Limitations of System dynamics?

[zenabraham@aol.com]

Greetings, My personal point of view, due to spark controversy, is that nothing is too complex to represent with System Dynamics. The only limitation is the knowledge and imagination of the researcher / programmer, and the software that person is using. Moreover, and perhaps for reasons known best to you, I didnt receive a great, detailed explanation of what youre doing. Its a bit laden with jargon that seems to be unique to your business. If you can explain what youre doing in baby steps of detail, that would help. Zennie Abraham From: zenabraham@aol.com

[Latin Hypercube]

In my experience SD is very applicable to modelling business situations, that is it can be used to map out the critical aspects of a business and evaluate options - strategic or tactical. The key seems to be not to try to model the whole business as it operates but to pick out the key resource stocks and flows and build an explanatory model. Ideally, one will also have historical data from the business against which to test the fit of your model. I greatly recommend Kim Warrens book Competitive Strategy Dynamics for its focus on using SD in business/strategy with John Stermans Business Dynamics an excellent resource for detailed models of SD in use and some of the more technical aspects of the approach. Alan Coughtrey From: Latin Hypercube <latinhypercube@ntlworld.com>

["Jos De Neve/ESG"]

Some advice Be conscious of dynamics; (time, variations over time, delays) no dynamics ... No SD (a non cartesian way of thinking is mandatory; that is the most difficult) Be conscious therefore: no loops, no system, no dynamics Many business problems are quite sizeable; is yours sufficiently fence-able/containable? Do you have sufficiently feasible data? Do you sufficiently know the problem and the problem construct? Make an exercise using a rather simple problem Hope to have helped a bit _______________________ jos de neve From: "Jos De Neve/ESG" <executive.support.group@pandora.be> executive support group snepkaai 6 B 9000 gent - belgium http://www.esg-int.com

["Woncheol Kwak"]

Dear SD-er worldwide, I am new to System dynamics and struggling to find some ways to applying it to a commercial project. The project is agent-based simulation building and I dont have much experience in it either. -.-; (My profession was Database and mobile programming) At first I tried to use UML as in most other SW projects but soon I realized that its not that fit for this area, and it threatened the effective communications between my customer and programmers. Then I came across System Dynamics (John Stermans book) and was immediately attracted in it. However, it was not that easy for me to model my customers extremely complex business logic with SD. Now what Im wondering is, whether this difficulty is due to my lack of experience in SD or another bad choice. My question: Does anyone have a comprehensive list of pros and CONS of using System Dynamics in business modeling? Can anyone tell me the limitations of System Dynamics and the cases in which Id better not use it? The entities in my customers business process seem to be too complex to me to model with System Dynamics and I dont have much time to thoroughly examine the potentials of SD...if its the right choice, everybody will be happy. But if its not, itll be too late after I realize that. Thanks in advance. Woncheol Kwak Director/Project Manager HUMANEX, Inc. Korea woncheol.kwak@humanex.net

["Jim Hines"]

Woncheol Kwak wonders about whether he should use system dynamics: >The entities in my customers business process seem to be too complex to >me to model with System Dynamics and I dont have much time to >thoroughly examine the potentials of SD. Woncheol, The good news is that the process is not too complex. The bad news is you dont have enough time to learn system dynamics and then take a system dynamics approach to the problem. Jim Hines From: "Jim Hines" <jhines@MIT.EDU> MIT

[Timothy Fong]

Alan, Im trying to work with a client to model the marketing of a new product. They dont have historical data or even normalizing industry-wide data, and so cant see how SD can help them understand and make decisions on how they can market and what the range of revenue could be based on various variables. Have you had experience with that in business simulations? I think the problem is relatively well-defined. But the challenge is they want data. Thanks. Tim --------------------------------------------------------- Timothy Fong timfong@ureach.com ---------------------------------------------------------

[Bill Harris]

Woncheol Kwak wrote: > Does anyone have a comprehensive list of pros and CONS of using System > Dynamics in business modeling? > Can anyone tell me the limitations of System Dynamics and the cases in > which Id better not use it? Better than pros and cons, I sometimes find Gomez and Probsts taxonomy of problems and approaches to solutions helpful: Simple problems: small number of variables, low coupling among them. I find approaches such as Kepner-Tregoes helpful for these. (Simple is not necessarily the same as easy.) Complicated problems: large number of variables, relatively strong cross-coupling. Stable structure with little dynamic behavior. Operations research often deals with this sort of issue. Complex problems: complicated problems in which the interaction among variables changes over time. This is the domain where system dynamics shines. Thats summarized from pp. 14-15 of Die Praxis des ganzheitlichen Problemlösens, 2. Auflage. I agree with Jim Hines and others, though. If youre just learning system dynamics from a book, its too risky for a time-critical customer project. Youd certainly be better off doing it your own way this time or bringing in outside help. You might want to consider preparing for future opportunities by getting some training and experience. Regards, Bill From: Bill Harris <bill_harris@facilitatedsystems.com> -- Bill Harris 3217 102nd Place SE Facilitated Systems Everett, WA 98208 USA http://facilitatedsystems.com/ phone: +1 425 337-5541

[Alan Graham]

Trouble is, SD is an extremely malleable process -- at heart, its just the scientific method, with many potential tools and processes available. Effective SD engagements can look like everything from group facilitation to econometrics. In practice, the limitations that most often constrain applicability (along with schedule and budget) are much more with the system dynamicist than the disciplines collective tools and processes. As Jim Hines has remarked in this forum, beginners are pretty much constrained to applications for which there are already well-known precedents and models that can be imitated directly and fully. So the most practical answer to Woncheol Kwaks question is: Ask whoever is likely to be working the problem (potentially, just yourself). Ask them. If they dont know, youre back to the beginners track. cheers, alan Alan K. Graham, Ph.D. PA Consulting Group Alan.Graham@PAConsulting.com One Memorial Drive, Cambridge, Mass. 02142 USA

["Tom Forest"]

UML produces static, detailed, non-runnable flowchart-like models. Think Visio. System dynamics produces dynamics, general, runnable simulation models in time series. They complement each other, but in my experience do not overlap much. Think Powersim, Vensim, or iThink. Having spent most of the last dozen years developing custom software, I know what Woncheol Kwak is referring to. UML is Unified Modeling Language. Quoting from Grady Booch, "UML User Guide:" "[It] is a graphical language for visualizing, specifying, constructing, and documenting the artifacts of a software-intensive system. The UML gives you a standard means of writing the systems blueprints, covering conceptual items such as the business processes and system functions as well as concrete items such as classes written in a specific programming language, database schemas, and reusable software components." To my eyes, it looks like flow charting. Its a language for building static models, of listing the steps in a process (say, ordering or production), which one can then use as a blueprint from which to write software. I use it after having done the system dynamics model to identify the critical process(es) that need improved implementations. Do you have a reference mode of problematic behavior that you want to explore? Do you have question about priorities and what the critical processes are for improvement? Do you want to try different policy implementations and explore their effects on the system as a whole? Do you want to build a shared consensus between people about any or all of these thins? System Dynamics can help. Having decided on the critical process(es) for improvement and what the new policies and procedures in that process are, do you want to clarify what software objects youll need, what properties/methods/events they should have, and how they should be organized to maximize software performance, flexibility and maintainability? UML can help. Tom Forest tforest@prometheal.com

[Latin Hypercube]

Sorry for the long delay in replying, my travel schedule has kept me away from email lists. Ive made comments on my experience and your situation below. Its good to hear from someone wanting to use the best tools available in client engagements, I applaud you. That said, and as no doubt others will have pointed out, Id advise strongly against trying to learn and apply SD at the same time. Either get some help or spend some of your spare time(!) on this project learning and trying to use SD to generate insights for yourself, not your client. I think you can expect a great benefit from doing that. Regards UML etc. I think it can be useful to pick bits out of that method to approach your problem. I can see that you can generate generic use cases which map on to the flows from one stock to another, e.g. potential customer places and order :: potential customers -1, customers awaiting delivery +1. Otherwise I dont think UML fits your situation. Ive quite extensive experience with UML from having run several large software development projects but I dont attempt to use it in any way for business strategy/planning work which is what I think youre trying to tackle. First, some comments on my experience with using SD. I can tell you that I have used SD as part of several larger engagements with some success. These have included studies for european mobile telecoms operators on the introduction of GPRS (2.5G) and UMTS (3G) services to the market, the product/service development process for a very large mobile operator and (in a more limited way) the R&D process of a pharmaceuticals manufacturer. These are all mission critical issues with huge financial implications which will take time to play out but the clients and I have been very happy with what weve achieved. Of course, using SD techniques in a management situation is not necessarily straightforward. I can tell you the way I have approached things, at the risk of opening myself up for (deserved ) criticism: 1. At the strategy level, we use some of the tools with management to explore the business situation. This is but a method by which to explore a "perceived reality" and generate some agreement on key aspects of the situation. What is useful here are the concepts of stocks and flows and how they are affected by a variety of factors and the consequences of delays in feedback. This allows us to get at the important aspects of the situation and build understanding of the dynamics. One important achievement at this stage is to generate in the team, a sense of the issues for resource allocation as the business situation develops through time. In my view, organizations can too easily lock-in early resource allocation decisions through their organizational and career structures -- if you have a head of marketing, a head of fulfilment, a head of customer support, each will naturally try to do their job and further their career by maximising/minimizing the usual things of sales, expenses etc. and easily make decisions which dont recognize the changing state of the business as a whole. This brings us to another key issue -- the business situation can be expected to change through time and you already have a mental model of how that might happen -- you have to collect data and analyze information continuously now to tell you where you are in the process (and potentially to challenge your assumptions/business design). Data gathering needs to built into the business and will itself require some substantial resource allocation. (Note that Ive not usually tried to introduce SD as a method to management teams. Certainly, if one is not a master of this topic, and I am not, to do so can be very hard going. That said, my pharma client did get into SD in some detail which was helpful but my telecoms clients really didnt want to know; probably a consequence of managing in businesses that havedoubled in size each year of operation! This is where I expect criticism from the list -- one shouldnt really be using a "black box" to give answers to a management team that hasnt understood the scope and limitations of the mechanism contained in that box. In my defence, I see that managing a business all comes down to an exercise of managerial judgement and Im just doing the best I can to inform that judgement and that my recommendations are only a part of the whole decision process.) 2. At the planning level, we use tools and techniques that help us refine estimates of resource allocations, e.g. how much to spend in marketing, what size of call centre to build and when, etc. These naturally flow from (1.) above and really take us down to a level of detail where we can check our model of perceived reality for consistency and -- with the availability of real world data and plausible estimates -- check it for sanity. The various pieces of software that exist to speed up this process are quite helpful here, e.g. IThink. As the business rolls forward, it is important to build in data collection to allow you to see where you are on the map and whether to challenge the map. Eventually, this is all about doing the best to plan the necessary resource allocations to support the business goal, manage changes to those resource allocations and to occassionally see whether the map needs to be re-drawn. Having a model built up in software that allows you to run what-ifs is of great utility here as it will save you much time. Secondly, some comments on the situation you describe. It is unusual not to have data of any kind, since a "new" product is rarely unique. In the case of mobile services of course, we are having to look at all kinds of consumer data to guess/estimate plausible spending potentials for the new services. (Obviously, we check our assumptions against real data as we get it from the business in operation.) The discipline really must be to exhaust yourself looking for data which give you the ability to make planning investments. Since I dont know anything about the product/service under consideration here, I can only offer some very general pointers. Look at the reference literature for better help. Anyway: One model of the marketing chain is "AIDA" which describes the journey of an individual from potential to actual customer. The acronym stands for Awareness, Interest, Desire, Action. You might use this as a jumping off point for your thinking about the central chain of stocks and flows involved. For example, there will be a stock of consumers who are unaware of your product (the whole universe at this point!). They will presumably become aware of your product through your marketing efforts and word-of-mouth recommendations from other customers. Considering your marketing efforts, quickly brings you to considering the actions of your rivals in this product market. So as well, as your marketing effort, you possibly need to look at their marketing effort to work out your effective "share of voice". (Referring back to an earlier point, if you make a resource allocation to marketing spending in ignorance of your rivals and they decide to spend more money you may not generate the awareness per dollar you assume. You will have to adjust your marketing approach -- spend more or go about it in a different way. In any case, having a model built in software that can run what-ifs with will pay dividends.) And so on along the customer acquisition chain. For example, are you seeding influencers who will assist word-of-mouth? Do you have a partner programme which adds utility to your product? When a customer decides to buy, can they do it easily? Are you filling orders in the time customers expect to wait? This is all standard marketing stuff but hopefully you can see how you might look at it afresh with an eye to identifying important stocks, flow and influencing variables. Finally, good luck and enjoy! From: Latin Hypercube <latinhypercube@ntlworld.com>

["Ray on EV1"]

The system dynamics methods are suitable in general. But changing paths in
the middle of a project might be more catastrophic than your current
situation.

You should re-think where you are: Your experienced in UML, your client is
not communicating well with your programmers. Communication difficulties in
UML have two typical sources - the facilitator doesnt know UML that well,
the participants dont really know the field they are expected to be
experienced in. SD has the advantage of not requiring either but you must
understand the methods. Since you are familiar with UML, you might consider
that your client isnt actually a domain expert in the material he is
supposed to provide. This is common for an executive on a tight budget not
wanting to allocate personnel to a project. You might also consider where
as you are familiar with UML and understand the process, you may not be well
versed at carrying (facilitating) others along.

So you are in the middle of a project and dont see the resources you need
to progress. Consider canceling the project and cutting short your clients
losses, or bring in a facilitator for one of the two methodologies.

Ray
From: "Ray on EV1" <rtjoseph@ev1.net>