How to Build an Assumption Register

Scope note
This is a general Ninth Atlas decision guide. It is not client-specific advice, statutory audit, regulated assurance or a substitute for appropriately qualified legal, tax, accounting, investment or technical professionals.

Direct answer

To build an assumption register, begin by defining the decision, the consequence of being wrong and the assumptions carrying the result. Gather evidence that can confirm or disconfirm those assumptions, test alternatives and document the conditions under which the recommendation remains valid.

Why this matters

For boards, founders, investors, strategy teams and business-unit leaders, the practical question is rarely whether a document looks complete. The question is whether the proposed decision can survive scrutiny, uncertainty and execution. How to Build an Assumption Register matters because it creates a disciplined bridge between the claim being made and the evidence needed to rely on it.

A strong review makes hidden assumptions visible, tests the commercial consequence of being wrong and converts findings into a clear choice with conditions. It should sharpen judgement rather than manufacture certainty.

When to use it

Use this approach when the decision is material, the case is being advanced by its own advocates, or the evidence is fragmented. It is especially useful before investment, board approval, market entry, launch, scaling, pricing action or operating-model change. It is less useful when the decision is easily reversible, delay is more costly than the likely risk, or the organisation has not defined what it wants reviewed.

Business questions it should answer

What must be true for the conclusion about build an assumption register to hold?
Which claims are facts, which are assumptions and which remain unknown?
What evidence would materially change the decision?
Where could selection, optimism or confirmation bias have entered the case?
What conditions should be attached to approval, investment or launch?

How the validation works

1. Frame the decision

State the decision, owner, deadline and consequence of being wrong. Define what build an assumption register must help the organisation decide.

2. Map the assumptions

Decompose the recommendation into demand, economics, capability, timing and execution assumptions. Record dependencies between them.

3. Grade the evidence

Review the available market demand evidence, customer and channel evidence, business-model logic. Distinguish direct evidence, proxy evidence, expert judgement and unsupported assertion.

4. Seek disconfirming evidence

Look deliberately for observations that weaken the preferred case. Test alternative explanations and competitor responses.

5. Stress the conclusion

Run downside, delay and execution scenarios. Identify thresholds at which the recommendation changes.

6. Convert analysis into conditions

Summarise the conclusion through an assumption register, evidence map, risk and resilience view, with owners and monitoring triggers.

7. Review after action

Compare realised evidence with the original assumptions and update the decision record. Validation should improve future judgement, not end at approval.

Evidence to gather

Direct evidence relevant to build an assumption register, rather than only industry averages or management assertion.
Recent and attributable sources covering market demand evidence, customer and channel evidence.
Evidence on business-model logic, commercial and operating assumptions.
Contrary or negative evidence that tests the preferred interpretation.
Documentation of data limitations, sampling limits, model choices and unresolved uncertainty.
A clear link between each material assumption and the source used to support it.

What a decision-ready conclusion looks like

A decision-ready conclusion should not stop at 'the evidence is positive' or 'more work is required'. It should state the recommended choice, strongest evidence, material risks, exposed assumptions and conditions that must be met. Typical outputs include an assumption register, evidence map, risk and resilience view, approval conditions, board-ready validation memo.

The reviewer should separate four levels of confidence: established fact, well-supported inference, plausible judgement and unresolved uncertainty. This prevents language from sounding more certain than the evidence allows.

Common mistakes

Treating a polished explanation of build an assumption register as verified evidence.
Collecting more information without identifying the assumption it is meant to test.
Using only evidence supplied by advocates of the plan.
Reporting a single number where a range or scenario is more honest.
Confusing statistical precision with commercial relevance.
Failing to specify what new evidence would reverse the recommendation.

Limitations and professional boundaries

This work is an independent commercial and strategic review. It is not statutory audit, legal advice, tax advice or financial-statement assurance.

Independent validation is constrained by access, time, data quality and stakeholder willingness to expose contrary evidence. A rapid review may identify material weaknesses without resolving every issue. The final page should state those limitations explicitly.

Practical checklist

Is the decision stated in one sentence?
Are material assumptions visible and owned?
Is each assumption linked to attributable evidence?
Has contrary evidence been sought?
Have downside, delay and execution scenarios been tested?
Are limitations explicit?
Does the recommendation include conditions, owners and triggers?
Is the regulated-services boundary clear?

Frequently asked questions

Does independent validation guarantee the outcome?

No. It reduces avoidable uncertainty, exposes weak assumptions and improves decision conditions, but cannot remove market, execution or timing risk.

Must every review include primary research?

No. Primary research is warranted when customer, channel or behavioural evidence is material and unavailable from reliable existing sources.

How should conflicting evidence be handled?

Preserve, grade and explain it. The goal is not forced consensus but a transparent view of what is known, disputed and decision-critical.

When the decision is live

Use this guide to frame the issue, identify the evidence required and decide whether the question can be resolved internally or needs independent challenge. Ninth Atlas engagements are scoped around the decision at stake rather than a fixed consulting menu.