Decisionmaking stress test
A practical test of whether the machinery behind consequential decisions actually works together.
Decision quality is a systems problem
Operationally complex organizations can have all the ingredients associated with strong decisionmaking and still make consequential decisions suboptimally.
A complete set of parts for a 747 is not a functioning aircraft.
The parts must be assembled into a system. The system must be operated. Its performance depends on how well the system is designed and run.
Decisionmaking is no different.
Evidence, expertise, alternatives, tradeoffs, uncertainty, authority, and feedback do not self-assemble into an optimized decisionmaking process.
They must be connected through mechanisms that cause the right information to enter, the right alternatives to remain open, the right conflicts to be resolved, and new evidence to change the commitment when warranted.
That machinery can be examined.
The machinery of optimized decisionmaking
The standard is not whether the ingredients exist. The standard is whether mechanisms cause them to interact while they can still change the decision.
- 01
Relevant reality reaches the decision
Material information must surface while it can still change the commitment, including information that would not otherwise reach the decision. Information that arrives too late, remains isolated, or cannot affect the recommendation is functionally absent from the decision.
- 02
The right problem enters consideration before momentum forms around an answer
A rigorous process can still optimize the wrong problem. Decision quality depends partly on whether the right problems and opportunities enter consideration before attention, ownership, or momentum narrows the field.
- 03
The opportunity set reflects current reality
The available choices should reflect what is actually feasible, valuable, and constrained now. The opportunity set should neither exclude viable options nor privilege them because they are familiar, fashionable, already funded, or institutionally convenient.
- 04
Materially different alternatives remain alive
A real comparison includes materially different courses of action. Delay, partial commitment, experimentation, a different sequence, preservation of an option, or no action may all belong in the set. A detailed comparison among variants of one predetermined answer is still a narrow decision.
- 05
Facts, assumptions, and unknowns remain distinct
Consequential decisions combine observed facts, estimates, forecasts, assumptions, and unresolved uncertainty. Those categories have different evidentiary weight. Precision does not turn a forecast into evidence. Repetition does not turn an assumption into fact. Approval does not eliminate uncertainty.
- 06
Material uncertainty changes the decision
Not every unknown deserves attention. The relevant uncertainty is uncertainty capable of changing the recommendation, the timing or degree of commitment, the value of learning more, or the need to preserve reversibility.
- 07
Causal logic is explicit and testable
Every recommendation rests on some account of how an action is expected to produce an outcome. That logic must be visible enough to challenge before commitment and test against evidence afterward.
- 08
Relevant considerations are integrated against the enterprise objective
Representation is not integration. Different considerations can all be valid and still point toward different choices. The process must resolve those conflicts against the objective governing the decision. If conflicting inputs never alter one another, they have been collected, not integrated.
- 09
Competing objectives are resolved above the level that owns the initiative
The objective of the part is not automatically the objective of the whole. A process optimized for enterprise value cannot allow the function sponsoring or owning the initiative to define the tradeoff by default.
- 10
Opportunity cost is visible
A positive return does not establish the best use of scarce resources. Every commitment consumes resources that could be used elsewhere. The relevant questions are not only what the commitment produces, but what it displaces.
- 11
The value of additional information affects action
Some uncertainty is worth resolving before commitment. Some is not. Information has decision value only when it can arrive in time to alter the decision and is valuable enough to justify obtaining it.
- 12
Dependencies, sequence, reversibility, and optionality shape commitment
The best initiative is not always the best next initiative. One commitment may enable another, constrain another, consume a scarce capability, create lock-in, generate information, or preserve future choice. Today’s commitment changes tomorrow’s option set.
- 13
Existing mechanisms do not predetermine the answer
Ownership, vendors, platforms, budgets, and institutional habits all constrain what appears natural. The decision process must prevent those constraints from defining the opportunity set before the objective has been resolved.
- 14
Decision rights are explicit
The process must establish who can decide, challenge, escalate, reopen, stop, defer, accelerate, or reverse a commitment. Authority discovered only when disagreement appears is not clear authority.
- 15
Evidence that would change course is defined in advance
A commitment should specify what future evidence would justify changing it. Otherwise evidence is judged against a moving standard shaped by sponsorship, sunk cost, and hindsight.
- 16
Execution evidence can change the decision
Execution produces information that did not exist at commitment. That information must be capable of changing priorities, scope, sequence, architecture, resource allocation, timing, or the commitment itself. Feedback that cannot change action is observation.
- 17
Commitments remain genuinely revisable
Formal authority to stop, defer, reverse, accelerate, or reallocate is insufficient if sponsorship, sunk cost, politics, or inertia make those actions impractical. The ability to revise a commitment has to survive the forces that make revision difficult.
- 18
The machinery survives changes in people
A mechanism that operates only because one unusually capable person remembers to invoke it is not yet a system capability. The process must carry the mechanism.
Having the parts does not mean having the system
The presence of every required capability does not establish a functioning decision system.
The parts must interact.
New evidence changes assumptions or alternatives.
Uncertainty changes commitment.
Opportunity cost changes allocation.
Dependencies change sequence.
Execution evidence reopens decisions when warranted.
Those interactions are the system.
A 747 does not become an aircraft because every component is present. The components must be assembled in the right relationships and operated as one machine.
Decisionmaking is no different.Decision process has economic leverage
Consequential decisions allocate scarce resources and alter the conditions under which later decisions are made. An investment can create capability, consume capacity, establish a dependency, generate information, increase switching costs, eliminate an option, or create several new ones.
Affordable can still be expensive.
A commitment can fit comfortably within budget while consuming a resource whose alternative use would create more value.
Positive ROI does not prove best use.
An initiative can create attractive returns and still rank below another use of the same constrained resources.
The best initiative is not always the best next initiative.
Sequence changes value.
The cost of being wrong is partly designable.
Staging, reversibility, experimentation, and preserved optionality can change the economic consequences of an incorrect decision.
Information loses decision value after the decision can no longer change.
Learning that arrives too late may improve explanation without improving the commitment.
Confidence requires evidence
The machinery should be demonstrable. For every material mechanism, the evidence falls into three different states.
Can point to it
Evidence shows that the mechanism operated.
Establishes process operationCould reconstruct it
The reasoning can probably be recreated afterward, but there is no contemporaneous evidence that the mechanism governed the decision.
Would be guessing
There is no reliable basis for saying the mechanism operated.
Only the first establishes that the mechanism was part of the process.
Evidence must also exist at the point when the mechanism could affect the decision. A rationale constructed after the outcome is known cannot establish what governed the original commitment.
What materially different alternatives were actually alive?
Which assumptions were explicit?
Which uncertainties changed commitment?
What information changed the recommendation?
Which opportunity costs changed allocation?
Which dependencies changed sequence?
Which conflicts were resolved, and against what objective?
What evidence was defined in advance as sufficient to change course?
What happened when that evidence appeared?
If those questions cannot be answered with contemporaneous evidence, confidence in the process exceeds the evidence supporting it.
A good decision does not prove a good system
A strong decision can emerge from weak machinery when necessary questions, evidence, challenges, or dependencies enter the process through individual initiative or circumstance rather than system design.
What the result does not establish is that the same mechanisms will operate on the next decision.
A decision system reduces dependence on the particular combination of attention, memory, influence, judgment, and circumstance surrounding one commitment.
The mechanisms reside in the process.
Consequential decisions recur. The leverage increases with the value, frequency, irreversibility, and downstream consequences of the decisions governed by the system.
Stress-test the machinery
The same standard can be applied at three levels.
One consequential decision
A recurring class of decisions
The standing enterprise decision system
The test is whether the machinery required to optimize the decisionmaking process can be shown to exist, interact, and operate.
If it can, the confidence is earned.
If it cannot, the confidence is an assumption.
