2013 August — Discerning Action

Principles for Adaptive Iteration – crystalized by a comparative review of similar approaches

August 23, 2013 By Markus Leave a Comment

I have just completed a Discussion Draft of comparative review of eight existing approaches that are similar to Adaptive Iteration. I have published the comparative review as a White Paper here. One of the benefits the review is that it has helped clarify and crystalize the essence of Adaptive Iteration in the form of a set of principles – still in draft at this stage. I listed the principles towards the end of the White Paper and repeat them here for you to review and comment.

Principles for Adaptive Iteration

Adaptive Iteration applies to all non-random situations where behaviour and performance cannot be predicted with confidence in advance. In the main, this applies to situations that are partially structured and emergent. Because of dramatically increased connectivity generated by developments such as globalization and by technologies such as the internet, situations that require Adaptive Iteration are increasingly prevalent and of increasing significance.

Adaptive Iteration involves the co-evolution of learning and action in an iterative four phase approach:
- Clear and unbiased observation of the situation, including contextual influences
- Interpretation of the observed data to improve understanding of the trajectory of and the influences on the situation
- Incorporation of the interpretation insights into a hypothesis for an improved design
- Conduct of one or more experiments to assess the impact of the proposed improved design.
Adaptive Iteration is underpinned by a clear and shared understanding of the fundamental drivers for and constraints on the adaptive decisions. This provides coherence and alignment for all decisions and provides lower level scope for the adaptive process to explore novel and unconventional options. It also provides a clear driving force for adaptation and against the status quo.

Adaptive Iteration may be a nested approach in which a lower level Adaptive Iteration cycle is used as the experimental phase for a proposal for a higher level design change. Because of its multi-level applicability, Adaptive Iteration should be a core capability throughout any organization.

Rapid iterations of Adaptive Iteration are important during the early stages of an adaptive transition. These iterations promote critical early learning and integration across the four phases. They also help shake-out any lack of shared clarity about the fundamental drivers and constraints.

Adaptive Iteration co-exists with analytical action. Where the impact of a design variable can be predicted efficiently and with confidence, Adaptive Iteration is not required and the design changes should be subject to ‘analytical action’.

Filed Under: overview

How many ball passes between the white team?

August 16, 2013 By Markus Leave a Comment

Before you read any more of this blog post, please watch the video below (even if you think you have seen it before).

(Video courtesy of http://www.dansimons.com/videos.html)

As the developers of this experiment, Christopher Chabris and Daniel Simons, emphasize in their book, The Invisible Gorilla: How our intuitions deceive us, the key message is that we are strongly biased to only see what we pay attention to and what we expect to see. They call this inattentional blindness (blindness to what we do not pay attention to).

A related experiment demonstrated that the percentage of people who fail to see a prominent, but unexpected, object directly in their field of view increases to 90% when the complexity of the ‘attention’ task increases. In other words, the likelihood of inattentional blindness increases if we need to concentrate hard to pay attention.

Now let us translate these findings to the Observe phase of Adaptive Iteration. Adaptive Iteration applies primarily to situations that are unpredictable, uncertain and emergent – precisely the situations in which the unexpected may occur. Add to this the characteristic that these situations are usually complex – unclear cause and effect relationships, many elements, interactions and decision points, dynamic context, etc – and require much effort (attention) to attempt to determine what is happening. So, in these situations, we have a relatively high likelihood of something unexpected occurring at the same time as we need to concentrate hard to attempt to understand what is happening. We have the perfect conditions for inattentional blindness.

What can we do to reduce the potential for inattentional blindness? I think there are several things:

Deliberately observe the situation from multiple perspectives. I have summarized this approach, including a range of observational perspectives, in a previous blog post.
Separate, as much as possible, understanding the situation from observing the situation. We can reduce the distraction of attempting to analyze and understand the situation by separating it from the effort to observe the situation. We focus on observing what is happening rather than attempting to explain what is happening. In particular, we suspend our theories and preconceptions so that we do not focus on those observations that confirm them. (This point summarizes one of the key reasons why we have explicitly separated Observe and Interpret in the Adaptive Iteration framework.)
Capture a data rich record of the situation. As with the basketball situation, if we capture a data rich record we can replay it multiple times so that we can observe things that we may have missed. Replaying the basketball video provides very strong proof that a gorilla did walk through the players even we are convinced from our initial observations that it did not. One word of caution with this approach – we need to ensure that we do not focus our data capture only on those things that we think beforehand are relevant and important. This will formalize our observational bias. The record capture needs to be as broad, detailed and unbiased as possible.
Don’t attempt to understand the situation – just focus on what is and what is not working. Many unpredictable, uncertain and emergent situations are so complex that it is impossible to understand and analyze them in detail. Often, we should just accept this and rather than attempting to understand and explain the dynamics of the situation, focus on what is obviously working and and what is obviously not. During the subsequent Interpret phase of Adaptive Iteration, we can then identify what sort of things we can do (design changes) to promote that things that are working and to suppress those that are not.
Use multiple observers, some of whom are not specific experts. The researchers experience with the basketball video is that approximately 50% of observers see the gorilla walk through the players. Increasing the number of observers will increases the likelihood that at least one observer will see the unusual and unexpected. However, it is important that there is diversity among the observer. Especially, it should contain some people who are not experts in the situation (see earlier blog post) and there should little or no opportunity for the group to develop ‘group think’ beforehand.
Practice mindfulness. Sam McNerney in an article in TheCreativityPost highlights research that suggests that mindfulness, defined as “paying attention to one’s current experience in a nonevaluative way,” may provide an effective means of observing a situation or circumstance without seeking to confirm our intuition and expectations. In other words, by being consciously aware of our own perceptions and thoughts we can learn to recognize and compensate for the filters they place over our observation and attention.
Iterate rapidly. The iterative nature of Adaptive Iteration means that we do not have to be ‘right first time’. If the unusual or unexpected event is not just a result of natural variation in the situation (i.e., ‘noise’), it is likely to occur again in a future iteration, especially if the iterations are rapid and involve relatively small design changes. Although this is not the main benefit of rapid iteration, it is a significant one.

Filed Under: observe

Dancing Guy – when would you join in?

August 15, 2013 By Markus Leave a Comment

The ‘Dancing Guy’ video below provides a clear and striking example of non-linear behaviour. While you are watching the video put yourself in the shoes of one of the observers. What would be your reactions? What sort of things would influence them? Would they change over time? What would it take to trigger you to join in?

Are there parallels here with change in organizations and in businesses? Perhaps the adoption of organizational change sometimes happens this way. Or possibly the video mirrors the progress of the introduction of a new product, or the introduction of an existing product into a new market.

For me, the video highlights a range of key observations about creating change in unpredictable and uncertain contexts:

For the first third of the video all we see is one man and then a second dancing enthusiastically by themselves. But is this all that is happening? It is all we can see externally, but a lot is also happening in the brains of the observers. I suggest that many of them are starting to, consciously or subconsciously, think about what it would be like to join in. They are weighing the fun and exhilaration of dancing against the risk of standing out and potentially appearing silly. In a sense, the observers are progressively being ‘primed’ to join in. So, in our change initiatives, we should not just rely on our external observations to monitor progress. There is likely to be much happening below the surface.
What would have happened if the ‘dancing guy’ had stopped just before the first person joined him? Would that second person have still got up and danced, this time by himself? Would a cascade still have happened? I think not? This suggests to me that in complex change perseverance is critical. If we are convinced that the change is important and valuable, we need to continue to, metaphorically, ‘dance and wave our arms’ to ensure that we are noticed and our message is heard.
If creating a dancing crowd was his objective, considerable effort was expended by the ‘dancing man’ before he saw any return for it. When he did start to get a return for his effort, he achieved a very large return for little additional effort. A traditional benefit-cost review at any time during the first minute would have suggested that he ‘cut his losses’ and stop dancing.
Would the ‘dancing man’ have attracted the same attention and had the same impact if he was not dancing so extravagantly and enthusiastically and so obviously enjoying himself? I think not. Sometimes, if we want to create change in uncertain and unpredictable situations, we need to be extravagant and enthusiastic to rise above the noise and the fear. This creates risks, because the crowd might not join us, but they are much more likely not to join us if we don’t take that risk.
Would there have been the same impact if a small group of people, say six, had started dancing together in the same way? I am not sure. From one perspective, I think group may have been perceived as a clique and it would have been more difficult for the first additional person to join in. On the other hand, the risk of standing out would have been reduced for the first additional person. I think there is a lesson here when we start a change initiative with a pilot, especially if the pilot is in a distinct and very coherent area or group.

Perhaps the ‘Dancing Guy’ video triggers other thoughts and perspectives for you. If so, please respond with a comment.

Filed Under: examples, interpret, non-linear

The ‘snowball effect’ – a powerful force in emergent situations

August 13, 2013 By Markus Leave a Comment

Have you ever wondered what is happening when a social media video goes viral? Viral videos are the result of self-reinforcing feedback (sometimes called the ‘snowball effect’ or ‘virtuous/vicious cycles’). If a video is sufficiently engaging, those who watch it will share it with their friends. There is a good chance the friends will also enjoy it and share it with their network of friends, and so on. Similarly, the more hits or ‘likes’ a video gets, the more likely it is to move up popularity based ranking lists which will, in turn, stimulate more hits and ‘likes’ that will drive it further up the ranking lists. So we see at least two positive self-reinforcing feedback loops driving the success of viral videos. One based on awareness generated through social media and the other based on attention generated by rising up ranking lists. Because of the multiplier effect each time around the self-reinforcing loop, rapid non-linear growth occurs.

My experience tells me the snowball effect is one of the most significant features of complex and emergent situations. It is one the key reasons why small initial changes (weak signals) can sometimes have unexpectedly large consequences or impacts. These, in turn, contribute to the unpredictability of such situations.

The snowball effect (self-reinforcing feedback) is at the heart of the development of most deep personal expertise. Several feedback loops are happening in parallel:

I develop initial expertise. This proves valuable for me in work or social settings, so I am motivated to develop the expertise further and to apply it with more confidence and more deeply. This, in turn, makes the expertise even more valuable to me, and so on.
My expertise is noticed by others. This makes feel good and so motivates me to develop the further and apply it more widely which, in turn, increases the number of people that notice my expertise, and so on.
The more I apply my expertise, the more I understand the nuances of the expertise and how it applies in various circumstances and this, in turn, creates more aspects that I can learn and develop (I like to think if it in terms of creating more ‘edges’ to my expertise that I can work on) which creates more and deeper opportunities to apply the expertise, and so on.
The more I apply and develop the expertise, the more likely it is that it will come to the notice of people who can help me develop further, either through coaching or advice, or the provision of more challenging opportunities to apply it. This creates further opportunities to develop and apply the expertise, and so on.
The more I apply my expertise, the larger the library of related patterns and associations I develop in my brain and this, in turn, increases the speed and confidence with which I can apply my expertise. This, in turn, grows the opportunities to apply my expertise which builds an even larger and more nuanced library of patterns, and so on.

As an aside, I think we see the snowball effect evident in many students that excel in a particular area. They are not learning to get a good grade, they are learning because one or more of the above self-reinforcing feedback loops have been triggered. The excellent grade is just a by-product or confirmation.

The growth and development of expertise via the snowball effect is not controlled, managed or predictable. Although the broad progression path for the type of expertise may be able to be anticipated, the specific opportunities for development are not prescribed, planned and programmed in advance. They are determined by the unfolding (emergent) situations triggered by one or more of the above self-reinforcing feedback loops.

The proceeding discussion has focused on the development of individual expertise, but I believe that it also applies to the development of capabilities at the group or organisational level. It also applies to the development of many successful product or business models. Initial success creates increased learning, exposure and reputation and that drives increased funding (sales and investors) and product improvements that, in turn, creates further success, and so on.

I believe the snowball effect (self-reinforcing feedback) is at the heart of successfully navigating many emergent and complex situations. I will explore this further in future blog posts.

Filed Under: interpret, non-linear

Experiments change as we progress through Adaptive Iteration

August 9, 2013 By Markus Leave a Comment

One of the key characteristics of situations that are uncertain, unpredictable and emergent is that we cannot rely on knowledge, expertise and experience to deduce what is happening and how to change the situation to achieve our objectives. The interactions are usually too numerous, complex and/or too subject to human choice or unforeseen events to be able to determine in advance the nature and sequence of all the cause and effect relationships. In these situations we need to rely on observing what is happening to determine ‘what works’ and to discern the broad-based influences – we need to let “the situation talk back”.

Observing natural behaviour will give us some insights about the situation and how we may be able to influence it to achieve our objectives (create organizational change, launch a successful new product, improve the performance of a team, etc), but often we also need to perturb or probe the situation to test our ideas and insights and to generate further insights. This is the ‘experiment’ phase of Adaptive Iteration.

An interesting way to look at experiments is by analogy with natural selection in the process of evolution. By conducting experiments and then retaining what works in the next round of design improvements, we are imitating the evolutionary process in which the naturally occurring variants that survive selection pressures become the basis for further variants, and so on. In Adaptive Iteration we let the situation give us feedback about what is working and was isn’t, and we then incorporate this learning into the next design variant. Using this analogy, it often may be appropriate to conduct a number of experiments in parallel to maximize the feedback we get. Note that these experiments are not like scientific experiments that attempt to determine and quantify governing relationships. In scientific experiments we need to be careful that we don’t change more than one variable at a time. In Adaptive Iteration this is much less of an issue, we are looking for the best design among a landscape of options, not the formula for the one right design.

What sort of experiments can we do? Initial small, low risk experiments include:

Thought experiments (see earlier blog post)
Sketches
Story boards
Prototypes (mock-up style)
Probes or ‘sighter’ trials

The objective of these initial experiments is to quickly canvas a broad range of design options and to learn more about the design context. The focus is on establishing and testing the broad design concept and structure. The objective is to develop one or more outline designs that are likely to satisfy the design objectives, boundaries and constraints. This also helps test early in the design process whether we need to question and review those objectives, boundaries and constraints. It is much more efficient to learn at this early stage that the stated design objective unnecessarily constrains the design.

A key aspect of these early low risk experiments is to make the design ideas as tangible as possible – to give a clear voice to the creative and conceptual thoughts of those involved in the design process. Forcing the early design ideas to become tangible has two key benefits. First, as noted above, it tests the whether the high level design criteria are appropriate and whether those criteria are clearly and commonly understood by all those involved. Second, it creates a common and tangible focus for evaluation and improvement of the design. It is much easier to identify and explain possible improvements if you are working with a representation of the design that is tangible and contains a practical level of detail than if you are working with a broad description. (Although thought experiments are not tangible, they should include sufficient descriptive detail to enable those involved to become immersed in the imagined situation. They must also be structured as a forward looking exploratory ‘experiment’, not as a rationalization or explanation of a desired outcome.)

One the characteristics, and risks, of these early adaptive iterations is that all four phases – ‘design’, ‘experiment’, ‘observe’ and ‘interpret’ – are often tightly coupled and performed by the one individual or group. This enables rapid early iterations and the early testing of a wide range of design options. However, it also runs the risk of ‘group think’ leading to premature narrowing of the design options. It is important that these early experiments are observed and interpreted with open minds and from multiple perspectives (see earlier blog post).

As Adaptive Iteration moves into detailed design, the experiments use more detailed and realistic representations of the design. The design is increasingly tested in the context in which it will used. The experimental options include:

Functional prototypes
Low risk trials
Simulations
Pilot implementation

These increasingly realistic experiments need to be structured to evaluate how the design performs in context. They need to reveal any unintended consequences, any unanticipated constraints, any emerging systemic behaviour (especially non-linear behaviour) and any lessons for wider implementation. At this stage it is also important to be open to any surprises – positive or negative. In summary, these experiments need to be sufficiently realistic and broad to assess what works and what does not. They also provide another opportunity to test the foundational design choices (objective, boundaries/ constraints, architecture and metaphor). The results of these more realistic experiments may indicate that some or all of the foundational design choices need to be fine tuned, or even substantially revised.

The final form of ‘experiment’ is the real implementation of the design. Although, it is not structured as an experiment, it is, by definition, the most realistic opportunity to observe what works and what does not. This is especially important for the situations of interest to us, those that are uncertain, unpredictable and emergent. In these situations, the introduction of the design is likely to have a ripple effect on the context (resistance to organizational change might increase – or decrease, competitors may bring forward the timing of a new product, reinforcing complementary initiatives may be stimulated in the industry, etc) . Consequently, it is important to continue the Adaptive Iteration approach throughout the life of the design, but especially during the early stages of introduction. So the final experimental option is:

Evolving implementation.

Filed Under: experiment

Composing an email – an example of foundational design choices

August 8, 2013 By Markus Leave a Comment

How does the concept of foundation design choices that I introduced in the previous blog post apply to a real design situation? To explore this let us look at it in a context familiar to most of us – the challenge of designing an important email. (If you do not think that writing an email is design, please keep reading with an open mind.) As a manager for many years, a significant percentage of my time (probably more than 20%) involved composing, reading and responding to emails. I believe the ability to compose (design) a clear, succinct and relevant email is a very underrated personal skill in most organizations.

In the previous post I proposed four foundational design choices:

Objective
Boundaries/constraints
Architecture/game plan
Metaphor.

Let us see how each of these could apply to composing an email.

Objective: Each email will (should) have a specific principal objective. It could be to persuade, to propose, to inform, to update, to report, to call to action, to instruct, to admonish, to praise, to entertain, to enlist, etc. The overall objective must be determined before we start composing the email so that the entire email supports that intent and so that the intent is absolutely clear to the reader. (I will leave aside here the question of whether an email is the best way to achieve this objective. That is a higher level design choice that I assume we have already taken.)

Boundaries/constraints: Any email is subject to a number of practical and contextual constraints. They include the time available to write the email, the likely time and attention that reader will give to the email (both these constraints will influence the length of email), the confidentially of any information in the email, the level of familiarity of the readers with relevant technical terms and acronyms, the existing level of familiarity of the reader with the subject matter and context, scope of subject matter, etc. Some of the these boundaries/constraints, such as the length and scope of the email, are within your control as a writer. Others, such as level of familiarity with technical terms and acronyms, are characteristics of the readers. However, the extent to which the characteristics of the readers is taken into account is still a choice of the writer (designer), that is, it is still a foundational design choice.

Architecture/game plan: The style of the email essentially defines its structure (architecture). Possible styles include formal report style, journalistic style, ‘call to action’, conversational, and narrative. Exactly how these styles are implemented will often depend on context. For example, how you ‘call to action’ will depend on factors such as whether the action is mandatory or discretionary, your level of authority or influence, the complexity of the action, and the level of reader self-interest in undertaking the action.

Metaphor: If the email is particularly complex and/or important, it may be valuable to identify a metaphor for the email before composing the detail. If the email is instructional, a road map metaphor may be appropriate. If the purpose of the email is to admonish or correct, picturing yourself as a coach or parent may help set a firm but constructive tone (although this may be better done face-to-face rather than use email). If you are calling to action to address a very specific challenge, then picturing the email as a arrow may help create the required focus and emotion.

Clearly, not all emails require the effort to develop foundational design choices. Many of our emails are routine responses to requests for information or routine elements of the business processes of our organizations. The design for these emails are essentially determined by the business context and/or routine (best?) practice in our organization. However, foundational design choices are much more relevant and significant when the email is part of situation that is uncertain, unpredictable and emergent. This is often the case when we are seeking to use the email to persuade and influence in the context of organizational or business change.

Filed Under: design, examples

Design involves choices – not all are equal

August 6, 2013 By Markus Leave a Comment

Whenever we create something to achieve a purpose or objective, we are designing. In the context of Adaptive Iteration we are interested in designs that need to achieve their objectives in situations that are uncertain, unpredictable and emergent. In these situations the scope and complexity of the interactions are such that we cannot predict all the cause and effect relationships in advance. Not only are they complex, but they are also changing dynamically and are likely to change further in response to the introduction of our design.

The types of situations that require this type of emergent design include business plans and strategies, organizational change initiatives, new product development, new market entry, career development and organizational development. In other words, almost any situation that is forward looking and involves interactions of people or groups of people whose decisions will influence the outcome.

Designing involves choices. There is no one right design. The design choices will determine the extent to which the design meet its objectives. If the design context is uncertain, unpredictable and emergent, the best design choices are not at all clear. For example, we cannot be certain how customers or competitors will react to a new set of product features. Can we be confident that our new social media strategy will attract sufficient attention from our target markets? It may not be clear what coaching and training to give to our middle-managers to ensure that they can make the transition to our new project-based organizational structure?

How do we make these choices? If we use the Adaptive Iteration approach we do not have to be ‘right first time’. With Adaptive Iteration, we form hypotheses about the best design choices and then conduct low risk experiments to assess how those choices work in practice. We use the resultant observations and interpretations to refine the design choices (hypotheses) that we again test experimentally, and so on. By starting small with low risk and iterating rapidly, we home in on the best design to meet our objectives in the particular context.

However, in Adaptive Iteration, our initial design choices (hypotheses) are not random. Foundational design choices set the relatively stable core of the design on which choices about the design detail are made and adaptively iterated. The foundational design choices are not unchangeable during Adaptive Iteration but will change much less frequently than choices about the design details. A significant shift in any of the foundational design choices should stimulate a major rethink of the overall design. The foundational choices, in general order of significance are:

Objective: The objective describes the intent of the design, and of the associated Adaptive Iteration, and forms the key anchor for all other design choices. All other design choices must enable the design to achieve its objective. The clearer the design objective, the more likely that Adaptive Iteration will proceed efficiently and effectively. It is important to select an objective that does not implicitly constrain the design options. In my blog post that summarized the start up stories of HP, Sony and Microsoft we saw that the objective of each of the founders was to create a profitable business based on their particular expertise and interests, rather than create a business in any particular product/market segment. The founders of each company adaptively iterated the product/market segment until they found one that met their design objective.

Boundaries/Constraints: The boundaries and constraints can apply to any aspect of the design. They set the scope of design variables (choices) available to the designers. They also set any limits to or targets for the performance of the design, and they set the context within which the design sits. In some cases, tight design constraints can stimulate focus and can help challenge the status quo. For example, a design constraint that requires that the manufactured cost of a product design be 20% less than its predecessor is likely stimulate new a search for new perspectives and step-change ideas. In other cases, widening the design boundaries may create the scope to develop more systemic and more fundamental design improvements. Roger Martin provides a good example of this in his book Opposable Mind where he describes how IDEO recognized that its design brief from Amtrak (redesign their railcar) was inherently limiting and convinced Amtrak that, to achieve its design objectives, it needed to look at the end-to-end train ride experience.

Architecture/Game plan: The architecture or framework of a design will usually have a significant impact on our ability to adaptively iterate the design. Key decisions include how much operating flexibility and how much modularity to build into the design. For example, an organizational design that is based on a strong and commonly held set of operating values and principles will empower employees to use their discretion to meet customer needs as long as what they do is consistent with those values and principles. On the other hand, employees in an organization that emphasizes rules and procedures will be constrained to only the situations and responses anticipated by those rules and procedures. The nature and extent of modularity will influence the ability to adaptively iterate the design. Coarse-grained modularity will constrain design changes to large steps (changing one or more of the coarse grain modules), whereas fine-grained modularity provides much more flexibility but may lead to costly customization and integration of the design after each adaptive iteration. Ideally we want to align coarse-grained modules with any relative stable aspects of the context and use finer-grained modules for those aspects of the context that are uncertain, unpredictable and emergent.

Metaphor: Use of a design metaphor can help unify a design by stimulating coherence for the myriad of detailed design choices. It can help create a distinct personality for the design and stimulate a positive emotional response from the user. A metaphor can also help identify the mix of design features that will create both functional and emotional coherence for the design. It can help overcome the tendency to load a design up with all manner of features and capabilities “just because we can”. For example, a metaphor may assist the redesign of an under performing group or small organization. Depending on the situation and objectives, the group could use the metaphor of a sporting team, an orchestra, a jazz ensemble, an operating theatre, improvisation theatre, etc.

Filed Under: design