# Locate uncertainty in the work

Break the option into inspectable slices and attach assumptions, dependencies and learning actions.

Duration: 5 min

Break the option into independently discussable slices. For each one, record its outcome link, included work, assumptions, dependencies and evidence basis.

Classify uncertainty only after naming its driver. Relevant comparison data and no material unresolved dependency may support a lower-uncertainty label. Novel technology, external systems or unanswered design questions may make uncertainty material. If evidence cannot support a useful range, investigate first.

Avoid automatic percentage uplifts unless your organisation has calibrated them against its own outcomes.

## initial-estimates-m1-l2-reading

### Uncertainty belongs to a driver

For each work slice, ask what assumption, dependency or novelty could materially change the implementation. Record a learning action next to every material driver. If no feasible action can make the estimate decision-useful, reduce the option or stop and investigate.

This lightweight practice borrows transparency principles from GAO guidance; it is not a substitute for quantitative risk and uncertainty analysis.

### Worked example and transfer practice

All examples below are fictional teaching material. Use synthetic notes; no workplace or personal data is required or uploaded.

**Locate the uncertainty before adding money.** For the fictional reconciliation option, one export format is documented, while the second may omit transaction identifiers. Name the second connector as a material uncertainty. A five-day synthetic extraction test will check whether records can be matched reliably. Stop or reduce the option if the identifier is unavailable. A uniform buffer would not reveal which evidence could change the estimate.

**Try before reading the guide:** Record one uncertain work slice, its assumption, an evidence action, an owner and an end condition. Explain the decision if the action cannot resolve the unknown.

<details>
<summary>Practice review guide</summary>

The action is bounded and tests the named driver. A generic request to ‘research more’ has no stopping rule. A range may remain appropriate when useful evidence cannot support one precise value.

</details>

## initial-estimates-m1-l2-check

Public answer keys are intentional. This check does not establish competence.