A better research result usually starts with a better question

A sales leader asks, “Which companies should we target?” It sounds like a research assignment. It is not—at least not yet.

Does “company” mean a corporate parent, a legal entity, an operating location, or a buyer account? Which territory matters? What makes a prospect worth targeting? Is the goal a long-term account list, next quarter’s outreach plan, or a decision about where to add sales coverage?

Without those answers, researchers can still produce a polished list. The problem is that the list may answer a different question from the one the team needed to make a decision.

Better research starts by treating the initial question as a design tool, not a search prompt. Before collecting records, define what an answer must mean and what someone will do with it.

Begin with the decision

A useful first question is: What decision could change after we know the answer?

That separates a topic from a research question. “We need to understand our competitors” is a topic. “Which competitors introduced products in our defined category during the past 12 months, and does that affect our launch timing?” gives the work a purpose.

This does not mean every request needs a formal study. A quick territory scan and an investment decision deserve different levels of rigor. But both benefit from clarity about the action at stake, the person who owns it, and the deadline.

If a team cannot say what it would do differently after receiving the results, it may need problem framing before it needs more data.

Write a short question brief

A one-page brief can prevent weeks of well-executed work aimed at the wrong target. It should capture a few essentials:

• Decision and owner: What choice is being made, by whom, and when?
• Population and unit: What does one row or one case represent—parent company, location, product, account, or transaction?
• Measure: What observable fact will stand for terms such as growth, active, large, or high-value?
• Time frame: What dates define the relevant period?
• Comparison or threshold: Compared with what, or above what cutoff?
• Evidence standard: How complete, current, and corroborated must the evidence be for this decision?
• Output: What form will make the answer usable: a ranked list, count, map, memo, or set of flagged candidates?

The U.S. Census Bureau’s data-quality guidance takes a similar disciplined approach to data requests. Its materials ask requestors to identify the entity level, reference period, coverage, and variable definitions needed. The goal is shared understanding of intended use and limitations, not paperwork for its own sake.

See how the question changes the work

Consider the request: “Find fast-growing manufacturers in the Northeast.”

It has a recognizable topic, but it leaves too much open. “Manufacturers” could refer to firms, legal entities, or individual plants. “Northeast” may mean a customary regional label or a specified set of states. “Fast-growing” could mean revenue, employment, facility footprint, hiring activity, or something else.

A decision-ready version might read:

> For Q1 territory planning, which manufacturing establishments with 50 to 500 employees in the specified states show publicly evidenced employment or facility expansion between January 1 and December 31 of the defined year? Provide the evidence date and source for each candidate, and flag records where establishment-level evidence cannot be confirmed.

Every added phrase changes the research plan. “Establishments” points to locations rather than corporate headquarters. An employee range sets an inclusion rule. “Employment or facility expansion” defines the signals being sought. A date window rules out stale events. Requiring evidence dates and flags makes uncertainty visible rather than burying it in a final spreadsheet.

That is not needless precision. It is a way to make sure the eventual answer can support territory planning.

Define labels before they define the result

Business language often carries hidden measurement choices. Ask “Which companies grew?” and different researchers may reasonably return different answers.

One may use annual revenue reported by public companies. Another may use location-level hiring or expansion announcements. Both could be careful researchers. They would still be studying different units, periods, and measures.

The same problem appears with labels such as “active,” “competitive,” and “high-value.” Before looking for data, decide what field or observable evidence will represent the concept. A label without an operational definition invites the available data to decide the question after the fact.

Also distinguish descriptive, comparative, and causal questions. “How many locations opened?” is descriptive. “Which region had more openings?” requires a defined comparison. “Did our campaign cause more openings?” requires a stronger design than observing two events that occurred around the same time.

Government Accountability Office guidance on evaluation design makes the broader point: useful planning begins with decision-makers’ needs, and different questions require different methods. A comparison is not, by itself, proof of causation.

Match the evidence to the consequence

A question brief should include an honest statement of acceptable uncertainty. A preliminary prospecting list may be useful despite incomplete coverage, provided its limits are clear. A pricing, compliance, or investment decision may require more current and more complete corroboration.

A precise question does not make research infallible. Evidence may be unavailable, incomplete, outdated, or poorly suited to the measure. Scope can also be too narrow if it excludes cases that matter to the decision.

But a defined question makes those limitations easier to see, discuss, and manage before a team has committed time to gathering information.

A practical test is this: if two competent researchers could interpret a request differently and both return plausible but different results, the question needs another round of definition.

Better research does not begin when someone opens a spreadsheet. It begins when a team can say what would count as an answer—and what it plans to do with one.