Dave Snowden gave a talk titled "Context is Everything" at the Scaling Agile for the Enterprise 2016 congress in Brussels, Belgium. InfoQ interviewed him about applying leadership models, the Cynefin model and how it can be used for capturing requirements, scaling agile, and sustainable change.
InfoQ: Can you explain why there's a need for multiple leadership models?
Snowden: I think because you have multiple situations. So under conditions of certainty where you can apply constraint, you can control by constraints. You can have a very directive style of leadership. On the other hand, if the situation becomes highly uncertain, then there isn't a previous plan that you can rely on, you need people to actually have the right sort of experience, to make some intuitive decisions but also to carry people with them in following that decision.
Now those are very different styles in my HBR article, we identified four different general styles of leadership and you could look at subsets of those.
InfoQ: So different situations which call for different models to use in there?
Snowden: Yes and I think this is not the one size fits all and you see this in the management literature. So you get a leader who has been in the right place at the right time and succeeded. They then assume they have a general model that will work for many other contexts. For example, you get the mayor of New York first time it ever happens to do the right thing in the crisis, he then writes a book about leadership but he can't even get on the Republican short list. So the trouble is we abstract from the single context into a general model and that's not going to work, right? If you look at historical models of leadership, and this actually links to brain facility, there’s a difference between leading the tribe and holding the wisdom of the tribe let alone asking naive questions to the leaders of the tribe. So even within a collective construct; there are different models of leadership.
InfoQ: For those not familiar with the Cynefin, can you briefly explain it?
Snowden: Cynefin works off three ontologies so it talks about ordered systems with a high level of constraint where you've got predictability and it divides those into two obvious where everybody knows what the constraints are, everybody buys into then and complicated where experts understand the situation (and you have to trust them) or analysis is necessary to understand what to do.
So in the obvious domain you can impose best practice; in the complicated domain, you can apply good practice. The distinction is important.
Then we have the complex domain which has its basis in complex adaptive systems theory. In a complex system, there's so many interacting dependencies that future states cannot be predicted. There constraints can provide a degree of coherence and direction but they can't provide predictability and that's where we manage dispositional states.
Then we have the domain which is always temporary, is where you have no meaningful constraints or connections. If it happens accidentally it is a crisis, but created it can allow for innovation and distributed decision support.
So the whole function of Cynefin is to say which domain are we in, and then engage with different styles of behaviour based on that domain. I often use the metaphor of a children's party to explain the differences. Search on my name and ‘children’s party’ in YouTube and you will find that.
InfoQ: Can you explain what makes the chaos state temporary?
Snowden: Well, because it's very difficult for anything to be unconstrained. If you get a crisis, somebody will impose order pretty damn fast. If you look in randomness in chemistry, you get some order within nanoseconds, some sort of structure will emerge. So if I hold something in that state it takes a lot of energy. Like nuclear fusion, the energy to maintain the magnets to hold the plasma is more than the energy you get out of it. So we use chaos for example for distributed decision support and for innovation but it's a very structured process because to create randomness, it takes a lot of energy.
InfoQ: That's why it will self organize?
Snowden: I mean you can fall into it by accident and then it's catastrophic but sooner or later, somebody will impose constraints pretty damn fast.
InfoQ: Can you give some examples of how the Cynefin framework can be used to keep the user requirements?
Snowden: Well, that's what we're working on at the moment so what are we doing? We have user requirements against known needs (for example going back a few years we were writing a computer program to actually automate a double entry bookkeeping). In Cynefin terms, that's obvious.
In complicated we have a range of options to choose from. And that's a classic business analyst problem.
Then we hit the complex. In these days technology is advancing faster than user needs evolve to require the technology. What we do in the complex space is we're getting panels of users to continuously recording observations, frustrations, day to day issues using SenseMaker®(our software product). When a pattern emerges from those observations, people from IT can see if they got technology which will make life easier and only if that proves itself does it move into a Scrum or a development process. So what you're doing is to match unarticulated needs with unknown capabilities. That gives considerable competitive advantage and is going to be more and more important over the next few years in technology enhancement.
InfoQ: What are the benefits of using this Cynefin approach for capturing the requirements?
Snowden: You radically reduce the costs in terms of wastage. If you go to any agile conferences, people talk about high levels of failure, now there are two responses to that. One is to accept it, the other is to say, well, maybe it's because we haven't allowed enough ambiguity early on.
So what we think will be the benefits (and this is early days) is before you even commit to a two or three week sprint, you might have had five pair programming teams work for a day each on a set of unarticulated customer narratives. By this you radically reduce the uncertainty associated with user requirements definition and the bias of the analyst. You reduce the cost of development, you also increase the possibilities, you use technology to leapfrog your competitors because you'll see an opportunity to use technology that users haven't articulated yet. So you're talking about reductions in cost but also radical reductions in opportunity cost.
InfoQ: It's both for reducing waste in your process but also speeding up the process?
Snowden: Yes and I think it's more a lean approach to waste as well. So by moving programmers closer to user stories, that's kind of like the lean principle in operation.
InfoQ: Can you elaborate about the key steps that are needed to scale Agile methods to cross an organization?
Snowden: Well step number one is do not go anywhere near a pyramid planning scheme like SAFe. It's a backwards step. I think the issue on the scaling in the complex system is you scale by decomposition and recombination not by aggregation. So one of the things we’re doing for example is take strategic needs, operational needs and technology capabilities, capturing those in real-time, emerge them together in fitness landscapes to see where they start to cluster together and so that scaled by decomposition reactive recomposition.
I think the other thing is to recognize that strategic functions in companies and operational functions in companies’ area actually more agile than IT. IT often acts as an inhibiter so the idea you can take an agile practice and apply it to the C level outside of IT will get a poor response. Their response is, “Guys, we've been here for years” time you caught up, right? It's called rapid response, there's a whole body of names for this. So I think one of the issues is to find new exchange mechanisms between strategic needs, operational needs, technology capabilities and that's where we're also taking these requirements capture.
I think the other issue is to recognize there's value in waterfalls, waterfall design, and there's value in time boxes that last more than two weeks. So bringing back waterfall development, bringing back in the DSDM time box techniques together with Scrum, developing pre-Scrum processes, and then creating new approaches to Kanban which represent all of that, you scale by creating organic models of how things are combining and recombining. You don't scale by huge engineering diagrams with lots of pictures and references to trains: trains run on tracks and the last thing you want is tracks in a complex system.
InfoQ: You're not talking about a top-down approach, you're talking of a much more organic approach?
Snowden: I'm talking about top-down enablement of bottom-up actions. So the role of top is to create constraints and relax constraint as needed and then dynamically allocate resources where in patterns of success emerge; you don't do that through micromanagement.
InfoQ: Is it like setting the conditions top down and then filling it in bottom up?
Snowden: Yes but you also need real-time feedback loops. so we are working at the moment to allow executives to ask questions to the whole work force in five minutes and represent the results as a fitness landscape. So you're talking about much more into connectivity and much more necessary ambiguity. Lots of leadership models say that you should remove ambiguity and be very clear, we're saying that if you're in complex, you should actually increase the ambiguity because decreasing the ambiguity is a bad idea, it means you miss things you need to see.
InfoQ: Is this the same kind of thinking that you see in antifragility?
Snowden: Yes. I have mixed feelings about Taleb all right? He's one of the best of the popular authors, 'Surprise by chance’, is a brilliant book, Black Swan less so. Antifragility for me is a sub section of resilience. I think the trouble is he wants to make antifragility unique instead of making a hugely valuable contribution to the field as a whole. Resilience is surviving with continuity of identity at the time. Systems which become stronger when they fail are sub-function of that and not the totality.
InfoQ: What are your thoughts on sustainable change? How can leaders make it happen?
Snowden: They need to know when to apply directive leadership and when they need to change the constraints. There's times where you take control and we have whole models on this: sub-models within Cynefin, there's a separate set of models for complex settings and models of chaos and there's times when you move in to take control and there's times when you stand back. It's called multi ontology (type of system) sensemaking.
Different types of system, different leadership styles, you need to be able to understand -- you need to be ontologically aware so you can change the style and apply distributed or centralized leadership. Sometimes you're indicating directions of travel, sometimes you're indicating specific goal. If you indicate goals, you'll achieve what you ask for but that might not be what you needed. If you need achievement and you indicate direction, you might discover valuable alternatives you could not have planned for. There's a British phrase "horses for courses." If you got wet ground, you bet on different horses than on dry ground and leadership needs to be more flexible right now.
InfoQ: I recognize the styles of leadership. I'm working with teams all around the world, which are translating my books and I usually start off with giving them as much freedom as they want and usually at some point in time, they say, “Okay, but we need a little bit more direction, we need a little bit more telling us what to do" and then we find a way for that until we get to a point and they say, “Okay, well we can do this ourselves, so can you back off.”
Snowden: In traditional models the apprentices will become a journeymen at which point they teach. Then some of the journeyman will become masters. Experience and peer interaction is key. You can't reduce it to simplistic training courses, competence models and airport books.. That approach is dangerous, and we are seeing it in Design Thinking, management development and many fields, as it is trying to codify an artisan process which cannot be codified. They are trying to reduce something which takes two or three years to understand into a two week training course. That's criminal.
About the Interviewee
Dave Snowden is founder and chief scientific officer of Cognitive Edge. His work is international in nature and covers government and industry looking at complex issues relating to strategy, organisational decision making and decision making. He has pioneered a science based approach to organisations drawing on anthropology, neuroscience and complex adaptive systems theory. He is a popular and passionate keynote speaker on a range of subjects, and is well known for his pragmatic cynicism and iconoclastic style.
