Record the source, owner and validation for each MUST; agree the scoring scale and reasons for weights.
EVIDENCE LAB / 03 / DECISION ANALYSIS
Choose deliberately.Protect the decision.
A preference score is never a license to ignore mandatory requirements. Evaluate three fictional alternatives and challenge the apparent winner.
FICTIONAL3 STAGESCHALLENGE THE SCORE
> open /lab-03/decision.txt
Three hosting options. One defensible choice.
A fictional reporting service needs a new hosting platform. A procurement summary recommends the lowest monthly price, while an engineering team favors the solution with the most features. The organization has three mandatory requirements that must be met before preferences are scored.
3OPTIONS
3MANDATORY MUSTS
60mMAXIMUM RECOVERY TIME
Mission: screen eligibility, make value judgments transparent, and reduce the downside risk of the chosen alternative. All scores, tests and offers are fictional.
THE INVESTIGATION BRIEF_
Make a decision defensible, not merely numerical
Which eligible option best fits the stakeholders' priorities and still has a credible recovery and exit strategy?
Useful lens: Kepner–Tregoe-style decision analysis: MUST eligibility, WANT comparison and explicit adverse-consequence review.
What the record establishes
- Option C fails a stated residency MUST despite a favorable price.
- Options A and B satisfy the three fictional eligibility tests.
- The initial weights favor A, but a reasonable sensitivity change can favor B.
What remains open
- Whether the original weights accurately capture stakeholder priorities.
- Whether the tested export and recovery results generalize to production volumes.
- Which contractual and operational risks remain after the migration.
Evidence boundary
Scores and weights are illustrative preferences—not measured probabilities or empirical facts. Requirements must be validated by accountable business and technical owners.
STAGE 01 / NON-NEGOTIABLE CONSTRAINTS
The lowest price may not be eligible.
The decision owner sets three explicit MUST criteria: (1) EU-only data residency, (2) recovery time objective demonstrated at or below 60 minutes, and (3) complete data export demonstrated within 48 hours. These are illustrative exercise constraints, not legal claims about all businesses.
EVIDENCE A / Eligibility verification
| Option | EU residency | Tested recovery | Full export |
|---|---|---|---|
| Managed A | Yes | 45 min | 18 hours |
| Self-hosted B | Yes | 55 min | 8 hours |
| Budget C | No | 30 min | 12 hours |
In this fictional exercise, the test results are recorded against representative full-sized datasets and a defined recovery drill.
DECISION / WHICH OPTIONS ARE ELIGIBLE?
A — Exclude C, then compare A and B
Correct screening. C fails a stated non-negotiable condition. Its attractive price or recovery result cannot compensate for ineligibility. A and B meet the verified MUSTs and can advance to preference scoring.
Continue to the evidence →B — Allow C because the low cost is overwhelmingly attractive
Invalid trade-off. A MUST is an eligibility condition, not a weighted preference. Reconsidering a legitimate requirement is a separate governance decision, not something a high score can silently override.
Continue to the evidence →STAGE 02 / MAKE PREFERENCES EXPLICIT
Scoring helps, but cannot decide for you.
The stakeholders score eligible options from 1 (poor) to 5 (excellent). Higher is better. These values reflect fictional judgments and must be challenged for bias and sensitivity.
EVIDENCE B / Weighted preferences
| WANT | Weight | Managed A | Self-hosted B |
|---|---|---|---|
| Operational simplicity | 5 | 5 | 2 |
| Cost predictability | 3 | 3 | 4 |
| Customization | 2 | 2 | 5 |
| Weighted total | 10 | 38 | 32 |
Totals: A = 5×5 + 3×3 + 2×2 = 38. B = 5×2 + 3×4 + 2×5 = 32.
EVIDENCE C / Sensitivity check
Suppose customization turns out to matter more than stakeholders first assumed, increasing its weight from 2 to 5. Recalculate:
Managed A: 5×5 + 3×3 + 5×2 = 44.
Self-hosted B: 5×2 + 3×4 + 5×5 = 47.
The ranking reverses. A six-point preference advantage under the original weights is not a robust proof of the “best” option.
DECISION / HOW DO YOU INTERPRET 38 VERSUS 32?
A — Prefer A provisionally, then test sensitivity and risks
Good judgment. A leads under the stated preferences, but B can lead when priorities shift. Confirm stakeholder values, assess actual downside risk and make an explicit decision record before committing.
Continue to risk planning →B — Announce that 38 mathematically proves A is objectively best
False precision. Multiplying subjective weights by subjective ratings provides a transparent comparison, not an objective fact. A sensitivity analysis shows the conclusion can reverse.
Continue to risk planning →STAGE 03 / ANTICIPATE THE DOWNSIDE
A promising option still needs protective actions.
The exercise decision owner leans toward A because of lower operating effort, but worries about export lock-in and a difficult rollback if service recovery fails in real operations. The earlier tests meet stated MUSTs, but system change still creates new risk.
DECISION / WHAT MAKES THE PREFERRED OPTION SAFER?
A — Pilot a representative migration, export, recovery and rollback rehearsal
Risk-aware choice. Establish measurable acceptance criteria before contracting. Run a complete pilot on representative data, verify recovery and exit routes, and decide on triggers that stop or reverse the migration. Reassess whether the preference advantage remains credible.
Continue to the debrief →B — Sign immediately; the requirements and scores are enough
Under-controlled. Passing a test does not mean all future changes will be harmless. Review contractual exit rights, operational readiness, new dependencies and a tested rollback path before high-impact commitment.
Continue to the debrief →> lab-03 / decision-review
Debrief: eligibility before preference.
- MUSTs: EU residency, documented recovery objective and tested full export. C fails the residency requirement and is excluded.
- WANTs: A scores 38, B scores 32 under the original preferences; this is not a universally correct or objective ranking.
- Sensitivity: customization weight 5 reverses the ranking (A = 44, B = 47), revealing where stakeholder values matter.
- Downside: exit difficulty, recovery uncertainty, contractual dependencies and hidden operating costs.
- Protection and verification: representative migration rehearsal, complete export and recovery tests, rollback criteria, post-migration service objectives and a scheduled review.
Reasoning workshop
Why the ranking changes
With weights 5, 3 and 2, A scores 38 versus B's 32. Increasing the customization weight from 2 to 5 changes those totals to A = 44 and B = 47. The reversal shows that the outcome depends on the importance assigned to customization—not that either option was calculated incorrectly.
What deserves a separate risk review?
Document credible failure scenarios, probability ranges where defensible, severity, early-warning indicators, ownership and contingent actions. Weighted WANT scores should not silently absorb recovery, exit, security or concentration risks.
Verification plan
Run sensitivity checks, inspect vendor exit terms and rehearse representative data export, recovery and rollback.
Measure service objectives, recovery readiness and actual operating burden; review assumptions when volume or business priorities change.
Takeaway: A strong decision record makes disagreement and uncertainty visible; a higher weighted score is only one part of the decision.
TRANSFER THE METHOD_
Kepner–Tregoe-style decision analysis makes assumptions visible. It does not erase subjective judgments or the obligation to evaluate the risks of a preferred alternative.
You've completed the three investigations.
Evidence Lab overview →Compare your approach across incident diagnosis, quality control and decision analysis.
This scenario is invented. Criteria and observations are for training and are not legal, procurement, contractual or infrastructure advice.