The Structure of Unstructured Decisions

Pure curiosity and a desire to improve the decision processes within my own company, twenty--twenty. I was going through the Amazon shareholder letters to understand the 'Amazonian thoughts' of Jeff Besoz. In the 2005 Shareholder letter there's a footnote that caught my eye coming from this paragraph:

As you would expect, however, not all of our important decisions can be made in this enviable, math-based way. Sometimes we have little or no historical data to guide us and proactive experimentation is impossible, impractical, or tantamount to a decision to proceed. Though data, analysis, and math play a role, the prime ingredient in these decisions is judgment. [Footnote 1]

[Footnote 1]

“The Structure of ‘Unstructured’ Decision Processes” is a fascinating 1976 paper by Henry Mintzberg, Duru Raisinghani, and Andre Theoret. They look at how institutions make strategic, “unstructured” decisions as opposed to more quantifiable “operating” decisions. Among other gems you will find in the paper is this: “Excessive attention by management scientists to operating decisions may well cause organizations to pursue inappropriate courses of action more efficiently.” They are not debating the importance of rigorous and quantitative analysis, but only noting that it gets a lopsided amount of study and attention, probably because of the very fact that it is more quantifiable. The whole paper is available at www.amazon.com/ir/mintzberg.

So, I gave the paper a read and also found it fascinating. It was a new topic area for me so I decided to set the ideas into my mind by reading and consolidating the paper into a language I understand better.

In this summary, I try to reduce the number of words and the reading difficulty where I could. I summarise the key points and at the end I put myself through an exercise of thinking how one could change an organisation to help it make better decisions.

Let me know what you think!

2005 Amazon Shareholder's Letter

The Structure of ‘Unstructured’ Decision Processes

This paper forms a model which has 3 phases. You can break each phase into routines. Phase 1 is identification, broken down into routines of decision recognition and diagnosis. Phase 2 is development, which have routines of search and design. The final phase, selection, is broken down into screen, evaluate-choice and authorise routines. There are some routines which span across all the phases, such as communication, decision control and politics. Though this is explained in a linear fashion, it's more of a cyclical process, with its 'length' determined by dynamic factors such as: interrupts, delays, natural cycles of comprehension and even failure.

General

Limitations

Next

You could quite neatly put empirical literature on this topic (to date, i.e. 1973) into three groups:

1 - individual decision making in 'lab' games by cognitive psychologists

2 - group decision making by social psychologists

3 - real organisational decision making

This paper's posed structure is 'phased', which takes inspiration from Dewey (1910) who said there are 5 phases of reflective thought:

Future research continues, with phases ranging from 3 to 8. The most well know is probably Simon's intelligence-design-choice trichotomy. [link]

In comes Witte (72) who challenges 'phase theorem' by saying that there's no such thing as distinct, sequential phases. He found in his research that the decision processes of data processing equipment did relate to some of Simon's phases but:

Witte also finds:

Mintzberg and friends agree with Witte. Phases make sense, but it doesn't have to be sequential. Here's how they name things. 3 phases at the top level: identification, development and selection. Phases are broken down into routines. Alongside, there are dynamic factors. Together this makes up the 12 basic elements of the strategic decision process.

Decisions in this paper were labelled in 3 ways:

1) the stimuli that evoked them - a spectrum from (voluntary) opportunity decisions to (involuntary) crisis decisions, with problem decisions falling in the middle. Problems can move up and down the spectrum: delays, temporary solutions, converting crisis into opportunity. Note that the problem definition depends on perspective, i.e. a crisis for one may be an opportunity for another.

2) their solutions - 4 ways:

3) process for arriving at solutions - the whole point of the paper, read on!

Important decisions aren't obvious. Through the heaps of verbal communication, if there's a gap between the actual and expected situation, decisions need to be made! Pound (69) says an expected situation is formed from:

This routine is where we recognise opportunity from problem from crisis decisions.

Moment of action is when cumulative amplitude of stimuli > action threshold

Determinants of stimuli amplitude

Also:

Threshold of action changes a lot and depends on:

Okay - we cross the action threshold! The decision maker has some partially ordered data and an interesting situation.

The first step is often tapping and creating new information channels to clear up and define the issues.

Interestingly, diagnosis doesn't have to be formal explicit activity.

Explicit diagnosis emerged most in cases in the middle of problem range continuum but not really at the extremes (crisis / opportunity).

2 ideas: opportunites don't really need diagonising (it's an activity of creativity); and in crisis mode there is no time or energy to do formal diagnosis.

Finally, Drucker noted that what makes the Japanese managers different from the American ones was that they pay careful attention to the daignosis routine.

The greatest amount of resources are consumed here.

For ready made solutions, using divergent thinking. There are 4 search behaviours:

The search routine is hierarchical and stepwise. An individual would start with memory and trap search. Here it's cheap to consider more alternatives so many arise. You may get a top pick and a secondary to compare (called a "confirmation candidate").

If nothing is adequte, they'll venture out to people and information sources further away - from familiar to unfamiliar sources. Repeated failure would lead to choosing a Design routine, which would yield custom-made solutions.

To develop or modify custom-made solutions (convergent thinking) through cycles of iteration, ultimately leading to a single solution.

As it's expensive and time-consuming you can't afford to create more than one.

Not always the conclusive phase. Decisions can be made of sub-decisions, and each sub-decision has a selection phase.

Mintzberg prefer these three routines: screen, evaluation-choice and authorisation.

Selection phase is typically multi-stage, iterative and investigations that get deeper.

If your search routine (in the dev phase) yields many potential results, then you'll screen so that the next decision makers save time. It's more a task of removing those are that un-realistic rather picking those that are realistic. Search and screen can happen together.

This is, by far, where academics have focussed the most. But ironically, it seems less significant than the diagnosis and design routines.

This routine can happen in one of three modes:

Findings:

Other studies' findings:

When selecting strategic solutions there are a great number of factors and they're often "soft" (such as emotions, politics, power, personality), so the evaluation-choice routine is actually quite crude. There are also dynamic factors and uncertainty. Finally there are also cognitive limitations & biases, intended or not. These factors are also 'scaled' / 'weighted' individually and are independent of each other. To cope, decision-makers use proxies of choice, e.g. imitating or following tradition (Pfiffner, 60).

To make an evaluation, one may use a method of first measuring against a primary goal (never more than three) and then next constraints. It's also important whether the individual is optimising or satisficing (Soelberg, 67). Hopefully we find an option which hits our goals, within our constraints. If there multiple in the 'active roster' then often a crude scale is used, such as much better vs a little better.

This is a yes/no answer to a whole solution and happens when the individual making the choice lacks the power to commit a course of action. It may have to pass through multiple tiers, and may get blocked. Typically you'll ask for authorisation towards the end of the decision process, where you reach a confident solution, but sometimes auth is asked from the outset or during development.

Acceptance = commiting resources, or presenting the solution to the next level.

Rejection = abandonment / redevelopment

Authorisors will often lack the in-depth knowledge that the developers have, yet they manage budgets. Hmm the developer wants it to be approved, the authorisor doesn't want to seem ignorant... capital budgeting is more a distorted political process and less analytical than you think (Carter 71a and 71b, Pettigrew 72, Bower 70)

These routines support the three phases:

This is decision making about decision making - what are the steps to a solution but also, what's our approach, what resources do we need?

FIrst part is decision planning:

Then, the decision maker would jump to a routine - do we diagnose the problem? Or search for solutions?

Three sub-routines:

Political activities reflect the influence of individuals who seek their personal and institutional needs by the organisational environment. They may be inside or outside the org, because what ties them in, is that a decision outcome involves them.

Political activities clarify power relationships, but also:

This supporting routine generally appears during the bargaining routine (in the evaluation-choice phase).

Can happen:

Levellers:

The previous describes process, phases and routines. Pretty standard. What makes things strategic are these factors which change the timeline.

Six groups:

These cause changes in pace or direction. Due to:

Solutions

Often interrupts beget interrupts. More frequent in high pressure environments.

Managers are time constrained. Without big blocks of time, they'll break complex decisions into chunks, which can introduces time delay.

It just takes time to get the results/ reactions of a previous step. Some activities need some incubation period before insight arrives.

Not written up much, but realise that managers control pace to:

Smooth execution doesn't mean do things as fast as possible.

Remember the theme of this paper. Decision processes aren't an assembly line, rather more like a cycle, where we often jump back to earlier phases. You can cycle within a single routine, or between multiple to gain further insight into complex decisions.

It may not be appropriate to continue a process, or to implement a solution. The payoff may be too low. They can't stay within constraints. Maybe it just doesn't appeal to the authority. So, we delay, wait or go back to development with some different constraints or goals.

That's all the components, now we develop a model.

A bit overwhelming huh? Here's the gist of it:

The paper then goes through different decision processes, explained by drawing a timeline on the general model.

Figure 2 - the decision process of passing a policy of mandatory retirement at age 65

The line shows what happened:

A more difficult example: acceptance of a new treatment in a hospital.

To flesh out the diagram:

Features of this:

The authors continue, and go over three more examples. I'll save you the time and summarise:

A regional airline wanted to acquire new jet aircraft. This decision was interupted by a new CEO. They then dived deep into an evaluation-choice routine, but ultimately when another airline went into administration the CEO quickly took the opportunity to buy up these craft at a good price. Multiple routines needed authorisation on three successive levels.

In the cases which needed the most design activity (often new product development or marketing), they were driven by opportunity, and the new profit motive outweighed political power.

In some cases, a basic design process were jolted into a dynamic one.

Conclusion

Strategic decision processes are complex and dynamic, but we can still put it on a diagram (which structures the concepts). But they've barely scratched the surface.

They haven't dived into the routines. Nor looked into how multiple decisions inter-relate (formulating a strategy). What about comparing decisions at the top versus at the bottom.

What could a company do to improve each element of the decision process?

The greatest amount of resources are consumed here.

For ready made solutions, using divergent thinking. There are 4 search behaviours:

The search routine is hierarchical and stepwise. An individual would start with memory and trap search. Here it's cheap to consider more alternatives so many arise. You may get a top pick and a secondary to compare (called a "confirmation candidate").

If nothing is adequte, they'll venture out to people and information sources further away - from familiar to unfamiliar sources. Repeated failure would lead to choosing a Design routine, which would yield custom-made solutions.

To develop or modify custom-made solutions (convergent thinking) through cycles of iteration, ultimately leading to a single solution.

As it's expensive and time-consuming you can't afford to create more than one.

Not always the conclusive phase. Decisions can be made of sub-decisions, and each sub-decision has a selection phase.

Mintzberg prefer these three routines: screen, evaluation-choice and authorisation.

Selection phase is typically multi-stage, iterative and investigations that get deeper.

If your search routine (in the dev phase) yields many potential results, then you'll screen so that the next decision makers save time. It's more a task of removing those are that un-realistic rather picking those that are realistic. Search and screen can happen together.

This is, by far, where academics have focussed the most. But ironically, it seems less significant than the diagnosis and design routines.

This routine can happen in one of three modes:

Findings:

Other studies' findings:

When selecting strategic solutions there are a great number of factors and they're often "soft" (such as emotions, politics, power, personality), so the evaluation-choice routine is actually quite crude. There are also dynamic factors and uncertainty. Finally there are also cognitive limitations & biases, intended or not. These factors are also 'scaled' / 'weighted' individually and are independent of each other. To cope, decision-makers use proxies of choice, e.g. imitating or following tradition (Pfiffner, 60).

To make an evaluation, one may use a method of first measuring against a primary goal (never more than three) and then next constraints. It's also important whether the individual is optimising or satisficing (Soelberg, 67). Hopefully we find an option which hits our goals, within our constraints. If there multiple in the 'active roster' then often a crude scale is used, such as much better vs a little better.

This is a yes/no answer to a whole solution and happens when the individual making the choice lacks the power to commit a course of action. It may have to pass through multiple tiers, and may get blocked. Typically you'll ask for authorisation towards the end of the decision process, where you reach a confident solution, but sometimes auth is asked from the outset or during development.

Acceptance = commiting resources, or presenting the solution to the next level.

Rejection = abandonment / redevelopment

Authorisors will often lack the in-depth knowledge that the developers have, yet they manage budgets. Hmm the developer wants it to be approved, the authorisor doesn't want to seem ignorant... capital budgeting is more a distorted political process and less analytical than you think (Carter 71a and 71b, Pettigrew 72, Bower 70)

These routines support the three phases:

This is decision making about decision making - what are the steps to a solution but also, what's our approach, what resources do we need?

FIrst part is decision planning:

Then, the decision maker would jump to a routine - do we diagnose the problem? Or search for solutions?

Three sub-routines:

Political activities reflect the influence of individuals who seek their personal and institutional needs by the organisational environment. They may be inside or outside the org, because what ties them in, is that a decision outcome involves them.

Political activities clarify power relationships, but also:

This supporting routine generally appears during the bargaining routine (in the evaluation-choice phase).

Can happen:

Levellers: