Ask The Outcomes Guru - 'No outcomes question too simple nor too hard'


M&E and capacity building specialist Lesley Williams (Linkedin Profile) asks: How do linear models address the complexity in which we work?


There are two issues here, the first is the size of the model being built and the second is whether or not a model is ‘linear’. What I’m striving for in my work with visual outcomes modeling is to encourage the development of models that are ‘simple but not simplistic’. This relates to both the size of the model and the number, and type, of interactions (lines and arrows) drawn between boxes within the model. Here’s an example of multi-page model that I built in the form of a best practice template for a mental health service

Size of models

In terms of size, a multi-page model such as the mental health one provides enough room to model a fair degree of complexity. I think that a model should include as many sub-pages as are needed to fully describe the program that’s being modeled.  

In contrast to multi-page models, people often drawn single page visual models of programs. In my view, regardless of their usefulness as an overview, a single-page model does not adequately convey the complexity of what’s happening within most programs. Don’t get me wrong, I’m not complaining about people drawing program overviews - it’s just that you should be able to instantly drill-down to lower-level pages to get access to as much detail as you need. 

Should models be ‘linear'

While I think that models can be large, we need to be careful about making them too visually complex at this stage in socializing the widespread use of visual modeling in planning and evaluation. 

The first point is that the flow of time is generally viewed as linear and I don’t think there is a problem with this being reflected in our modeling. However the causal relationships between boxes that are linked within models is by no means linear in the mathematical sense that an increase in one box will be followed by the same amount of increase in another box. All sorts of causal relationships are possible between boxes with thresholds and cutoff points etc. In addition, there may be feedback loops in operation. 

Ultimately I would like us to be able to model all of this visually and I am hoping that we will gradually built the appetite for this amongst stakeholders at the same time as we improve our software visualization tools to do so (animation and Virtual Reality would be cool in this regard). At the moment I think that you need to at least be able to connect any box with any other box (if you have a multi-page model this needs to be able to be done across pages) and be able to model feedback loops where one box influences, and is influenced by, another box in any software that you use to draw your model.

While I use box-linking extensively to show the relationship between projects and higher-level outcomes in ‘line-of-sight’ alignment, at the moment I normally don’t do a lot of linking of boxes within the higher-level modeling of outcomes and the steps leading to them in the models I use when communicating with stakeholders. 

The way I put it to stakeholders who are looking at my models is that there are many links and potential feedback loops between the boxes on the page. I actually put a note to that effect on the bottom of most pages in my models. I then just show a general movement of time with single arrows pointing from left to right (as in the mental health example). I use a gray filled arrow to signify a ‘then’ statement and a non-filled arrow to represent an ‘and’ statement and leave it at that.

First rule, ‘Don’t scare stakeholders'

As I said before, I think that further down the track we will get better ways of visually modeling complexity which will be less overwhelming. Equally importantly, stakeholders will have more of an appetite for more complex modeling. At the moment there is a trade-off between modeling complexity and turning off stakeholders.

The cause of visual model can potentially be put back by years if just a few high-level stakeholders don’t get our models. This happened in the case of the diagram shown to General McCrystal in Afghanistan where he said something along the lines of, it would be easier to win the war than understand this diagram!

I talk about this example in this Fulbright Seminar.

Keep it simple for stakeholders but go to town in backroom work

Here is an example of an outcomes model with links and one without. I think that the second model is more likely to be accepted by stakeholders at the present time. However this does mean that the more complex model could not be used as a back-room analytical tool. 




In conclusion, I think that we need to allow models to be as large as they need to be. There will be a linear aspect arising from the fact that time passes in a linear fashion. But there will be non-linear aspects from thresholds, cutoffs and feedback loops. While we can try to model as much complexity as possible in back-room work, there is an argument for trying to keep models reasonably accessible in stakeholder-facing work until we increase stakeholders appetite for visually modeling complexity and we get better at it from the software point of view. 


If you would like to discuss this post you can do so on the Linkedin Pulse version of it.

Lift your outcomes game by subscribing to the DoView School of Outcomes Newsletter

Ask the The Outcomes Guru any question about outcomes.

Full disclosure: Dr Paul Duignan is involved in the development of DoView software.

Dr Paul Duignan