Question and evidence map · Lesson 1

From problem to question

7 min read

Lesson objectives

  • Turn a practical problem into a researchable question and documented evidence-search plan
Reading tools

Research starts with a practical problem, but a problem is not yet a research question. “Too many people leave” identifies neither the population, time frame nor outcome to be explained. Before selecting a survey or statistical test, identify the decision maker and the decision the evidence should inform. A request to count departures is descriptive. A request to examine how shift arrangements relate to departure is associational. A request to estimate the effect of changing shifts is causal and calls for a stronger design. Replacing “describe” with “prove an effect” changes the evidence required, not merely the title.

Build a question card with four fields:

  • the population to which the finding may apply
  • the phenomenon, exposure or intervention
  • a comparator if the question needs one
  • and an outcome with a measurement window

A qualitative question about workers' experiences does not require two comparison groups. When a comparison matters, specify it before collection. “Employees” is broad; “service-centre staff employed at the start of the study quarter” defines whom the records can follow. “Satisfaction” needs an operational definition. “Leaving employment within six months” is checkable if a trustworthy record exists. A precise definition does not make the record complete; test its limits separately.

Separate the study aim, question and hypothesis. An aim states the work to be done, such as estimating departures over a defined period. A question states what the data should answer. A hypothesis is a specific expectation that can be tested when the design permits; it is not mandatory for every exploratory or qualitative project. Specify the unit of analysis as well. Is one record a person, a shift or a team? A person appearing in several shifts does not become several independent people when calculating a departure rate. Name the start and end points and the possible data source so feasibility can be checked before a conclusion is promised.

Stress-test the question with three challenges. Can the available data distinguish the proposed outcome from a nearby one? A contract-end field may combine resignations and scheduled contract endings. Does the data frame cover the people to whom the result will be generalized? One centre does not support a claim about all centres. Does the design support the verb? Comparing people who chose flexible shifts with those who did not can show an association, while differences in prior circumstances offer competing explanations. Narrow the verb or plan how to address those differences. Refer any participant-risk or data-access issue to the responsible authority; a well-written question is no permission to collect data.

Make the card's completion criterion observable: population, analysis unit, phenomenon or variables, time window, allowed type of inference, proposed data source and at least one limitation. If a field remains uncertain, record an open decision with an owner and date rather than inventing an answer. A reviewer can then compare the eventual report with the intended inquiry. Clear reporting of purpose and procedures matters for that comparison.

A researchable question and a decision-ready question differ. You may estimate departures accurately while a manager needs to know whether to alter a particular policy. Keep “intended use” separate from “study finding” on the question card. If departures exceed an agreed threshold, the appropriate next step may be investigating reasons rather than immediately cancelling shifts. Set a minimum data requirement too: a register that cannot distinguish resignation from planned contract ending cannot answer a resignation question, however correct its arithmetic. This requirement drives feasibility assessment before collection.

A team may ask how many people leave, why they leave and whether a new policy would prevent it. These are three questions. The first needs reliable description, the second may need qualitative understanding or analysis of factors, and the third needs an effect comparison. Prioritize and connect them. A descriptive pattern may guide interview selection; interviews may reveal that “flexibility” means different things and change the intervention definition before effect testing. An answer to one question is not evidence for another without a methods bridge.

Check feasibility as well: authorized access, sufficient follow-up time for the outcome, enough events to describe without identifying individuals, and an owner prepared to act on the type of evidence available. If not, narrow scope or propose a small feasibility study instead of promising a decisive report. Express assumptions so they can be tested, such as completeness of departure codes, rather than “the data are accurate.” Such precision prevents a later dispute about what the study actually meant.

Lesson summary

From problem to question: define population, phenomenon or intervention, comparator where applicable, outcome and feasible scope.