Start with the job the answer needs to do
Suppose a team wants to understand why fewer people finish an online application. It could ask, “What proportion of applicants complete each step?” “Why do applicants leave before submitting?” or “Which applicants are likely to leave?” These questions concern the same process, but a useful answer to one would not necessarily answer the others.
The first question is descriptive: it asks what is happening. The second is explanatory: it asks what accounts for an outcome. The third is predictive: it asks what might happen in a future case. Before choosing a method or gathering data, decide which job the answer must do. Otherwise, it is easy to produce a careful description when the decision actually requires an explanation, or to offer a plausible explanation when someone needs a forecast.
Use descriptive questions to establish a clear picture
A descriptive question asks about the characteristics, frequency, distribution, or sequence of something. Examples include “Which steps take applicants the longest?” and “How do completion rates differ across application types?” These questions do not need to claim that one factor causes another. Their value lies in making a situation specific enough to discuss and investigate further.
Good descriptive questions name what will be observed and the boundaries of the observation. “How often do people abandon the form?” leaves open which form, which people, and what counts as abandonment. “Among applicants who start the housing form, how many reach the final review screen before leaving?” gives the researcher clearer terms to define and check. The wording still needs care: a person who pauses and returns later might not belong in the same group as someone who never returns.
Description can be the right endpoint, not merely a preliminary step. A team deciding which instructions to clarify may first need to know where people encounter difficulty. If your example involves planning information for an audience, Content Strategy · Bolster Digital offers a related area to browse for more questions about what people need to understand.
Use explanatory questions to investigate possible reasons
An explanatory question asks why a pattern occurs or how a process produces an outcome. “Why do applicants leave at the document-upload step?” invites investigation of possible factors: unclear instructions, file requirements, interruptions, or something else. A count of departures can locate the problem, but it cannot establish the reason on its own.
Phrase an explanatory question so that more than one answer remains possible. “How does confusing upload guidance cause applicants to leave?” assumes both that the guidance is confusing and that it causes departures. “What factors contribute to applicants leaving at the upload step?” keeps those possibilities open. The next task is to consider what evidence could distinguish competing explanations, such as observations of the process, applicants’ accounts, or comparisons between relevant groups.
Be cautious with causal language. If people who use a phone leave more often than people who use a computer, device type may be related to completion without being the cause. The groups may also differ in their circumstances or in the tasks they are trying to complete. A useful explanatory question helps you look for alternatives rather than treating the first association as the answer.
Use predictive questions to anticipate future cases
A predictive question asks what outcome is likely under specified conditions. For example: “Given what is known when an applicant starts, which applications are more likely to remain unfinished?” Prediction can help a team decide where to offer additional support. It does not, by itself, explain why an application remains unfinished. A factor can help anticipate an outcome even when its role in producing that outcome is uncertain.
Specify when the prediction would be made and what information would be available then. A question about identifying unfinished applications at the start cannot rely on information recorded only after an applicant leaves. Also define the outcome carefully: “unfinished after one session” and “never submitted” are different outcomes. If you are interested in the measurement choices behind such questions, Digital Measurement · Bolster Digital is a related topic to browse.
One project can include all three types of question, but keep their answers separate. First describe where applicants leave, then investigate possible reasons, and finally ask whether an early signal could help identify who might need assistance. To check your own wording, finish this sentence: “An answer would let me describe a pattern, explain a reason, or anticipate an outcome.” If you cannot choose one, split the question or decide which answer your immediate decision requires.
