking and trying to understand their own behavior in the system is what drives me to start small (surely this will grow with time and understanding and learning, I am quite positive on that).
Most people I am dealing with are engineers (I am just a plain economics guy;-)) and they love to have a running simulation model that they can "play" with and that gives out numbers.
I will have a look at the mentioned papers and perhaps there are also the models available as well (this would be great).
Thanks so much for commenting.
Best
Ralf
thanks for you comment on my recent postings.
You are right, there has been intensive work done by Nelson Repenning, John Sterman (some of his students).
The difficulty I sense while in constant discussion with members of the "Lean Thinking" group (where managers are also be part) is that they are looking for the quick solution and the larger models just scare them off and they don't even start talking about it.
So the intention is to learn from scratch in a very familiar situation (actually I am in this field for almost a decade now) and it hasn't changed much for years. Managers are driven by stakeholders (they take mistakenly only for shareholders;-() and there has to be quick results and positive numbers.
To understand the drivers in the model and the drivers that scare managers away from looking and trying to understand their own behavior in the system is what drives me to start small (surely this will grow with time and understanding and learning, I am quite positive on that).
Most people I am dealing with are engineers (I am just a plain economics guy;-)) and they love to have a running simulation model that they can "play" with and that gives out numbers.
I will have a look at the mentioned papers and perhaps there are also the models available as well (this would be great).
Thanks so much for commenting.
Best
Ralf
Ralf Lippold
Lean thinking development
Hi Ralf
To summarize your problem;
You want to build a set of policies that increases or maintain
the use of lean thinking techniques.
I think that your definition is too long and should only mention what is necessary, especially if you want to start with a simple model first.
It is much easier to throw away a simple model than a complicated one that needed a lot of work.
I will give an example of definition even if it does not correspond to your problem.
A company gets a steady flow of work to do. Depending on the productivity of the workers, the time to do the job varies as is varying the backlog of works to do and the stock of works that are currently done.
The project of the managers is to increase the productivity
by using lean thinking techniques. The current use of lean thinking techniques is influenced by lean regarding policies and too by the past variation of the productivity. The increase of productivity justifies the increase of the use of lean thinking techniques that increases the productivity (a positive loop).
This is a very simple and easy to model starting definition.
If you want you can change it but at the condition that it
does not increase in complexity.
If we can start with a simple definition first even if it avoids most of the reality at the beginning, we will transform it in a workable model and decide after that what to add etc...
Regards.
JJ
To summarize your problem;
You want to build a set of policies that increases or maintain
the use of lean thinking techniques.
I think that your definition is too long and should only mention what is necessary, especially if you want to start with a simple model first.
It is much easier to throw away a simple model than a complicated one that needed a lot of work.
I will give an example of definition even if it does not correspond to your problem.
A company gets a steady flow of work to do. Depending on the productivity of the workers, the time to do the job varies as is varying the backlog of works to do and the stock of works that are currently done.
The project of the managers is to increase the productivity
by using lean thinking techniques. The current use of lean thinking techniques is influenced by lean regarding policies and too by the past variation of the productivity. The increase of productivity justifies the increase of the use of lean thinking techniques that increases the productivity (a positive loop).
This is a very simple and easy to model starting definition.
If you want you can change it but at the condition that it
does not increase in complexity.
If we can start with a simple definition first even if it avoids most of the reality at the beginning, we will transform it in a workable model and decide after that what to add etc...
Regards.
JJ
Focus on the essentials
Hi Jea-jacques,
thanks for boiling down my rather lengthy statement to the essential. Your definition of the problem pretty much covers what I have intended to say.
I had the feeling the model would be already small enough -as we initiallyY started with an about 20-level-model;-)
Cheers
Ralf
PS.: I have come across some smaller model patches concerning some questions more deeply.
[Edited on 6-14-2008 by Ralf]
thanks for boiling down my rather lengthy statement to the essential. Your definition of the problem pretty much covers what I have intended to say.
I had the feeling the model would be already small enough -as we initiallyY started with an about 20-level-model;-)
Cheers
Ralf
PS.: I have come across some smaller model patches concerning some questions more deeply.
[Edited on 6-14-2008 by Ralf]
Ralf Lippold
How to get reality piece by piece into a working model
Hello,
after a workshop at the beginning of the week on SD modeling, seeing that even PhD student are struggling with similarly heavy (and on the other side easy seeming) questions in their models I would like to come back on the problem.
The connections between processes, stakeholders and there feedback reactions in the organizational system is easy as long as you don't ask deeper questions. As soon as you start asking questions, such as, "Why is the workforce not commit themselves to the improvement programs? (Mental Model of the questioner: That's the best way to work without hazle, putting in new ideas, being motivitated)?".
Soon you realize that this underlying reasons are not easy to see and finally -after really intense discussion, personal experience, etc.- you see that the structure giving from higher hierarchy and pretty fixed installed in the organization makes the workforce's behavior quite understandable.
Here comes into play what hinders almost any system dynamics project/thinking (coming from the bottom of the organization):
Do I get a promotion, if I get involved into the project?
We can't change these set rewarding policies as we are not HR!
This makes it even more difficult to understand the different drivers and in which way (equations) they are connected in the model.
The Causal Loop Diagram looks pretty straight forward (at least for my eyes;-)) and the most challenging part of the transfer to a "working" model is to find the functions that emulate reality closely enough to see current reality (in the reference model run!).
Any comments and questions are very welcomed:-))
Cheers,
Ralf
PS.: How to split the given model into submodels which can be analysed seperately and then could be put together to run as a larger one?
after a workshop at the beginning of the week on SD modeling, seeing that even PhD student are struggling with similarly heavy (and on the other side easy seeming) questions in their models I would like to come back on the problem.
The connections between processes, stakeholders and there feedback reactions in the organizational system is easy as long as you don't ask deeper questions. As soon as you start asking questions, such as, "Why is the workforce not commit themselves to the improvement programs? (Mental Model of the questioner: That's the best way to work without hazle, putting in new ideas, being motivitated)?".
Soon you realize that this underlying reasons are not easy to see and finally -after really intense discussion, personal experience, etc.- you see that the structure giving from higher hierarchy and pretty fixed installed in the organization makes the workforce's behavior quite understandable.
Here comes into play what hinders almost any system dynamics project/thinking (coming from the bottom of the organization):
Do I get a promotion, if I get involved into the project?
We can't change these set rewarding policies as we are not HR!
This makes it even more difficult to understand the different drivers and in which way (equations) they are connected in the model.
The Causal Loop Diagram looks pretty straight forward (at least for my eyes;-)) and the most challenging part of the transfer to a "working" model is to find the functions that emulate reality closely enough to see current reality (in the reference model run!).
Any comments and questions are very welcomed:-))
Cheers,
Ralf
PS.: How to split the given model into submodels which can be analysed seperately and then could be put together to run as a larger one?
Ralf Lippold
Hi Ralf
It is difficult to split reality in piece especially when the parts are related to each other, through feed back loops.
To make a model understandable, I start with a simple model, eventually with one level,
and try to understand it fully by using it with all its different possibilities.
It is then easier to get to a slightly more complex model and to use it again fully and so on.
I have just finished a model that has unfortunately 130 loops and I have not enough studied the earlier easier versions and I am too lazy to start studying it fully with the lack of experience of the earlier versions. I have a model that I can use but I do not really understand it fully and intuitively and I am therefore not completely confident with the results it delivers. To bad. The next model I build, I will try to follow strictly the method.
Regards.
JJ
[Edited on 14-6-2008 by LAUJJL]
It is difficult to split reality in piece especially when the parts are related to each other, through feed back loops.
To make a model understandable, I start with a simple model, eventually with one level,
and try to understand it fully by using it with all its different possibilities.
It is then easier to get to a slightly more complex model and to use it again fully and so on.
I have just finished a model that has unfortunately 130 loops and I have not enough studied the earlier easier versions and I am too lazy to start studying it fully with the lack of experience of the earlier versions. I have a model that I can use but I do not really understand it fully and intuitively and I am therefore not completely confident with the results it delivers. To bad. The next model I build, I will try to follow strictly the method.
Regards.
JJ
[Edited on 14-6-2008 by LAUJJL]
-
bob@vensim.com
- Senior Member
- Posts: 1107
- Joined: Wed Mar 12, 2003 2:46 pm
Hi Ralf,
You might start with a project model - eg Richardson & Pugh prog.mdl in the others directory). Add to that an input to quality based on training in quality management principles so that you add to fraction satisfactory a new effect
fraction satisfactory = normal fraction satisfactory*
effect of experience on fraction satisfactory*
effect of schedule pressure on fraction satisfactory *
effect of training in quality management
where
effect of training in quality management = etqf(retained training hours / reference retained training hours) ~ dmnl
eftqf((0,0),(1,.9),(10,1)) ~dmnl
retained training hours = INTEG(training hours - forgetting training,initial training hours) ~Hours/Person
forgetting training = retained training hours / time to forget training ~ Hours/Person/Month
then you can experiment with investing time in training versus just having people trying to do work to see how that changes things.
You might start with a project model - eg Richardson & Pugh prog.mdl in the others directory). Add to that an input to quality based on training in quality management principles so that you add to fraction satisfactory a new effect
fraction satisfactory = normal fraction satisfactory*
effect of experience on fraction satisfactory*
effect of schedule pressure on fraction satisfactory *
effect of training in quality management
where
effect of training in quality management = etqf(retained training hours / reference retained training hours) ~ dmnl
eftqf((0,0),(1,.9),(10,1)) ~dmnl
retained training hours = INTEG(training hours - forgetting training,initial training hours) ~Hours/Person
forgetting training = retained training hours / time to forget training ~ Hours/Person/Month
then you can experiment with investing time in training versus just having people trying to do work to see how that changes things.