Reflections on Systemigrams

Hello Systems Thinkers,

As promised last week, here is a post containing an blog comment and then email dialogue I had with STA friend and regular commenter Julian Johnson.  If you want to look him up, here is a link to Juian’s LinkedIn profile and here is a link to the website for his consultancy business, holistem.

The conversation is about what a “Systemigram” ought to be, and how that differs from an “Information Model”.  Julian has very kindly put together some slides showing how the canteen Systemigram we produced is actually an Information model, and another set of slides showing what a Systemigram might actually look like.  So, here are Julian’s slides showing how:

  1. The canteen set-up might be expressed as an Information Model – here
  2. The canteen situation might be expressed as a Systemigram – here

And here is the dialogue we’ve had.  It’s well worth reading. Thanks Julian.

Comment Julian left on STA blog week 27 on 3rd December 2016.  Here.

Tim, another great post from you, very interesting and thought provoking. A couple of (again I hope constructive) observations:

  1. On the nature of your systemigram: When I look at what you have above, this seems to be a less a systemigram (Boardman et al), and more a conventional information model (IM). So what is a conventional IM? Basically it identifies concepts in the domain of interest, and declarative relationships between them. It isn’t really about influence relationships, which is arguably the essence of a systemigram. In fact, if I take your diagram, and replace ellipses by rectangles, and arrows by (depending on notation) lines with a directed annotation label (UML class diag) or labelled ‘lollypop’ line (STEP modelling language), AND ellipse-in-ellipse by the appropriate supertype-subtype notation (open triangle notation in UML class diag), I’m about there. Its not that the diagram is wrong in any way, (in fact it appears very useful), it just (to me) seems to have moved from a systemigram underlying paradigm, to something else (IM paradigm).
  1. Pick your time comment: Cannot agree more. It often seems to higher up the organisation you go, the less attention span we find (rightly or wrongly), and the more things have to be reduced to primary colours. We’re talking seriously low end of ‘magic number 7 +/- 2’, if you know what I mean. However, in my experience, when getting something that can appear complicated / complex in a single diagram over, it is better not to through the whole diagram at the audience at step 1, and then explain all the elements, but more to build up the diagram on the blank paper / screen, so you end up with the whole picture, but you have brought an audience along, step by step. In fact, this is partly why some tools, like Insight Maker, have a storyboard capability…

Hope that helps…

Tim’s email response on 5th December 2016

Thanks Julian.

I do appreciate these comments!

I must admit that my head is now in my hands though (metaphorically).  I couldn’t quite get Rich Pictures, so I had a go a Systemigrams, and it turns out I’ve not “got” them either.

I wonder, reading your explanation, are Rich Pictures and a Systemigrams much the same thing, or rather serve a similar purpose form a similar perspective, but obviously produced with different media in a different environment?


Julian’s email response on 5th December 2016


“my head is now in my hands though”. Oh dear.

“I wonder, reading your explanation are Rich Pictures and Systemigrams much the same thing, or rather serve a similar purpose from a similar perspective, but are obviously produced with different media in a different environment?”

If I can interpret your question, there appear to be two parts to it:

  1.   That Rich Pictures and systemigrams are much the same thing, serving similar purposes and/or from a similar perspective;
  2.   They are produced with different media (notations) in different environments (tools).

First to respond to 2:

They could indeed be produced with the same tool, for instance, pen and paper / flipchart etc. (But so could any diagram regardless how ‘well formed’.) In terms of computer based tools, it would then start to depend how flexible the tool is. Basically both Rich Pictures and systemigrams could be said to be essentially node-and-link diagrams (essentially a visualisation of a ‘graph’, in maths of computer science speak). However, systemigrams appear to be stricter to the node-and-link paradigm, than Rich pictures, because they are essentially about one item (node, semantically a concept) being linked to one or more other items (nodes) via links (representing relationships). Systemigrams have another type of relationship, type-of, represented by the ellipsi inside another ellipse layout. Anyway, any computer based tool worth its weight that allows you to put diagrams together will have some awareness of the underlying ‘paradigm’. For instance, on a data flow tool, it would not allow you to connect a flow line directly to another flow line, only a flow line to a process bubble, or a process bubble to a flow line etc. We can see this when we use Systemitool, which only allows certain diagram constructs to be connected to certain others. I guess the nearest we can get to a rich picture tool is only something like powerpoint that generically has a large pallet of icons, symbols and various connector types, but is pretty free about where you put things.

As to 1:

Rich pictures and Systemigrams are similar to each other in that they are both pictorial views that a) help a group of people represent and communicate an understanding of a domain and b) are both used for early phases tackling complex / complicated or chaotic systems, often social technical, where ‘conventional’ systems development approaches tend to struggle. However, they differ (as representations) in that a) systemigrams are rather more constrained than rich pictures in the strict node-link metaphor above and b) rich pictures allow carte blanche in types of symbols, AND how they are placed. For instance, having a symbol representing a group of people labelled ‘project management’ and a second symbol representing a group of people labelled ‘engineering’ in some proximity on the diagram, and a drawing of a brick wall between them usefully communicates: these groups don’t talk. There is no explicit node-link construct, it is the simple placement that we see as a pattern.

My comment below: “…identifies concepts in the domain of interest, and declarative relationships between them. It isn’t really about influence relationships…” is the essence in a sense of difference between an IM view, and a systemigram view. This email is already too long, but if it helps [for your understanding], I’ll happy try find / create some examples to try to illustrate my expectation of the difference of typical IM and systemigram examples.


Julian’s follow up Email on 12th December 2016

I’ve backed up from the week 27 canteen ‘systemigram’, given my observation that it appears closer to an information model than an systemigram, by building up an information model applicable to a canteen/ meals on wheels domain.

Attached is the result, which is basically a step by step build-up of a fledgling information model. I say fledgling as:

  •     I’ve not been explicit about capturing either requirements of what we are trying to achieve with such a model, nor with the use cases;
  •     I’ve not elaborated the nature of relationship ordinality (one-to-many, many-to-many etc)
  •     I’ve not elaborated many attributes that one would need to make the model ‘useful’.

However, you will see by the time you get the slide 8 which has the ‘full’ model (subject to caveats about AND any entity should have a definition) that the result is very similar (ignoring boxes rather than rectangles for concepts/entities, and straight versus arcs for relationships, and subtype representation is different) that there is much similarity; indeed I then hand marked with red ticks my paper copy of your systemigram, and may be 80% of items in your diagram are ticked (=covered).

In a separate response I’ll look at how my expectation on a systemigram would actually focus on how well (or poorly) a given canteen works (or doesn’t work) and it part why (or works or doesn’t work).

Hope this is of value to you… let me know!

Julian’s final email on 13th December 2016


Not hampered by tackling a real example, I’ve just tried to speculate a potential canteen and some of its issues. Attached is a slideset, where two of the slides are a prose explanation of some aspects that are ‘challenging’, and then the build-up of systemigram fragments. As you will see in the last slide, I came across some interesting issues, which probably means I need to (completely) read Boardman’s book properly, rather than just wing it. Anyway, I learned something by doing this anyway…

Hope it perhaps moves you a step closer to practical systemigrams, or at least, the distinction of systemigram to (say) information model?

Let me know if you go something (anything) out of this material…


Julian Johnson, BSc, PhD, CSEP MINCOSE
Director, Chief Scientist, Holistem Ltd.
+44 1254 209421
+44 797 442 8697


3 thoughts on “Reflections on Systemigrams”

  1. Tim

    I can’s say anything other than this is a great post. Thanks for the promotion!

    Just come back from facilitating a third workshop with a multi-disciplinary healthcare (NHS and others) group, building on our systems thinking analyses applied earlier that captured large range of issues and proposals / opportunities, and now moving to prioritising proposals (some done). Need to validate versus some of our systemigrams – attempt to confirm no unintended consequences!

    Cheers, Julian.

  2. Hello Julian,

    I was originally a member of Tim’s group, but I now merely dip in from time to time, and try to keep up with the blog. I did have time over the holidays to work through the two presentations you kindly put together, and I shared my thoughts with Tim, who suggested I share them with you – so here they are:

    ~ What’s the difference? ~

    I found the “Motivation” slides of the two presentations a useful way to understand the difference between an Information Model and a Systemigram. In my simplified universe, I am now labelling the Information Model as “a tool to identify things we can measure and the relationships between them” and a Systemigram as “a tool to help understand the interconnected effects of changes to a system”.

    As you can see, I approach both of these from the point of view of the function of the tool, because what is the point of developing and understanding a tool which isn’t used?

    For example, when I invent my tool for arranging raisins in an orthogonal grid, it may not be useful to spend time arguing over the accuracy of the placement. One obvious counter-argument would be Einstein’s theory of general relativity which became useful 50 years after its development, but I’m hoping you won’t point that out.

    ~ Information model arrows ~

    As the arrows on the diagram are, by definition, directional, it follows that the direction has meaning – otherwise we could simply connect with lines.

    In your 6th slide, “Some relationships 1”, the statements are in different voices:
    • Active – The kitchen prepares the food item
    • Passive – Food item is purchased by customer

    My question is this: If both statements had been expressed in the active voice, (Kitchen prepares Food, Customer purchases Food)
    a) would both arrows point toward the food item?, and if so
    b) would that make any difference?

    … or do we draw the arrows to make sense and then formulate the statements to support the diagram?

    ~ Systemigram issues ~

    I also made notes of my questions as I went through your Systemigram presentation, but then I came to the last slide and found that you had already posed all the same questions yourself.

    The result of these questions then suggests that this systemigram’s main function is in communicating complexity, rather than understanding impacts of changes.

    ~ Quod Erat Demonstrandum ~

    I was going to close with a comment on how this still feels to me like learning Latin at school – in that it doesn’t relate directly to my everyday life – but then I realised that Latin has influenced my thinking and understanding of language in general, even though I don’t directly translate. I have a feeling that systems thinking may leave its mark on me in a similar way.

    I hope none of this exposes my ignorance, and I would be interested to read your responses to my questions.


  3. Peter

    great questions, and by the way I always interpret the term ‘ignorance’ with the non-prejorative form of ‘not knowing’, and nothing more (I’m in that situation all the time 🙂 ).

    I am now labelling the Information Model as “a tool to identify things we can measure and the relationships between them” and a Systemigram as “a tool to help understand the interconnected effects of changes to a system”
    The latter yes, but also to cope with complexity, to some extent. The former? Typically info models underpin IT systems because they are often more mundane that perhaps ‘measure’, in that they can simply be about handling existence of, tracking status of etc…from this can also come measurement. Think of an accounting system: simply recording balance of accounts, payments in, payments out, to whom and when etc.

    Function of a tool: yes, undoubtedly. What perhaps is interesting (to me) is when I’ve seen (frequently) a ‘tool’ (intellectural or IT) ‘designed’ for one purpose, then used for another…and (over time) the increasing chaos that can ensure (because the information we capture and present using tools is often about (trying to achieve effective) communication. Back to your comments:

    For example, when I invent my tool for arranging raisins in an orthogonal grid, it may not be useful to spend time arguing over the accuracy of the placement.
    Absolutely because there presumably is some reason for why we want to align raisons on the grid – like tokens on a draughts board – it doesn’t matter that a single token is exactly in the centre of a cell, but it does matter that it is unambiguously ‘in’ the cell, so may be up to 25% off centre is perfectly OK. I assume your referene to Einstein is that, for most common human experiences and work a day time / speed / distance calculations, Newton works just fine: if I average 35mph over a journey of 125 miles, I know how long it will take me – and this works for planetary orbit calcs. It’s only if we bust a gut to go speeds approaching speed of light does Newton break down…

    >>> ~ Information model arrows ~ >>>

    To start at the end, we should (any type of modelling) start with the real world, the concepts and interrelationships, and abstract and model as we need to; any model and then any diagram only follows from that.
    In principle whether the relationship is expressed active or passive I’m not sure is so crucial. Take the food item and customer relatonship (perhaps because I want to, for some reason, track food items and customers). I can say:
    – The food item is purchased by a customer;
    – A customer purchases a food item.
    To me the conceptual relationship is actually the same, and just expressed from one sense (one ‘end’), or the other.
    And don’t be mislead here by the use of ‘arrows’ on information models, I dislike them intensely. Using a notation to indicate the direction within which to read the label on the relationship *is* important; there have been various variants of IM notations that have used other conventions (than an arrow): the ER modelling technique used (see and scroll down) a variety of notations, as it evolved from Chen’s original ‘diamond’ notation, to the Crow’s foot notation, and others.
    I hate the ‘arrow’ (rather than crows foot of lolly pop) notation, as it can lead people down the trap of thinking the diagram is showing information flow rather than declarative relationship. (Which is a reference back to using the right tool for the job, AND not confusing tools. Data flow diagrams (DFD) conventionally use elipses for processes, arrows for flows; state transition diagrams (STD) conventionally use rectangles for states, and arrows for transitions. Some DFD forms can use rectangles and arrows, making them look uncannily like STDs; the point is the semantic is completely different.

    The result of these questions then suggests that this systemigram’s main function is in communicating complexity, rather than understanding impacts of changes.
    May be, but also may be these two aspects (communicating complexity, understanding impacts of changes) – and a third understanding influences – are so closely related that it may be not worth picking them apart. Information models can be used to provide insight into a domain, but can also then be used as the basis for a software implementation. You can do the first without the second, but it may be a big mistake to do the latter without the former.

    Ah, remember it well (well a bit anyway, amo, amas, amat…). You know when I was attempted to learn Latin and French I really wish I could have done some computer science (CS), language analysis, abstract and concrete syntax versus semantics at the same time – I think I’d have been considerably more motivated on the natural languages from the insight I subsequently got from the CS awareness!

    Hope this helps!? J.

Leave a Reply

Your email address will not be published. Required fields are marked *