Chapter 5: Decision Processes
The decision that moved rooms
Renee is part of a cross-functional group choosing a customer-support platform. The group has published criteria, tested three systems, and is preparing a recommendation. Two days before the final meeting, the executive sponsor cancels it. She says the purchase will instead be discussed during the senior technology steering meeting because “we need to move quickly.”
Only two members of Renee’s group belong to that meeting. The steering committee receives a four-slide summary, not the test results. It selects the vendor favored by the infrastructure director, citing “enterprise fit,” a phrase that did not appear in the original criteria. The minutes say the steering committee “confirmed the project team’s direction.”
The chosen system may be excellent. The infrastructure director may know risks the project team missed. The steering committee may hold the proper financial authority. Yet the process leaves several questions. Why did the venue change? What happened to the recommendation? Who introduced the new criterion? Did the committee choose a different option or confirm the same one? Can anyone reconstruct the basis?
Decision quality and procedural fairness are related but not identical. A fair process can produce an unpopular result. An unfair process can stumble into a good one. Research on organizational justice treats procedures, outcomes, explanations, and interpersonal treatment as distinct dimensions that affect how people experience work.1 Research on voice also suggests that having a genuine opportunity to provide input before a decision can matter even when the person does not control the outcome.2 Reviews of group decision-making and information sharing further show why membership and usable information deserve attention rather than being treated as administrative details.3
This chapter is not a formula for consensus. Organizations need people with authority to choose. It is a guide to making that authority visible, informed, and reviewable.
Six parts of a legible decision
A consequential decision becomes easier to trust when six parts are clear:
- Question: What is being decided now?
- Authority: Who recommends, who advises, and who decides?
- Participation: Whose knowledge or interests must be considered?
- Criteria: What standards will distinguish the options?
- Record: What was chosen, on what basis, and with what dissent or conditions?
- Review: What could reopen, appeal, or evaluate the decision?
Not every choice needs a committee or lengthy file. A supervisor assigning today’s work can answer these parts in a sentence. A major procurement, promotion, site closure, safety exception, or policy change requires more. The control should match the consequence, reversibility, and risk.
1. Choosing a venue that changes who decides
Venue is more than location. It includes the meeting, committee, workflow, jurisdiction, platform, or informal conversation in which an issue is handled. Moving a decision can change membership, rules, records, expertise, appeal rights, and the range of possible outcomes.
Venue changes can be legitimate. A project team may lack purchasing authority. A complaint may need an independent route. A local issue may cross an enterprise standard. An emergency body may act when the ordinary committee cannot meet. The essential question is whether the new venue’s authority and purpose are explained.
In Renee’s case, moving the purchase to the technology steering committee may be correct. The misleading part is describing its choice as confirmation if the project group recommended something else. A clean process would say: the project group evaluated usability and service requirements; the steering committee held final authority; the committee introduced enterprise fit as an additional constraint; and it chose a different option for stated reasons.
When a venue changes, ask:
- Which mandate gives the new forum authority?
- What participation or review is lost or gained?
- Will the evidence and prior recommendation travel with the issue?
- Does the new forum make the same decision or a different layer of decision?
- What record will connect the two?
Individuals should avoid arguing that the original forum “owns” an outcome simply because it invested time. Ask that its work be represented accurately and that the decision chain be recorded. Managers should prevent venue shopping: repeatedly moving an issue until it reaches a sympathetic audience. A governance map showing normal forums and escalation conditions is often enough to expose the pattern.
2. Shaping committee membership or quorum
Membership determines which expertise, interests, and votes enter a decision. Quorum rules determine when the body can act. Both are necessary. Both can be manipulated or simply neglected.
Not every absence invalidates a decision. People are ill, conflicts require recusal, vacancies occur, and emergencies cannot always wait. Examine the written rule, the reason for the change, the effect on representation, and the availability of a correction.
Suppose a hiring panel normally includes the role’s manager, a peer, an independent manager, and an HR adviser. The independent member withdraws because they know a candidate. The chair proceeds with three people because schedules are tight. The policy permits it. The process may be valid, but the safeguard provided by independent input has disappeared. A replacement or recorded rationale would better preserve the design.
Late additions deserve the same clarity. Was the new member appointed, advising, observing, or voting? Did they review the same evidence? Can a person who joins after interviews score candidates? The answer should come from the process, not from the status of the newcomer.
Quorum is a floor, not always a fairness test. A committee may technically reach quorum while an affected function, region, or worker representative is repeatedly absent. Managers should review whether the membership rule still serves the decision’s purpose. If representation is essential, make it part of quorum or require an explicit exception.
Do not infer intent from a calendar alone. Build a timeline of invitations, recusals, substitutions, attendance, and authority. If a membership change alters the outcome, focus first on whether the decision should be paused, reheard, or ratified under the proper composition.
3. Treating a decision as settled before review
Precommitment occurs when actions taken before formal review make one result practically difficult to change. The sponsor tells a customer the launch date, a manager announces the preferred candidate, a team begins implementation, or a slide labels one option “the plan” and the others “alternatives.”
Some precommitment is unavoidable. Decision processes often overlap with preparation. A pilot requires resources; a negotiator needs a mandate; a preferred bidder may need due diligence. The relevant distinction is between reversible exploration and commitment that creates cost, expectation, or reputational pressure.
A useful decision paper identifies status explicitly:
- Exploring: no organizational commitment; evidence gathering only.
- Preferred: current recommendation; open to change under stated criteria.
- Conditionally approved: may proceed when named conditions are satisfied.
- Approved: authorized commitment by the decision owner.
- Implemented: action taken; future review concerns continuation or correction.
These labels prevent “preferred” from becoming “approved” through repetition. They also protect a sponsor from being accused of bypassing a process when they were legitimately testing feasibility.
If you suspect a decision is settled, ask what remains open and what has already been committed. “Can this committee still choose vendor B without breaching a contract or public promise?” is more revealing than “Is this a rubber stamp?” If the answer is no, the record should identify the earlier decision and its authority.
Voice after a decision can still have value, but it is not equivalent to voice before it. Experimental research has found that both pre- and post-decision voice can improve fairness judgments, with pre-decision voice producing stronger effects in the studied setting.2 The practical implication is straightforward: consultation should occur while input can still affect the choice, unless its purpose is honestly described as implementation feedback or review.
4. Blurring who made a decision and why
Decision provenance means the trace from question to authority, evidence, choice, and rationale. The phrase sounds technical; the idea is ordinary. Six months later, can a reasonable person tell what was decided, by whom, under which authority, using what central information?
Provenance blurs through vague minutes, passive voice, shared documents, undocumented verbal direction, and polite language that hides disagreement. “It was agreed” may mean a recorded vote, the chair’s conclusion, no one objected, or the most senior person expressed a preference.
This ambiguity can be convenient. Leaders avoid appearing unilateral, groups avoid recording conflict, and participants preserve relationships. The cost arrives later. Accountability cannot be assigned, new evidence cannot be compared with the original basis, and people remember the same meeting differently.
A concise decision record contains:
- the date and decision question;
- the authorized decision-maker or body;
- the option chosen;
- the decisive criteria or constraints;
- material evidence and its version;
- dissent, abstention, recusal, or unresolved risk where relevant;
- conditions, owner, and review date.
This is not a transcript. It does not need to reproduce every opinion. It preserves the facts needed to act and learn. Personal or sensitive detail should be minimized and access controlled.
If the decision came through a verbal instruction, confirm it respectfully: “My understanding is that you authorized option C today because the regulatory deadline rules out the longer implementation. I will proceed unless I have misunderstood.” The confirmation gives the decision-maker a chance to correct the record without drama.
5. Changing criteria after seeing the options
Criteria often evolve for good reasons. Evaluation reveals a requirement no one anticipated. A regulator changes a rule. A candidate raises a conflict. A prototype exposes a reliability issue. A rigid process that cannot learn may be less fair than one that adapts.
The problem is criterion change that selectively favors an option after decision-makers know who benefits, without a general and documented reason. A hiring panel values management experience until its preferred candidate lacks it, then emphasizes “future potential.” A project is approved on revenue forecasts but defended after poor results as a branding investment. A vendor’s weakness becomes “nonessential” while a competitor’s similar weakness remains disqualifying.
When criteria change, apply a reset test:
- Is the new criterion genuinely relevant to the decision purpose?
- Would it apply regardless of which option it favors?
- Can all options be reassessed against it?
- Do participants need new evidence or time?
- Who authorized the change, and is it recorded?
If the answer is yes, change may improve the process. Return to the full option set and rescore. If the change cannot withstand the same application, it may be a post hoc justification.
Weighting also matters. Criteria can remain nominally unchanged while their importance shifts. Publish weights before evaluation where practical. Where judgment cannot be reduced to numbers, ask decision-makers to identify which factor controlled the outcome. A score is not objective merely because it contains decimals.
Fair process does not mean majority rule
Many workplace decisions properly belong to an accountable individual. A safety officer may stop work. A budget owner may reject expenditure. A manager may choose among qualified candidates. A board may act over substantial internal dissent. Participation informs authority; it does not necessarily replace it.
The decision-maker owes three things in return. First, receive relevant input while the issue remains open. Second, apply or explain the criteria. Third, own the result. “The group decided” should not shield a leader who exercised final authority, just as “leadership wanted it” should not erase a committee’s recorded vote.
Consensus has a place when collective commitment is essential and time permits. It can also give every participant a veto, reward endurance, and hide pressure under a friendly label. Define what consensus means: no reasoned objection, everyone can live with the outcome, or full agreement. If the process ultimately allows a chair to decide, say so before disagreement arises.
When the process itself becomes a weapon
Detailed procedures can protect fairness. They can also exhaust participation. Endless requests for formatting, repeated jurisdiction disputes, and serial re-reviews can prevent a proposal or complaint from ever reaching substance. Chapter 10 will examine procedural burdens in depth; here the point is that decision controls need service standards.
Each process should identify completeness requirements, a contact for clarification, expected time, and what happens if the organization misses its own deadline. Review should improve a decision, not provide unlimited opportunities for whichever party dislikes the likely result.
Conversely, calls to “cut bureaucracy” can remove the very safeguards that make power accountable. Before eliminating a step, identify its purpose. If three approvals duplicate one another, combine them. If independent review exists because the primary decision-maker has an interest, speed is not a sufficient reason to delete it.
Decision records and privacy
Good records are selective. They preserve the basis of the decision without turning every discussion into permanent surveillance. Personal data, health information, complaints, legal advice, security details, and commercially sensitive material may require restricted annexes or separate systems.
A public or widely shared record might say: “Candidate B appointed following panel assessment against published criteria; one panel member recused because of a prior working relationship.” It need not reveal private references or interview notes. An access-controlled file can contain the evidence required by policy.
Do not create unofficial dossiers to compensate for a weak decision log. Ask the process owner to maintain the appropriate record. If you keep a personal note, record your own participation, the decision communicated to you, and any action you must take. Follow retention and privacy rules.
A response ladder for a questionable decision
Clarify status and authority. Ask whether the issue is proposed, preferred, conditionally approved, or approved, and who holds the decision right.
Request the rule and record. Obtain the mandate, criteria, membership requirement, or decision note you are entitled to see.
Identify the material gap. Name the venue change, missing participant, prior commitment, unclear provenance, or criterion shift and explain its effect.
Propose the smallest repair. Pause, reconvene with proper membership, circulate the full evidence, rescore all options, correct the minutes, or ratify under the authorized forum.
Preserve urgent safety or rights. Use a formal route promptly where required. Do not wait for ordinary process repair if the decision creates immediate harm or a legal deadline.
Review the design. After the immediate issue, correct the governance map, templates, thresholds, and service standards that allowed ambiguity.
For managers: a one-page decision charter
For repeated or high-stakes decisions, create a charter before evaluating options:
| Element | Question |
|---|---|
| Decision | What exactly will be chosen? |
| Authority | Who recommends, advises, approves, and implements? |
| Participants | Which expertise and affected interests must be heard? |
| Criteria | What standards and weights apply? |
| Evidence | What common material will participants receive, and when? |
| Conflicts | What requires disclosure or recusal? |
| Record | Where will the choice, basis, conditions, and dissent appear? |
| Review | What event or evidence can reopen it? |
Do not fill the charter with broad phrases. “Business needs” is not a criterion until it is translated into capacity, cost, risk, user need, or another assessable property. “Leadership” is not an authority until the relevant role or body is named.
Renee asks for the platform decision to be recorded against this structure. The steering committee confirms that it held final authority and that enterprise fit was a valid but late criterion. It commissions a short reassessment of all three vendors, including the project team’s evidence. The same vendor still wins, but the new record no longer misstates the team’s view. The organization improves the process without pretending that the first outcome must have been wrong.
Delegation without disappearance
Senior leaders cannot make every decision. Delegation places authority closer to relevant knowledge and keeps the organization moving. It also creates confusion when people hear “the director decided” even though the director delegated the matter, or “the team agreed” when a manager retained a veto.
A useful delegation states five things: the decision being delegated, the person or body receiving it, limits on money or risk, required consultation, and the point at which authority returns upward. For example:
The program lead may choose the supplier within the approved budget after security and procurement review. Any exception to the data-residency requirement returns to the technology risk committee.
This statement allows the program lead to decide. It does not turn the reviewers into co-decision-makers, and it does not allow a reviewer to replace the program choice merely by expressing a preference.
Delegation should travel with accountability and resources. Giving a person responsibility to deliver while reserving all meaningful choices elsewhere creates an owner in name only. Conversely, a delegate who accepts authority should not attribute an unpopular choice vaguely to “leadership” unless leadership actually directed it.
Temporary delegation needs an end. Acting arrangements, emergency powers, and substitute approvals can outlive the event that justified them. Record the start, scope, and expiry. If the arrangement works well enough to continue, approve it as a normal design instead of allowing temporary authority to harden invisibly.
Review, appeal, and the difference between the two
Review asks whether a decision should be reconsidered under defined conditions. Appeal asks another authority to examine an alleged error, unfairness, or breach. Organizations often use the terms loosely, which leads people to expect a fresh decision from a process designed only to check compliance.
A credible route explains:
- who may use it;
- on what grounds;
- within what time;
- what evidence is allowed;
- who reviews;
- what outcomes are possible; and
- whether action pauses while review occurs.
Grounds might include a material factual error, undisclosed conflict, procedural breach, new evidence unavailable earlier, or a decision outside authority. “I disagree with the outcome” may not be enough. That limit is reasonable when it is stated in advance and applied consistently.
Independence is proportional. A routine expense dispute may be reviewed by the next manager. A complaint about that manager may require a different route. A highly specialized technical decision may need an independent expert rather than a general administrator. The reviewer should not simply ask whether they would have chosen differently; they should apply the stated grounds.
An appeal route that exists only in a buried policy, requires information the applicant cannot obtain, or returns every matter to the original decision-maker is formally present and practically weak. Chapter 10 considers obstruction more fully. In decision design, the important step is to test the route before conflict: can a person find it, understand it, and receive a reasoned answer within the promised time?
A good outcome does not cure a weak process
When Renee’s preferred vendor loses but the selected system performs well, colleagues may conclude that the process concern was pointless. That conclusion confuses two questions. The system’s performance tells the organization something about the option. It does not explain why the venue changed, why evidence was summarized inaccurately, or why criteria shifted.
The reverse is also true. If the selected system fails, the failure does not prove that the infrastructure director manipulated the decision. New information, implementation errors, market changes, or ordinary uncertainty may explain the result. Evaluate the process against what was known and required at the time, then evaluate the outcome against subsequent evidence.
This separation supports learning. A sound process can make a reasonable choice under uncertainty and still get a bad result. The organization should update its assumptions without looking for a scapegoat. A weak process can get lucky; the organization should preserve the benefit while correcting how it decides. Experimental work on outcome bias provides a further reason to keep process evaluation distinct from what later happened.4
A retrospective can use two columns:
| At the time | After the result |
|---|---|
| Information available | New information learned |
| Criteria stated | Outcomes observed |
| Risks accepted | Risks that occurred |
| Authority and rationale | Corrective decision now |
Keeping the columns separate reduces hindsight reconstruction and prepares the ground for the narrative issues in Chapter 12.
Decisions supported by algorithms or AI
Organizations increasingly use scoring systems, automated recommendations, and generative tools to screen, rank, summarize, or advise. The tool does not eliminate the six parts of a legible decision. It adds questions.
Who selected the system? What task is it authorized to perform? Which data and criteria shape its output? Can an affected person correct inaccurate inputs? Who exercises human judgment, and are they able to depart from the recommendation? What monitoring or review detects unequal or unsafe effects?
Do not accept “the system decided” as provenance. A system has owners, configurations, thresholds, and users. Nor should “human in the loop” end the inquiry. A person who lacks time, information, training, or authority to challenge the output may be present without providing meaningful oversight.
The National Institute of Standards and Technology’s voluntary AI Risk Management Framework organizes risk work around governing, mapping, measuring, and managing.5 An organization does not need to adopt that framework to use the underlying discipline: define the context, assign owners, examine impacts, test performance, and maintain a route for correction. Where law regulates automated employment decisions, obtain appropriate legal and specialist advice.
Practice: reconstruct one decision
Choose a decision from the last three months. Without consulting anyone’s private messages, answer:
- What exact question was open?
- Who believed they had authority?
- Which forum actually made the choice?
- What criteria existed before options were known?
- What commitment occurred before approval?
- What does the final record say?
- What would cause review?
Mark each answer as clear, inferred, or unknown. Choose one unknown that matters to future action and ask the process owner to clarify it. The goal is not to reopen every decision. It is to keep the next one from inheriting the same ambiguity.
Decisions convert attention and support into commitment. The next chapter follows the resources that make those commitments real—and the dependencies that can quietly determine who has leverage over them.