Collect evidence that enables a decision

Start with the decision and its conditions

An analysis activity should produce information someone can use. Before requesting another report, identify the decision, its owner, deadline and conditions the outcome must satisfy. Calling information interesting or confidence-enhancing is insufficient.

Explain which uncertainty it resolves and what might change if the answer differs. In a fictional project, Business must choose an option compatible with historical-query needs. Three states are admissible: S1 needs a short horizon without exception E; S2 needs a long horizon without E; S3 needs a long horizon with E.

The exercise assumes this list is complete for the decision, not that it represents every banking service. If another state becomes plausible, the plan needs revision. The analyst prepares collection without inventing authority to select the solution.

Also record whether the owner requires only an option feasible under existing information or wants to identify the least costly option for the actual state. Those requirements can justify different information plans.

Represent what may still be true

In the example, A costs 20 units and is feasible only in S1; C costs 25 and is feasible in S1 and S2; B costs 30 and is feasible in all three states. Costs are fictional totals within one horizon, with no other amounts; outcomes are otherwise equivalent.

Feasibility is mandatory. Before collection, B is the only option working in every still-admissible state. This does not prove B is least costly in the actual state: A costs less in S1 and C in S2.

The record should distinguish these conclusions. Do not assign probabilities merely because three states exist, or average costs as though states were equally likely. Also do not turn the option robust to current information into automatic approval.

The owner may decide to collect more information, accept the robust option’s cost or revisit conditions within their mandate. The plan’s role is to explain conclusions supported by evidence and uncertainty that remains relevant.

Define the observation and permitted interpretation

Collection H confirms, in the model, whether the required horizon is short or long and needs two work hours. Collection E confirms whether the exception is needed and requires three hours. Both use an authorized source applicable to the analyzed scope; correct classification is an exercise assumption, not a guarantee about real interviews or tests.

H=short retains only S1. H=long retains S2 and S3. E=no exception retains S1 and S2; E=exception retains S3.

Write these relationships before collection to prevent choosing interpretation after seeing a convenient answer. Nonresponse means neither a short horizon nor absence of exception. An inconclusive result retains states not excluded by valid evidence.

If a source describes another version or population, check applicability before reducing the set. The plan should identify the expected product, who confirms meaning and how insufficient answers will be communicated. Performing the activity does not automatically demonstrate that uncertainty was resolved.

Compare conditional plans

If the requirement is to identify the least costly option for the actual state, collection can adapt to the first answer. Start with H: a short horizon identifies S1 and A as its least costly feasible option. E is unnecessary for that decision because the supplied state list admits no short horizon with the exception.

If H indicates long, E must distinguish S2 from S3; the choice will be C or B. With one person, no waiting and sequential work, this plan needs two hours in the first branch and five in the second. Reversing order can also resolve the decision: E=exception identifies S3 in three hours; E=no exception then requires H, totaling five.

Without probabilities there is no basis for an average duration or claiming one order always superior. Compare maximum duration, answers permitting stopping and actual source availability. The stopping rule should follow the intended decision while retaining the possibility of uncertainties outside the model.

Plan within a deadline without inventing certainty

Suppose only three hours remain before the committee and one person is available, with no parallel work possible. H followed by E does not fit every branch. A short result resolves the decision in two hours; a long result leaves uncertainty unresolved at the deadline.

Starting with E can identify S3 in three hours, but no exception leaves S1 and S2. Show this contingency before the meeting. Do not promise complete identification in every case or add hours from unavailable people.

If the committee accepts an option feasible for all still-possible states, it can discuss B without collection or a less costly option supported by the obtained answer. After E=no exception, C is feasible in S1 and S2 and cheaper than B, although A could be preferable if S1 is actual.

Distinguish that commitment from proof of the optimal option for the actual state. There is no universal postponement rule here: the decision depends on mandate and the explicit information requirement.

Separate effort, cost and usefulness of information

H’s two hours and E’s three hours are collection effort; 20, 25 and 30 units are option costs. Do not subtract them without an agreed conversion and boundary. Information may reduce the chosen option’s cost, but the benefit depends on the answer.

H=short permits considering A instead of B, a ten-unit difference; H=long alone leaves B as the only option feasible for remaining states. This does not make H useless to all analysis: it narrowed states and may guide the next observation. For this decision, describe effects by branch.

Without probabilities, no numerical expected gain is defined. If an absolute 28-unit limit applies, B no longer fits. E=exception identifies S3, where none of the supplied options simultaneously meets feasibility and budget.

Collection exposed an impossibility among options, not a failure to conceal by changing data. Record the gap and route assessment of new options or conditions to the responsible owner.

Handle incomplete and contradictory results

The model assumes only S1, S2 and S3. If one observation confirms a short horizon and another confirms the exception is required, no row satisfies both answers. Do not automatically choose A or B or interpret an empty set as no requirements.

Answers, scope or model assumptions are incompatible. Check source version, population, meaning and authority; the state list may need revision. This differs from an inconclusive answer, which excludes no states.

It also differs from discovering a state for which no option fits budget: the need may then be clear and the problem lie in the alternatives. Plan distinct records for these situations so the committee can distinguish missing knowledge, an inconsistent analysis basis and insufficient options.

A spreadsheet may correctly calculate an empty intersection while knowing nothing about what caused the contradiction. The analyst’s communication should retain that boundary, with an owner and next investigative action. Mechanical consistency checks support judgment but do not supply missing facts or revise stakeholder commitments.

Turn reasoning into an analysis plan

Prepare a short plan covering the decision, conditions, still-possible states, proposed observations, owners, effort, availability, stopping rule and contingencies. In the example, H then E may suit the situation if H’s source is available and the deadline supports the five-hour branch.

Otherwise communicate which outcomes can be resolved and which need a conditional recommendation or another authorized step. Define the deliverable: a confirmed answer with scope and version, an uncertainty or inconsistency record, and a comparison applicable to remaining states.

Avoid measuring analysis success only through meeting or document counts. Completed but inconclusive collection does not authorize marking uncertainty resolved, though it may reveal a need for another method. Use the exercises to explain the plan to Business and APS in ordinary language.

For example: “If the first answer is long, we need another three hours to distinguish the remaining options.” All workshop times, states and costs are fictional. The model makes information commitments checkable without choosing the decision for its owner.

Common pitfalls

Promise the shortest branch; invent probabilities; treat inconclusive answers as absence of need; mix hours and money; treat an empty state set as approval.