Engineering judgment · guided practice

Build a decision another person can review.

A decision record preserves why a choice made sense at a particular time. It connects the problem, options, criteria, evidence, tradeoffs, and conditions that should reopen the decision.

An engineer compares a fractured component with two alternatives beside a test mechanism. — **Choose with the failure visible**Options · evidence · tradeoffs · review

Concept

A defensible decision is not the same as a certain decision.

Engineering decisions are made under limits: time, budget, available evidence, interfaces, safety, maintainability, and reversible or irreversible consequences. Judgment becomes reviewable when those limits are explicit and alternatives are compared against agreed criteria.

A record should explain the choice without implying that uncertainty vanished. It should also help a future reviewer distinguish a poor decision from a reasonable decision followed by an unexpected outcome.

Exercise

Complete the record in seven passes.

1. Decision statement

Write one sentence: “We must decide whether to ___ in order to ___ before ___.” Name the accountable decision-maker and affected parties.

One bounded choice

2. Context and boundary

Describe the system, current condition, constraints, interfaces, and what this decision will not resolve. Separate requirements from preferences.

What is in and out

3. Options

Include the current state when plausible. Describe at least two options fairly enough that a supporter of each would recognize it.

No straw alternatives

4. Criteria

Choose criteria before scoring options: safety, performance, cost, schedule, reversibility, maintainability, accessibility, or other relevant concerns.

Weight only with rationale

5. Evidence and uncertainty

For each material claim, name the source or test. Mark assumptions, missing data, confidence, and the cost of being wrong.

Trace the support

6. Choice and tradeoffs

State the selected option and why it best fits the criteria now. Name what is sacrificed, who bears the downside, and any dissent.

Own the consequence

7. Review trigger

Name an observable condition, date, threshold, or new evidence that reopens the decision. Add an owner for monitoring it.

Decisions have lifecycles

Finish with the boundary

State what the record does not approve, verify, certify, or predict. Link any required specialist or authority review.

Do not overclaim

Worked example

Choose a monitoring approach for a community archive pilot.

A small team must choose between manual weekly checks, automated alerts, and a hybrid approach. Manual checks are inexpensive to begin and easy to understand, but detection may be delayed and the process depends on consistent staff time. Automated alerts improve detection speed but create configuration, false-positive, maintenance, and access-control concerns. A hybrid option adds cost and coordination while reducing dependence on either mechanism alone.

The team chooses a limited hybrid pilot because recovery time and staff capacity matter more than minimum setup cost. The record names the sacrifice: additional configuration and review burden. It also names evidence still missing: alert reliability under representative conditions. The decision reopens after four weeks, after any missed critical event, or if weekly review time exceeds the agreed threshold. This example demonstrates structure only; it does not recommend a monitoring design for a real archive.

Artifact structure

Copy this compact record.

Use Markdown, a document, or the browser-local Signal Brief Builder. Keep each field concise enough that a reviewer can find the reasoning.

# Engineering decision record

Decision:
Owner and date:
Affected parties:

## Context and boundary
System and current condition:
Requirements and constraints:
Outside this decision:

## Options and criteria
Options considered:
Decision criteria:

## Evidence and uncertainty
Sources or tests:
Assumptions and unknowns:
Failure modes and consequences:

## Decision
Selected option and rationale:
Tradeoffs and dissent:

## Review
Monitoring owner:
Review trigger or date:
What this record does not establish:

Review

Try to reopen your own decision.

Ask a reviewer to identify one omitted alternative, one criterion that may be weighted incorrectly, one unsupported claim, and one stakeholder bearing a consequence. Revise the record or document why the challenge does not change it.

Professional boundary: this exercise is educational. It is not engineering approval, safety assurance, design authority, or a substitute for applicable standards, testing, licensure, or organizational review.

Source shelf

Continue with established systems practice.

The NASA Systems Engineering Handbook addresses decision analysis within broader systems-engineering practice. The INCOSE Systems Engineering Handbook overview provides another entry into lifecycle and technical-management practice. These sources do not endorse ARCTURA.

Next: model the domain Use the review rubric