Gold Standard · Intelligence specialist guide
Find the nearest question, then follow the evidence
An assessment can follow valid logic and still fail because one unexamined premise carries the whole argument. A Key Assumptions Check makes that load-bearing premise visible before it surprises the reader.
A key assumption is a proposition treated as true so that analysis can proceed, even though it may not be directly established. Assumptions are unavoidable. The danger appears when they remain hidden, become confused with evidence or persist after the conditions supporting them change.
This guide presents the Key Assumptions Check as a bounded diagnostic technique. It complements Assumption-Based Planning, which designs signposts and responses around planning assumptions, and Analysis of Competing Hypotheses, which compares explanations against evidence. Here the central task is to expose, support, challenge and rank the premises carrying one assessment.
Choose a reading route
Find the assumptions
Claims, premises and dependency chains.ROUTE 2 · CHAPTERS 4–5
Test support
Evidence, confidence and alternatives.ROUTE 3 · CHAPTERS 6–7
Work the check
Run a team example and reversal test.ROUTE 4 · CHAPTERS 8–10
Challenge safely
Bias, indicators, updating and ownership.ROUTE 5 · CHAPTERS 11–12
Transfer and audit
Planning boundaries and analytic records.
Open the full chapter index · Use the field check · Return to the Intelligence root
Chapter contents
Chapters 1–3 · Claims, premises and hidden load
Chapters 4–6 · Support, confidence and a worked check
Chapters 7–9 · Challenge, change and monitoring
Chapters 10–12 · Transfer, governance and audit
CHAPTER 1 OF 12 · Find the assumptions
1. Separate observations, inferences and assumptions
Observations describe reported or measured information. Inferences connect that information to a judgement. Assumptions fill gaps that analysis currently cannot resolve. The three may interact, but they should not be written as one seamless paragraph because readers then cannot see what could be wrong.
Take the draft assessment sentence by sentence. Label each proposition observed, inferred or assumed. Ask what would have to be true for each inference to hold. Convert those necessary but unproven conditions into explicit candidate assumptions.
A team observes delayed deliveries and infers that a supplier lacks capacity. Hidden premises include stable demand, reliable tracking data and no deliberate prioritization of other customers. Each premise could fail for a different reason.
Changed-context transfer check
Label the evidence chain behind a prediction of exam readiness. Distinguish observed performance from assumptions about future practice and test conditions.
Limit and misconception
The labels are analytical roles, not permanent properties. A proposition can move from assumption to supported judgement as evidence improves.
Evidence practice
Keep a proposition ledger with role labels and source links. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: writing an inference as though it were a direct observation. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
CHAPTER 2 OF 12 · Find the assumptions
2. Identify assumptions that are necessary, not merely convenient
A list of every uncertainty overwhelms the method. Key assumptions are premises whose failure would materially weaken, reverse or reframe the assessment. Convenience assumptions may simplify calculation without carrying the central conclusion.
Draw a dependency chain from evidence to judgement. For each premise, ask whether the conclusion still follows if it is false. Rank by impact and uncertainty. Keep a separate watch list for contextual conditions that matter but are not logically load-bearing.
An assessment that a project will finish on time depends on timely access to a test environment. The colour of the reporting template is uncertain but not key. Reversing the access assumption changes feasibility; reversing the template assumption does not.
Changed-context transfer check
Apply the dependency test to a health-service demand forecast and identify one key and one non-key assumption.
Limit and misconception
Materiality depends on the decision and time horizon. A small operational premise may become key when buffers are thin.
Evidence practice
Keep an assumption-to-conclusion dependency map. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: calling every unknown a key assumption. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
Contents · Previous chapter · Next chapter
CHAPTER 3 OF 12 · Find the assumptions
3. Phrase assumptions so they can be examined
An assumption such as “the situation remains normal” is too vague to test. A useful statement names the actor or mechanism, time period, boundary and expected condition. It should be clear what evidence would increase or reduce confidence.
Rewrite broad beliefs as falsifiable propositions. Avoid double assumptions joined by and. State the relevant horizon and population. Add a short rationale explaining why the analysis currently accepts the proposition.
“The partner will cooperate” becomes “The partner will provide weekly inventory data in the agreed format through the next eight weeks.” The narrower wording reveals what to observe and what a partial failure would mean.
Changed-context transfer check
Rewrite “students will keep improving” into testable assumptions about attendance, practice quality and feedback use.
Limit and misconception
Not every social proposition can be tested like a laboratory hypothesis. The aim is inspectability and disciplined update, not false precision.
Evidence practice
Keep a one-proposition-per-row assumption register. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: using vague positive language that no evidence can challenge. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
Contents · Previous chapter · Next chapter
Assumptions may rest on direct reporting, historical pattern, analogy, expert judgement or the absence of contrary evidence. These supports have different strengths and failure modes. Repeated reports may share one source and create an illusion of corroboration.
For each assumption, record supporting and contradicting evidence, source independence, recency, access limitations and plausible deception or error. Distinguish no contrary evidence from positive support. Note which evidence was collected specifically to test the assumption.
Three status reports repeat that equipment is ready, but all quote the same contractor. The count is three documents, not three independent sources. An inspection finding would contribute a different kind of support.
Changed-context transfer check
Audit support for an assumption about household demand using survey, transaction and interview data. Identify shared dependencies.
Limit and misconception
A support score can organize review but should not replace reasons. Weak evidence does not prove an assumption false, and strong evidence can become stale.
Evidence practice
Keep a support-contradiction matrix with source-dependence notes. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: counting repeated reporting as independent corroboration. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
Contents · Previous chapter · Next chapter
Confidence communicates how firmly the analyst holds a judgement after considering evidence quality, consistency, knowledge gaps and assumptions. It should not be copied automatically from a probability estimate, and it should change when a key premise lacks support.
Use a defined confidence scale and explain the main drivers. Record which key assumption limits confidence and what evidence would change it. If the judgement remains likely but the support is fragile, say both instead of choosing one reassuring label.
A team judges disruption more likely than not but has low confidence because one key capacity estimate is indirect and six months old. Probability addresses the outcome; confidence addresses the foundation of the judgement.
Changed-context transfer check
Write probability and confidence statements for a fictional maintenance failure where data are sparse but mechanisms are understood.
Limit and misconception
Confidence terminology varies across institutions. Use the governing standard and avoid converting labels into invented numerical precision.
Evidence practice
Keep a confidence rationale tied to named assumptions and gaps. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: treating high probability and high confidence as synonyms. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
Contents · Previous chapter · Next chapter
CHAPTER 6 OF 12 · Work the check
6. Worked example: review a fictional supply interruption assessment
A fictional school expects laboratory materials to arrive before a practical assessment. The initial judgement is that delivery will be on time. The team lists assumptions about supplier stock, transport capacity, customs clearance, order accuracy and the absence of competing priority orders.
Each assumption is tested for necessity, support, contradiction and sensitivity. Supplier stock has direct evidence. Transport capacity relies on an old schedule. Customs timing is historical but volatile. The team identifies transport capacity as both uncertain and capable of reversing the overall judgement.
The revised assessment remains cautiously positive but lowers confidence, names the vulnerable assumption and recommends checking confirmed carriage. The method does not manufacture a different answer; it makes the answer’s dependency visible.
| Assumption | Current support | If false | Action |
|---|---|---|---|
| Stock is reserved | Direct supplier confirmation | Delivery impossible | Verify reservation identifier |
| Transport capacity exists | Old schedule only | Delivery likely late | Obtain confirmed carriage |
| Customs timing is typical | Historical range | Buffer may vanish | Track clearance signposts |
| Order details are correct | Internal review | Wrong items may arrive | Independent order check |
Changed-context transfer check
Change the case to a digital service launch. Replace physical transport assumptions with capacity, integration, identity and support assumptions.
Limit and misconception
The example is fictional. A Key Assumptions Check improves analytic transparency; it does not guarantee that all hidden premises will be found.
Evidence practice
Keep a completed assumption register and revised judgement note. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: forcing the check to overturn the original conclusion. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
Contents · Previous chapter · Next chapter
CHAPTER 7 OF 12 · Work the check
7. Run an assumption reversal and sensitivity test
Reversal asks what follows if an assumption is false. It is not a claim that the opposite is true. The exercise reveals which premises change the judgement, which merely adjust timing and which have compensating evidence or options.
For every key assumption, write a plausible failure state and recompute the conclusion. Mark outcome reversal, confidence reduction, scope change or no material effect. Then test combinations, because two moderate failures can matter more than either alone.
If transport capacity is absent, the on-time judgement reverses. If customs takes one extra day, the delivery remains possible but the buffer disappears. If both occur, alternatives must be activated immediately.
Changed-context transfer check
Reverse one assumption in an educational progress forecast and show whether the conclusion, confidence or recommended action changes.
Limit and misconception
A reversed assumption should remain plausible. Extreme opposites create drama without testing the working model.
Evidence practice
Keep a sensitivity table showing conclusion changes under each reversal. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: treating reversal as evidence that the assumption is actually false. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
Contents · Previous chapter · Next chapter
CHAPTER 8 OF 12 · Challenge safely
8. Use diverse challenge without turning review into performance
Hidden assumptions are easier to see when reviewers bring different knowledge and incentives. Yet challenge fails when status discourages dissent, when the loudest skeptic wins or when participants attack authors rather than propositions.
Use a facilitator, circulate the draft, invite silent assumption generation before discussion and separate identification from evaluation. Record minority views and reasons. Ask subject experts where mechanisms could break and outsiders where language hides premises.
A junior analyst notices that “normal demand” assumes no competing national event. Anonymous collection brings it forward before group discussion, preventing hierarchy from filtering it out.
Changed-context transfer check
Design a short classroom assumption review in which students can challenge a solution method without challenging one another personally.
Limit and misconception
Diversity of viewpoint does not guarantee accuracy. Participants still need relevant evidence, scope and a method for resolving factual disputes.
Evidence practice
Keep a facilitated challenge record including minority assumptions. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: measuring quality by the heat of disagreement. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
Contents · Previous chapter · Next chapter
CHAPTER 9 OF 12 · Challenge safely
9. Connect vulnerable assumptions to indicators and collection
A check is useful beyond publication when vulnerable assumptions have observable implications. Indicators and collection requirements convert “we should watch this” into a specific plan, while keeping the assumption register distinct from the monitoring system.
For each high-impact, low-support assumption, identify leading indicators, thresholds, source owners, collection frequency and a review trigger. Include indicators that would strengthen the assumption, not only those that confirm fear.
Confirmed carriage strengthens the transport assumption; repeated schedule slips weaken it. A trigger requires reassessment when no booking exists by a named date, rather than waiting for final non-delivery.
Changed-context transfer check
Create balanced indicators for an assumption that a new study routine is sustainable.
Limit and misconception
Indicators can be noisy, manipulated or lagging. A threshold is a prompt for review, not automatic proof that the premise failed.
Evidence practice
Keep an assumption-to-indicator table with owners and review triggers. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: collecting only signals that confirm the preferred story. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
Contents · Previous chapter · Next chapter
CHAPTER 10 OF 12 · Challenge safely
10. Update the assessment without erasing its history
When an assumption changes, analysts should preserve the original judgement, the new evidence, the revision and the decision consequence. Silent rewrites destroy learning and make it impossible to distinguish poor reasoning from a changed world.
Version the register and assessment. Date each confidence change, identify the assumption affected and state whether the outcome probability, scope or action changed. Close assumptions only with a reason; do not delete them from the record.
A carrier cancels a route. The transport assumption moves from accepted with low support to contradicted. The team revises the delivery judgement and activates a local substitute, while preserving the earlier evidence and reasoning.
Changed-context transfer check
Update a forecast after a new test result contradicts a key learning assumption.
Limit and misconception
Version history can become burdensome. Keep material changes and decision-relevant reasons rather than every wording edit.
Evidence practice
Keep a dated assumption and judgement change log. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: replacing the old conclusion without recording why. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
Contents · Previous chapter · Next chapter
CHAPTER 11 OF 12 · Transfer and audit
11. Distinguish the check from competing-hypotheses and planning techniques
A Key Assumptions Check examines premises supporting one assessment. Analysis of Competing Hypotheses compares multiple explanations against diagnostic evidence. Assumption-Based Planning prepares signposts and actions around uncertain planning premises. They can connect, but each has a different job.
Use the smallest method that fits the question. Start with the check when hidden premises may carry a judgement. Move to competing hypotheses when rival explanations need comparison. Move to planning when uncertainty must be managed through shaping and hedging actions.
The supply assessment uses a Key Assumptions Check to find transport dependence, a competing-hypotheses matrix if delays have several explanations, and Assumption-Based Planning to prepare alternative supply actions.
Changed-context transfer check
Choose the right technique for a fictional drop in attendance and justify the sequence.
Limit and misconception
Combining techniques can improve coverage but also create paperwork. Do not run a suite merely to signal rigor.
Evidence practice
Keep a method-selection note tied to the analytic question. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: using different technique names for the same undifferentiated workshop. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
Contents · Previous chapter · Next chapter
CHAPTER 12 OF 12 · Transfer and audit
12. Audit whether the judgement reveals its load-bearing premises
The final audit asks whether a reader can identify the key assumptions, see their evidential support, understand confidence, know what reversal would change and trace later updates. The record should make uncertainty usable rather than bury it in caveats.
Select a completed assessment and ask an independent reviewer to reconstruct its dependency chain. Compare their assumptions with the original team’s register. Test one reversal and one new evidence item. Record any premise that only becomes visible during reconstruction.
A reviewer discovers that the delivery judgement also assumed staff would be available to receive materials. The assessment is amended, and a new indicator is assigned, before the premise causes operational failure.
Changed-context transfer check
Run the audit on an assessment from a different domain and compare what counts as support.
Limit and misconception
No audit can prove that every assumption is visible. Repeated challenge, outcomes review and institutional memory remain necessary.
Evidence practice
Keep an independent reconstruction with unresolved premise gaps. Label every field as observed, computed, assumed or chosen. Give the record to a second reader and ask them to reproduce the result without the original author explaining what was meant. Differences reveal hidden definitions, missing versions or decision rules that still need to be made explicit.
Deliberately test for this failure: declaring the assessment robust because a checklist was completed. Record the earliest point at which the process can detect it, the correction applied and whether later conclusions change. A method is transferable only when its controls work on an altered case, not merely on the example used to teach it.
Contents · Previous chapter · Connected articles
Advanced boundary cases
Strategic assumptions shared by an institution
A premise may be so widely accepted that no single assessment states it. Periodic cross-product reviews can surface these strategic assumptions and test whether consensus reflects evidence or inherited framing.
Adversarial or deceptive environments
Evidence may be shaped to sustain a false premise. Record source access, incentives, denial and deception possibilities, and seek indicators that are costly for an actor to fake.
Assumptions about absence
“We would have seen it” depends on collection coverage, reporting incentives and detection probability. Make those visibility assumptions explicit before treating silence as evidence.
Time-dependent validity
An assumption can be well supported for one horizon and fragile for another. Date the claim and schedule review before the evidence ages beyond its useful window.
A practical field check
Choose one consequential judgement. Build a concise register that shows what the conclusion requires, how each premise is supported and what would happen if it failed.
- Separate observations, inferences and assumptions.
- Map premises that are necessary for the conclusion.
- Rewrite each key assumption as one testable proposition.
- Record supporting, contradicting and dependent sources.
- State confidence and its main limiting assumption.
- Reverse each key premise and recompute the judgement.
- Assign indicators, review triggers and a versioned update owner.
Run the check as a small workbook
Complete four passes. First, use a clean case whose answer is known. Second, introduce one edge condition and predict the failure before running the method. Third, change the setting while preserving the underlying structure. Fourth, hand the evidence to another reader for independent reconstruction. Keep disagreements: they show where definitions, versions, values or decision rules remain implicit.
| Audit question | Evidence | Stop condition |
|---|---|---|
| Which premises carry the judgement? | Dependency map and ranked register | Uncertainties are listed without impact |
| How are they supported? | Source-aware support matrix | Repetition is counted as independence |
| What if one fails? | Reversal and combination tests | No conclusion is recomputed |
| How will change be noticed? | Indicators, triggers and version log | Assumptions are checked only once |
Chapter-by-chapter evidence audit
| Chapter | Evidence to keep | Failure to catch |
|---|---|---|
| 1. Separate observations, inferences and assumptions | A proposition ledger with role labels and source links | Writing an inference as though it were a direct observation |
| 2. Identify assumptions that are necessary, not merely convenient | An assumption-to-conclusion dependency map | Calling every unknown a key assumption |
| 3. Phrase assumptions so they can be examined | A one-proposition-per-row assumption register | Using vague positive language that no evidence can challenge |
| 4. Assess evidential support and source dependence | A support-contradiction matrix with source-dependence notes | Counting repeated reporting as independent corroboration |
| 5. Express confidence without hiding fragility | A confidence rationale tied to named assumptions and gaps | Treating high probability and high confidence as synonyms |
| 6. Worked example: review a fictional supply interruption assessment | A completed assumption register and revised judgement note | Forcing the check to overturn the original conclusion |
| 7. Run an assumption reversal and sensitivity test | A sensitivity table showing conclusion changes under each reversal | Treating reversal as evidence that the assumption is actually false |
| 8. Use diverse challenge without turning review into performance | A facilitated challenge record including minority assumptions | Measuring quality by the heat of disagreement |
| 9. Connect vulnerable assumptions to indicators and collection | An assumption-to-indicator table with owners and review triggers | Collecting only signals that confirm the preferred story |
| 10. Update the assessment without erasing its history | A dated assumption and judgement change log | Replacing the old conclusion without recording why |
| 11. Distinguish the check from competing-hypotheses and planning techniques | A method-selection note tied to the analytic question | Using different technique names for the same undifferentiated workshop |
| 12. Audit whether the judgement reveals its load-bearing premises | An independent reconstruction with unresolved premise gaps | Declaring the assessment robust because a checklist was completed |
Connected articles across the eduKate ecosystem
Start with the governing route
- The Gold Standard of Intelligence
- The Gold Standard of Education | Learning That Transfers into Life
- How X Works hub
Continue into verified specialist studies
- Assumption-Based Planning, signposts, shaping actions and hedging actions — manages planning assumptions after they are identified.
- Competing hypotheses, diagnosticity and evidence matrices — compares rival explanations rather than one judgement’s premises.
- Collection requirements, source validation and analytic gaps — turns vulnerable assumptions into focused evidence needs.
- Indicator validation, base rates and warning thresholds — tests whether monitoring signals deserve the role assigned to them.
- Red teaming, devil’s advocacy and premortem analysis — provides distinct challenge methods and boundaries.
Questions readers often ask
Are assumptions signs of weak analysis?
No. Analysis always relies on premises. Quality depends on making material ones visible, supported and revisable.
How many assumptions should a check contain?
Enough to expose the load-bearing premises, not every uncertainty. Rank them by impact and support.
Does reversing an assumption mean believing the opposite?
No. Reversal is a sensitivity test that asks what the judgement would become if the premise failed.
Is confidence the same as probability?
No. Probability concerns the assessed likelihood of an outcome; confidence concerns the strength of the assessment’s foundation.
When should the check be repeated?
Before finalizing a major judgement, when material evidence changes, when indicators cross review triggers or when assumptions age.
Can one analyst run it alone?
Yes, but a diverse, well-facilitated team is more likely to expose premises that the original author treats as obvious.
Sources and scope
CIA Center for the Study of Intelligence: A Tradecraft Primer — official primer describing the Key Assumptions Check and other structured analytic techniques.
CIA Center for the Study of Intelligence: Tradecraft Primer overview — official publication page and context for the primer.
CIA Center for the Study of Intelligence: Psychology of Intelligence Analysis — foundational treatment of mental models, bias and improving analysis.
Office of the Director of National Intelligence: ICD 203 Analytic Standards — official standards on objectivity, uncertainty, assumptions, alternatives and source quality.
The supplier, school and assessment cases are fictional. The method supports transparent reasoning; it does not disclose classified material, replace domain expertise or guarantee correct prediction.
“Gold standard” describes the educational benchmark explained in this guide. It is not an official accreditation, legal opinion, policy determination or professional certification.
Back to contents · Return to the Intelligence root · Return to the Gold Standard of Education