Framing Problems and Decisions

Review questions

Integrated summary: Framing Problems and Decisions

Good decision-making begins before choosing a tool. The first question is what we are trying to understand or choose, and what was reasonably available at the time. That is why this subject separates decision-process quality from outcome quality. A reasonable process can be followed by a bad outcome because uncertainty remains, while a weak process can receive a good outcome by luck. Outcomes matter because they add evidence, but they do not erase what was known or unknown at the time. A short decision record preserves the decision, desired state, alternatives, evidence, assumptions and uncertainty, rationale and review trigger, making later learning less vulnerable to hindsight.

Next comes problem framing. “We need a new system” jumps to an intervention before defining the gap. A stronger frame states the current condition, desired condition, scope, constraints and affected stakeholders. In the teaching case, rework in web and email requests rose from 8% to 17% over six weeks, with a goal of 10% or less while preserving response time. That is a testable gap. Saying “employees are slow” instead assumes a cause about people without evidence. A new platform, training or process change may later become alternatives, but they do not belong in the problem statement as facts.

Initial scope does not have to include everything. Start where the signal appears and define clear expansion triggers. If the rise appears only in web and email, those channels can be investigated first while phone remains temporarily outside scope; broaden the inquiry if a similar pattern or shared cause appears. The desired state describes improvement, not a solution name: “rework at 10% or less while preserving response time” is an outcome; “buy a platform” is a possible intervention.

Then examine information quality. In another case, 420 requests arrived in one month and 84 were after hours; 84÷420=20%. That is a valid historical observation within the month and measurement definition, but it does not mean next quarter will also be 20%, that every evening request is urgent, or that extended hours are the best choice. “The share will remain 20%” is an assumption; an 18-person survey is an observation from a small sample; and the effect of extended hours on satisfaction is unknown until measured. Labeling observations, estimates, assumptions and unknowns prevents evidence from being given more strength than it deserves.

Uncertainty does not require paralysis. Use ranges, scenarios or clear descriptions of limits. Seek additional information when there is a reasonable chance that it will change the choice, safeguard or timing, and when collection cost and time are proportionate to the importance of the decision. If the same choice remains suitable across a reasonable range, extra precision may have little value. If a small change flips the choice, the information is decision-sensitive and deserves more attention.

Finally, do not confuse a symptom with a cause. A rise in no-shows from 9% to 16% is a symptom. “The reminder redesign caused the rise” is a causal hypothesis, as are longer booking lead time, a changed customer mix, or more evening appointments. A useful hypothesis states what we would expect to observe if it were true and what would weaken it. In the case, when appointments were split by lead time, no-show rates were 10% then 11% for 1–7 days, and 15% then 16% for 8–14 days. The within-band rise is one percentage point, while the lead-time bands differ more. This weakens “message design alone” and makes lead time worth stronger investigation, but it does not prove a single cause.

Cause-and-effect diagrams and repeated “why?” questions generate hypotheses; they do not prove them. Comparison, measurement and proportionate small checks increase or decrease confidence. If evening errors rise among the same employees who also work mornings, while the validation system slows in the evening, the training-difference hypothesis weakens and the system-condition hypothesis strengthens, while further verification is still needed. The four lessons therefore connect as one process: preserve decision logic, frame the gap without jumping to a solution, weigh information with its limits, and move from symptoms to testable hypotheses before selecting a remedy.

Review questions

Test your understanding with these questions. Marking and saved results are available with a free account.

Question 1A team selected a vendor after comparing suitable alternatives and evidence. A rare event that could not reasonably have been anticipated later stopped the service. Which description is most accurate? The process was weak because the outcome was bad. The process may have been strong despite the bad outcome. The decision was necessarily correct because alternatives were compared. There is no value in reviewing the outcome because the event was rare.
Question 2Which statement best expresses the difference between a decision and an outcome? The decision is what actually happened, and the outcome is what was planned. Decision and outcome are synonyms when the objective is clear. A decision is a choice made at a point in time; the outcome emerges later and may be affected by other factors. A decision cannot be evaluated until the final outcome appears.
Question 3A manager chose the first suggestion heard, without comparison or verification, and it happened to succeed. What is the best lesson? Success proves the process was strong. The process was weak despite the good outcome. Nothing can be learned because the outcome was good. The first suggestion should always be chosen if it worked once.
Question 4Which two elements make a decision record especially useful for later review? The decision-maker’s name and the presentation color. Only the final outcome and the manager’s comment. The longest possible description of every meeting before the decision. Assumptions and sources of uncertainty, plus review triggers.
Question 5A team says, “Our problem is that we need a new platform.” Which reframing is the best starting point? Request-processing errors have risen above the desired level in two channels; we want to reduce them within the response-time constraint before comparing interventions. The current platform is old, so it must be replaced. Employees are not using technology correctly. We need a platform with more features than competitors.
Question 6Which element should not be presented as a fact inside a problem definition before verification? Rework rising from 8% to 17% over a defined period. The goal is to reach 10% or less. Poor employee training is the confirmed cause of the increase. The initial scope includes web and email.
Question 7The 8%→17% rise appears in web and email only, and there is no phone data yet. What is the most proportionate initial scope? Every organizational channel and every branch from day one. Web and email first, with a clear trigger to expand if a similar phone signal appears. Phone only because it has not yet been examined. Do not define any scope until the final cause is known.
Question 8Which statement describes a desired state rather than an intervention? Buy a new request-management platform. Add more people to the service team. Provide additional training to everyone. Reduce rework to 10% or less while preserving the current response time.
Question 9In one month, 84 of 420 requests arrived after hours. Which statement is an assumption rather than an observation? 420 requests were recorded in the month. 84 requests were after hours. The after-hours share will remain 20% next quarter. 84÷420 equals 20% for the observed month.
Question 10A service center is considering extended hours. Which additional information is most decision-sensitive? The dashboard interface color. The share of evening requests that cannot be deferred without material service impact. The number of presentations the team produced last year. The name of the first employee who suggested the idea.
Question 11What is the most accurate way to state the 84-of-420 result? 20% of all future requests will be after hours. Extended hours are the solution because 20% is high. The small survey proves all customers prefer evening service. In the observed month, 20% of requests arrived after hours; whether that share continues is uncertain.
Question 12When is gathering additional information most justified? When the information could change the decision and its collection cost and time are proportionate to the decision’s importance. Whenever any additional data can be collected. Only after the final outcome appears. When we want to increase the report’s page count.
Question 13Which statement is a testable causal hypothesis rather than merely a symptom description? The no-show rate rose from 9% to 16%. No-shows are a major problem. If longer booking lead time increases the chance that plans change, no-shows should rise as lead time increases. We should change the reminder immediately.
Question 14After a reminder redesign, no-shows rose. A group not exposed to the new design showed a similar rise. What does this do to the hypothesis that the redesign is the main cause? It proves the hypothesis conclusively. It weakens the hypothesis because a similar pattern appeared without exposure to the redesign. It changes nothing because comparisons are not useful. It proves booking lead time is the only cause.
Question 15What is the correct use of a cause-and-effect diagram or repeated “why?” questions? Prove the first cause that appears in the diagram. Identify the person responsible for the problem. Generate possible hypotheses and then test them with suitable evidence or comparisons. Avoid data because the team knows the process.
Question 16Errors rise in the evening. The same employees work mornings and evenings, and the validation system slows at night. What is the most disciplined conclusion? Evening-team training is the confirmed cause. System slowdown is the only confirmed cause. The training-difference hypothesis weakens and the system-condition hypothesis strengthens, but further verification is needed before claiming causality. No hypothesis can be weighted at all without a full randomized experiment.