Strategy Is a Learning Process

#model #method1 mention

Strategy is de facto always an iterative learning process, even if this is often not made explicit and information gaps between iterations make it less effective and efficient.

The process of strategy is not a linear process but an iterative cycle.Wardley (2018), 15

To maximise overall learning in a changing Environment, one needs to minimise cycle time, not maximise the information processed in each step of the process.Bungay (2012)

Since Strategy is based on models, strategic learning can be analysed in terms of creating and updating Generative Models.Friston et al. (2017)

This can happen consciously (based on an Explicit Model) or unconsciously (based on an Implicit Model).

The learning process has a generic structure.Generalised from Boyd (1995), Bungay (2012), Lafley & Martin (2013), Rumelt (2011), Scotland (2020), Sun Tzu (see Schön 2020 and Wardley 2018), and Wardley (2018). For a structurally equivalent model of learning in the brain (as a biological/cognitive system) see Friston et al. (2017), 19. The process diagram is adapted from Boyd’s “OODA Loop” (Boyd 1995).

Making it explicit helps identify and close gaps as well as minimise cycle time.


The system’s Purpose sets the overall goal of the strategy. In evolved Systems it is given or emergent over the system’s lifetime; in designed systems it is decided (or, to be more precise, strictly also emergent, but shaped by actors with Power).

Organisations have characteristics of both evolved and designed systems, so their purpose will be partly evolved and partly shaped by specific actors.

A strategy can only be successful if it pursues and furthers the realisation of the actual, not the espoused purpose of the system.“Purpose” is equivalent to: Purpose (Sun Tzu, Wardley, McCandless & Schartau), Ambition (Lafley & Martin), Aspirations (Scotland)


Perception, i.e. data gathering, informs the system about its own state and that of its environment. It forms the basis for context or situational awareness, which in turn is the prerequisite of successful strategising.

A highly complex context requires a high diversity of data sources and/or frequency of observations.“Perception” is equivalent to: Observe (Boyd), Outcomes (Bungay), Evidence (Scotland)


Based on perception, the context of the system is represented in an explicit or implicit Model. This generates context awareness and enables a situational assessment.

The diagnosis for the situation should replace the overwhelming complexity of reality with a simpler story, a story that calls attention to its crucial aspects. This simplified model of reality allows one to make sense of the situation and engage in further problem solving.Rumelt (2011)

This requires identifying and assessing salient context features:

The adequate mode for modelling depends on the characteristics of the system’s environment:The following is adapted from Kurtz & Snowden (2013) and David Snowden’s further development of the Cynefin framework.

Modelling informs perception as well as (directly and indirectly via decision) Friston et al. (2017)

“Model” is equivalent to: Landscape and Climate (Sun Tzu), Orient (Boyd), Diagnosis (Rumelt), Insight into the basis of competition (Bungay), Where to play (Lafley & Martin), Map and Where (Wardley), True North (Scotland)


Based on the model and the situational assessment it enables, policies for action are selected. This means prioritising and directing resources and thus the setting of Enabling Constraints.

Building blocks for policies can be taken from two different sources:

In choosing building block, trade-offs between simplicity and stability on the one hand and adequacy and adaptability on the other are unavoidable.

Decision generates immediate input for perception, thus feeding the learning process.“Decision” is equivalent to: Doctrine and Leadership (Sun Tzu), Decide (Boyd), Guiding policy (Rumelt), Plans (Bungay), How to win (Lafley & Martin), Doctrine and Gameplay (Wardley), Strategies (Scotland)


The ultimate realisation of a strategy (and the test of its viability) is action, i.e. the overt system behaviour resulting from the chosen policies.

The alignment and coherence of actions is based on and constrained by the previous steps of the process, i.e. they create internal Path Dependence. To mitigate this, two meta-strategies are useful:

Action generates input for perception, thus closing the learning loop.“Action” is equivalent to: Act (Boyd), Coherent actions (Rumelt), Actions (Bungay), Core capabilities & management systems (Lafley & Martin), Tactics (Scotland)