Methodologies

From the documentation of the field

Most organisations already use methods that sound close to the six primitives of Ismail's Primitives. This page relates ten familiar ones to them. The relationship is one of mechanism and requirement. A primitive is a structural requirement: a function that a decision structure must perform when it faces the matching situation. A methodology is a candidate mechanism: a way of performing a function, which may or may not perform it in a given organisation.

The mappings below are the field's reading. They are arguments, not theorems, and sit in the domain readings tier described on the verification tiers page. Each method gets a real entry for what it can contribute and an honest one for what it cannot supply alone.

How to read the entries

An assessment asks whether a structure possesses a mechanism that performs the function, not whether it owns an artefact that resembles one. The kind of verdict says the same. So a method that appears in an organisation's practice is a lead to investigate. It is not evidence of possession, because a template that is filled in once and filed performs no function. Error theory calls accepting such an artefact as the mechanism one of the four ways an assessment goes wrong.

The capabilities, in short, are X₁ telling apart live hypotheses that call for different actions, X₂ staying away from actions that lock in permanent loss, X₃ continuing to explore past a comfortable plateau, X₄ committing once the evidence supports one option, X₅ treating a hard constraint as a wall, and X₆ discarding evidence that went stale when the world changed.

Summary

Read together, the table records an observation. No method on the list, used on its own, supplies all six primitives of Ismail's Primitives. That says something about how each method is designed and nothing about its worth.

SWOT

SWOT can contribute to X₃, X₅ and X₆. The opportunities column is a place to name positions beyond the comfortable current one, which serves X₃. The threats column can surface hard limits, which serves X₅ when a limit is then treated as a wall and not as a weighting. A SWOT revisited when conditions change supports X₆.

It contributes when its entries change what the organisation does and are reviewed when the environment shifts. On its own it supplies no way to tell two competing explanations apart, which X₁ needs, and nothing in it forces a commitment, which X₄ needs. It is a snapshot, so a SWOT that is completed once does not perform X₆.

OODA

OODA, the observe, orient, decide and act loop, can contribute to X₄ and X₆. The decide and act steps turn a judgement into a commitment, and a fast loop shortens the time between the world changing and the structure responding.

It contributes when the orient step discards models that no longer fit, not when it merely reorganises the same picture. It supplies no mechanism for X₁, because it does not say how competing hypotheses are weighed against evidence. It does not identify irreversible actions for X₂, and it does not enforce hard limits for X₅. A fast loop with a stale model acts quickly on the wrong picture.

MECE

MECE, mutually exclusive and collectively exhaustive, can contribute to X₁. A partition of the live hypotheses into non-overlapping and complete cases makes it possible for evidence to discriminate between them, and an incomplete partition hides the explanation that is true.

It contributes when the partition is of hypotheses that call for different actions and is then tested against evidence. It is a property of how a problem is represented, so it says nothing about dynamics, uncertainty or what to do. A clean partition with no evidence-gathering tracks nothing. It also cannot detect that the partition itself has become out of date.

Stage gates

Stage gates can contribute to X₂, X₄ and X₅. Releasing resources in stages keeps an irreversible commitment from being taken before evidence supports it, which serves X₂. A gate is a point at which a decision is made and held, which serves X₄. Gate criteria that are binding serve X₅.

They contribute when a gate can stop the work and when criteria are written to be met or not met, instead of being negotiated at each review. They supply less for X₃ and X₆. A gate sequence can lock in the plan it was set up to protect, and criteria fixed at the start may not reflect a world that has since changed. A gate that always passes is an artefact, not a mechanism.

Pre-mortems

Pre-mortems can contribute to X₁ and X₂. Imagining that a plan has failed and asking why generates alternative explanations of what could go wrong, which serves X₁. Asking which failure routes could not be undone serves X₂.

A pre-mortem contributes when its output changes the plan and when the irreversible routes it finds are acted on. It is usually a one-time exercise before launch. That means it does not provide continuing detection after the world moves, it commits to nothing, and by itself it does not adapt anything.

RAID logs

RAID logs, which record risks, assumptions, issues and dependencies, can contribute to X₂, X₅ and X₆. Flagging risks with their reversibility serves X₂. Recording dependencies and limits serves X₅. An assumptions register can serve X₆ if an assumption that stops holding triggers a review of the decisions that depended on it.

They contribute when each assumption has an owner, a test and a trigger, and when the log drives decisions. A log records and does not decide. A RAID log with no triggers, however thorough, performs none of these functions, and it supplies nothing for X₃ or X₄.

Monte Carlo

Monte Carlo simulation can contribute to X₁, X₂ and X₅. Running outcomes under competing models shows how far the hypotheses differ in what they predict, which serves X₁. Estimating the probability of ruin, if the ruin state is in the model, serves X₂. Estimating the chance of breaching a limit serves X₅.

It contributes when the model includes the trap or the limit it is meant to test and when the results are used to change a decision. Its output is only as good as the model, and a model that omits a trap shows no trap. It cannot detect that the process generating the data has changed, which X₆ needs, and it commits to nothing.

Scenario planning

Scenario planning can contribute to X₁, X₃ and X₆. Holding several coherent pictures of the future at once keeps competing hypotheses alive, which serves X₁. Exploring futures beyond the current position serves X₃. Indicators that signal a regime shift serve X₆.

It contributes when the scenarios differ in ways that call for different actions, and when named indicators are monitored and tied to responses. It supplies no way of choosing between scenarios and no commitment, so X₄ sits outside it. A set of scenarios anchored on today's view can be a comfortable range that is not really different from the current plan.

Six Sigma

Six Sigma can contribute to X₁, X₄ and X₆. Measurement and root-cause analysis test competing explanations of a process failure, which serves X₁. Standardising a practice the evidence has settled serves X₄. The control phase, which keeps watching a process after the change, serves X₆.

It contributes when the root-cause step really tests alternatives and when the control step triggers a revisit if the process drifts. It optimises within an existing process, so it does not itself push beyond a comfortable plateau, which X₃ needs. It also tends to assume the process is stationary, and it can entrench a local optimum.

Agile

Agile can contribute to X₁, X₂, X₃ and X₆. Short iterations test hypotheses against real results, which serves X₁. Small, reversible steps limit exposure to actions that cannot be undone, which serves X₂. Experiments explore beyond the current approach, which serves X₃. Frequent feedback serves X₆.

It contributes when increments really test an assumption, when the team will stop or reverse a direction on the evidence, and when each step is in fact reversible. Left alone it supplies little for X₄: endless iteration can keep a settled question open indefinitely. It also does not by itself enforce hard constraints such as compliance, which X₅ needs.

What the entries share

No method on this page supplies all six capabilities, and every one of them can be run in a way that performs none. That is why the field assesses the mechanism and not the method's name. It is also why the field does not rank these methods. A method can be useful and still leave a primitive of Ismail's Primitives unmet.