More on that enterprise-architecture ‘help needed’

Given the responses to my previous post ‘Guess I could do with some help here…‘, seems it’d be useful if I clarify a bit more what kind of help I most need. (Or we need, rather, as an industry and discipline: probably the only ‘I’-part here is that I seem to be one of the few at present who’s asking these questions?)

To be honest, we don’t need much help in thinking about the nature of enterprise-architecture or the like: that’s already well covered, and it’s actually quite straightforward anyway.

Where we need help is in rethinking the toolsets that we use for enterprise-architecture and the like.

To me it’s clear that if we’re to make sense of the enterprise, and support viable decision-making about the true whole-enterprise-scope within which our organisations operate, we’re going to need some kind of holographic approach to modelling.

We already have plenty of frameworks for enterprise-architecture. We also have plenty of methodologies to work with those frameworks. Each of those is constrained to some variously narrow view into this holographic space, and usually constrained by placing some particular point as ‘the centre’ – hence IT-centrism, business-centrism, health-and-safety-centrism and so on. And those constraints are useful in practice: no doubts whatsoever about that. So to my mind there’s nothing whatsoever that’s inherently ‘wrong’ about any of those frameworks or methods, other than their own occasional internal inconsistencies and than that they too-often position their own very limited perspective on reality as ‘the only possible’ view on reality. That’s the only problem there, and it’s very easy to fix, by acknowledging the value of a single viewpoint as a viewpoint yet also insisting on cross-viewpoint generics. We can choose anywhere as ‘the centre’, for some given purpose, yet must also insist that everywhere and nowhere is ‘the centre’, all at the same time.

Hence, as such, we don’t need any new frameworks or methodologies for enterprise-architecture. We have more than enough of them already, to be honest.

What we do need are better ways to manage and make sense of the information we have about this ‘enterprise-hologram’.

Which is where toolsets come into the picture. And that‘s the part that I really need help on – or we need help on, rather, because it’s definitely too large a context for any one person, or even any one group, to tackle on their own.

What we need most is clarity on the question we’re aiming to address. I think it’s an Einstein quote that says “if I had an hour to solve a problem, I would spend the first fifty-nine of those minutes on the question, and only the last minute on the answer” – because the right answer tends to fall out all by itself when we have clarity on the question.

One thing we do know, though, is that there are many different players all trying to tackle different aspects of the same enterprise-holograph – for example:

  • enterprise-architecture and all the other domain-architectures and solution-architectures – an emphasis on structure and purpose, and how they link together to deliver useful value
  • knowledge-management – an emphasis on how knowledge-items link together (especially narrative-knowledge, stories and ‘subjective truth’)
  • change-management – how everything changes and can be changed over time
  • process-management – how activities link together, and entities flow between them, to add value across supply-chains, value-networks and so on
  • service-management – the human and technical aspects of keeping everything going and ‘on purpose’
  • futures and other business-intelligence – how to trawl through the enterprise-space for sensemaking and the like
  • simulation, ‘gamification’ and other skills-development – how to apply ‘what-if’ to any part of the overall context
  • skills-management – identifying the skills needed, and the training and/or recruiting needed to cover present or future gaps
  • performance-management – identifying required metrics and their impacts (especially, to avoid ‘gaming the system’)
  • governance – identifying requirements and responsibilities, assuring assignment and ability to comply, and verifying compliance

In principle, yes, they’re all views or viewpoints into the same overall context. Yet each of those views also carries information or requirements that are specific to that view alone – in other words, more like an orthogonal dimension. And there are a lot of those distinct dimensions in there – an n-dimensional space, where n could literally be any number. (Certainly a lot more than are accessible in the simple flat images so typical of so many current ‘EA’-toolsets, anyway.)

Then there are all the different uses for all those distinct views: an architectural view shows us relationships between structure and purpose, for example, but do we need that view to decide on future strategy, to plan investment, to compare different implementation-options in trade-off analysis, to guide scenarios and substitutions in planning business-continuity and disaster-recovery? The views we’d use might at first seem similar, but the focus and emphasis and model-dynamics in each case might be very different.

At first glance this all might seem impossibly huge. Yet it’s not as hard as it seems. Most of the technology we’d need to deal with all of this complexity does already exist. Despite the huge spread of the overall toolset-ecosystem, most of the key software components we’d need are already easily available, at little or no direct cost, running on a wide-range of easily-available existing devices. Conceptually speaking, the underlying data-structures are straightforward enough: for example, it could be done with little more than freeform tagging in an object-database of some kind, with some kind of filters applied to the tagging. The same applies to search, and filtering: pretty much everything we need already exists somewhere, if we knew where to look. And even if display-technologies are not yet quite capable of showing a true n-dimensional holographic space, they’re certainly capable of simulating one. Given the right people and the right ideas, the technology side really isn’t all that hard these days.

But if the technology isn’t hard, the user-experience part is hard. And seems to me that that’s where we most need to focus our attention at the moment.

In many cases, we simply don’t have the right metaphors for model-relationships:

  • some of it can be usefully described as a matrix – but it’s definitely more than a simple matrix as per Zachman
  • there are some contexts in which a metaphor of hierarchical layers will sort-of make sense – but it’s definitely more than a simple ‘stack’ as per TOGAF or Archimate, or even a multi-axis matrix-stack as per CapGemini IAF
  • there are flows of information and materials – yet it’s much more than a simple supply-chain
  • there are identifiable relationships, including realisation, aggregation and so on – yet much of it tends to follow a modal logic of possibility rather than solely a simple true/false logic of presence or absence
  • there are identifiable trails of derivation and decomposition – yet there’s more to it than a simple Zachman layering, or the classic ‘current state’ versus ‘future state’

And perhaps more important, we don’t yet have adequate user-interface metaphors:

  • drag-and-drop for entities can be misleading – is it a class or an instance? how do we link back an instance back to a parent class? are we editing the instance only, or the whole class? are we affecting other instances elsewhere? or other versions of the same nominal instance? what happens if the parent disappears over time, but the instance continues as re-linked to something else?
  • notations can be confusing – especially where the same nominal entity would appear in multiple views with different notations or visualisations or images
  • aggregation (as in Zachman primitives versus composites, TOGAF ABBs versus SBBs [Architectural Building Blocks versus Solution Building Blocks]) can be very confusing – especially where entities can recombine in many different forms, and at different levels of abstraction or realisation
  • zooming (as per Prezi) works well to describe a ‘containment’ or hierarchical concept of structural-decomposition – but it’s hard to make sense, if entities aren’t fully ‘contained’, or if there are more than two orthogonal axes
  • timelines (as in Gantt-chart relationships, or Apple OS-X Time Machine backup/restore) provide a good means to step through time-related views – yet the views themselves may change over time
  • multi-axis user-controls work well for up to three or four exactly-orthogonal dimensions – but become exponentially harder to use with increasing numbers of dimensions and other variables, and probably cannot cope when the dimensions themselves blur into each other

In short, we urgently need new user-interface metaphors to navigate through n-dimensional holographic space, where the nature of the space itself may change as we go.

(Oh, and it has to be easy to use, too, such that both the navigation and what’s visible in the view will make immediate sense to anyone. Of course.)

To use another Einstein quote, “Everything must be made as simple as possible, but not more simple”. Simple to use; yet actively dissuade overly-simplistic ‘solutions’ such as IT-centrism and the like.

That’s the challenge.

And that’s also where Agile and the like come into the picture – and, most of all, people who are experienced in Agile-style software-development for toolsets and the like. We seem to have quite a lot of methodologists and the like around here, who tend to be great on the theory. But what we need most are developers who know how to think seriously-sideways and put it into practice.

We can’t just talk it into existence: we need to get past the ‘talking about it’ stage now. Given the blunt fact that we’re very unlikely to get it right first time, we need something to play with, to test in real-world practice, to review and rethink our ideas, to help us clarify what the question really is that we’re facing here.

More thinking-about-thinking, about the theory and the like, well, yes, we’re always going to need it, it’s always going to be a ‘nice to have’, and so on. But right now, we’ve done more than enough thinking-about-thinking: it’s time to get down to the doing, of creating this new kind of toolset.

Anyone interested in helping with that? Please?

7 Comments on “More on that enterprise-architecture ‘help needed’

  1. Hi Tom,

    May I suggest to converge the relevant concepts / tools / frameworks / ideas in a nice overview? I have an excellent example: The Framework Quagmire.

    NB: Don’t look at the /content/ of the picture (the frameworks) but to the structure / idea behind it: it shines light on a quagmire by relating ‘things’, marking the ‘kind/type of things’. So the example should not be viewed as an answer to your question, but as a practical means to converge attention to what you actually (don’t) want and to put the result in position/perspective with its ‘environment’.

  2. Tom, I would advise to start with an existing open source platform and use that for a while in a plain and simple way before even starting to think about improving or extending it. It is crucial that you have a stable “small” foundation/backbone that is understandable for everybody who wants to join your initiative. This also forces you to think about the core capabilities you want to have right from the start in your new kind of toolset.

    I only have knowledge of one possible candidate for such a platform:
    The team behind Cohere are thinking about similar things as you do, creating an emerging suite of sensemaking tools

  3. Nice challenge. I would start from basics: Any complex system is a dynamic set of artefacts. Artefacts are interconnected and interdependent. We have to anticipate potential changes of the system and implementation of such changes necessitates the evolution of some artefacts and the relationships between them. So,

    – All artefacts must have formally described (i.e. be good as services)
    – All artefacts must be versionable throughout their lifecycle
    – All relationships between these artefacts are modelled explicitly
    – All models are made to be executable

    1. collect artefacts
    2. find and formalise relationships between them – Note: some artefacts are relationships
    3. run the simulation of that model (for a particular system)
    4. change (iteratively) the model to get the desired effect
    Also, we may consider refactoring of the system to be easy to modify all artefacts and relationships without causing any negative effects.

    So far, only the item 1 is more-or-less known (thanks to EA we know majority of types of artefacts). The majority of work will be in finding and formalising relationships (it may be a recursive “procedure”). Thanks to BPM we know one type of relationship — processes. Certainly more types to be discovered.


  4. @Ron van den Burg, @AS – Ron, Alexander – yes, strongly agree to all of that. I’ll do another post on that as soon as practicable (there may be a brief delay because I think I’m going down with a flu at the moment 🙁 ) Main point I’m looking at is that I think we need to go right back to the root-level, far beneath any specific framework or model-type. More on that when sanity permits, anyway.

    Thanks again, of course.

  5. @Peter Bakker – Thanks for those links, Peter – very useful indeed. A lot of trails to follow there (and elsewhere, such as CMAPTools and Southbeach) but yes, would strongly agree that the whole point of this kind of exercise is sensemaking and decision-support. That’s why, as above, I think we need to go right down to the roots – possibly beyond even a metametamodel such as MOF – in order to get the versatility we need. More on that in the upcoming post, anyway.

  6. Tom, I made a “Rethinking the EA Toolset” tubemap based on your articles and posted an article on the making process at with a bit of insight on my own ideas.

    I’m thinking about writing an additional article with some deeper insight on the additional lines. But at this moment I’m curious if you can recognize your ideas in the tubemap and how you feel about the additional lines of thinking (simulation and storytelling).

Leave a Reply

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